The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Techniques of Preparing Major BLS Statistical Series [From the Monthly Labor Review of the Bureau of Labor Statistics, Sep tember, October, and November 1949, and January, February, March, and April 1950 issues] Bulletin No. 993 UNITED STATES DEPARTMENT OF LABOR Maurice J. Tobin, Secretary BUREAU OF LABOR STATISTICS Ewan Clague, C om m ission er For sale by the Superintendent of Documents, U. S. Government Printing Office, Washington 25, D. C, Price 40 cents Letter of Transmittal U nited States D epartment op L abor, B ureau of L abor Statistics, Washington, D. C., June 21, 1950. The Secretary of L abor: I have the honor to transmit herewith technical descriptions of the methods used by the Bureau of Labor Statistics in the preparation of its major statistical series. In all, 13 technical notes are included, each of which indicates the sources of information, the method of collection, the limitations of the series, and the statistical procedures utilized in the computation of the periodic economic measures. These technical notes are the work of members of the staffs of the appropriate Bureau divisions. Bruce M. Fowler of the Construction Division of the Office of Domestic Commerce, U. S. Department of Commerce, is co-author of note No. X . E wan C lague, Commissioner. Hon. M aurice J. T obin, Secretary of Labor. m Preface The Bureau of Labor Statistics considers that it has an obligation to inform users of its statistics concerning methods used in the preparation of individual series. Therefore, a number of technical notes covering the major statistical series make up this volume. They have already appeared in various issues of the Monthly Labor Review and are collected here for the convenience of the users of BLS data. Professor Simon Kuznets, a recent president of the American Statistical Association, speaking before the members of that Association, made the following statement on the explanation of techniques: “ Indeed, as a rule, collectors and publishers of primary data do not deem it their obligation to accompany a series by a detailed description of how it was obtained; and users also, for the most part, tend to accept a series, particularly one issued by a governmental agency, at its face value without inquiring into its reliability. If this impression is correct, there is surely room for much additional work. It may legitimately be urged that compilers and publishers of series give full details on methods of collec tion, compilation, classification, and adjustment; that various compendia of basic series supply descrip tions of their origin as an indispensable part of the information; that users exercise their right to be in formed about the derivation of the series offered them; and that authors of textbooks on statistics become cognizant of the problem and cease confining their attention to tools of analysis while forgetting the elementary question of the character of the primary and derived data.” In an attempt to fulfill such requirements, this bulletin deals separately with (a) the scope of the respective surveys, (b) definitions and concepts used in the collection of these statistical data, (c) the sources of these data and their limitations, and (d) the methods used in the calculations of the various measures, such as indexes, averages, rates, etc. These discussions should enable persons employing the statistics to make more effective use of them, and to limit their use to the situations where applicable. These technical notes are written primarily from the point of view of the consumer and not the pro ducer of the data. Whenever possible, therefore, the notes have been written in narrative form. The occasional algebraic formulas merely supplement the text, which is complete and understandable in itself. Although written primarily for the consumer, these notes should prove of inestimable aid to both students and university instructors, particularly in courses in the field of labor economics and statistics. They also should furnish to the student, to the instructor, and to the writer of textbooks an appreciation of the problems faced by the producer of these data. As indicated by Professor Kuznets, present-day textbooks do not always give a balanced view of the statistician’s problems in conducting specific surveys. The major cost of conducting a survey, mainly that of collection in its broader sense, is ignored by present-day authors. It is hoped that these notes may furnish to the writers of statistical textbooks material which will result in more adequate coverage than heretofore of the problems of collection encountered in the statistics. Contents Page I. Construction of consumers’ price index____________________________________________________ Limitations of the C PI__________________________________________________________________ Methods of pricing______________________________________________________________________ Sources of price quotations______________________________________________________________ Calculation procedures__________________________________________________________________ Relative importance of items____________________________________________________________ II. Collection and compilation of work stoppage statistics___________________________________ Limitations of the series________________________________________________________________ Survey methods and sources_____________________________________________________________ Calculation procedures__________________________________________________________________ III. Estimating national housing volume_____________________________________________________ Limitations of the series_________________________________________________________________ Methods and sources of survey__________________________________________________________ Calculation procedures__________________________________________________________________ Tests of reliability______________________________________________________________________ IV. Measurement of labor turn-over_________________________________________________________ Limitations of the series_________________________________________________________________ Survey methods and sources_____________________________________________________________ Calculation of turn-over rates___________________________________________________________ V. Compiling monthly and weekly wholesale price indexes____________________________________ Comprehensive monthly index__________________________________________________________ Weekly index___________________________________________________________________________ VI. Preparation of union scales of wages and hours series____________________________________ Limitations of the series_________________________________________________________________ Study methods and sources______________________________________________________________ Computation procedures________________________________________________________________ VII. Measurement of industrial employment_________________________________________________ Limitations of the data__________________________________________________________________ Survey sources and methods____________________________________________________________ Calculation procedures__________________________________________________________________ VIII. Calculating hours and earnings of workers in industry--------------------------------------------------Survey sources and methods____________________________________________________________ Calculation procedures__________________________________________________________________ Interpretation and limitations___________________________________________________________ IX . Measurement of unit man-hour requirements___________________________________________ Limitations of productivity measures------------------------------------------------------------------------------Industry indexes: secondary source data________________________________________________ Industry reports: field-collected data------------------------------------------------------------------------------X . Estimating expenditures for new construction-------------------------------------------------------------------Limitations of the series_________________________________________________________________ Sources and general estimating methods--------------------------------------------------------------------------Adjustment procedures__________________________________________________________________ X I. Compilation of industrial-injury statistics-----------------------------------------------------------------------Limitations of the series_________________________________________________________________ Sources and methods of surveys-------------------------------------------------------------------------------------Computation procedures------------------------------------------------------------------------------------------------X II. Occupational wages: establishment sampling-----------------------------------------------------------------Collection of information________________________________________________________________ General sampling procedures____________________________________________________________ Limitations of sampling theory---------------------------------------------------------------------------------------X III. Occupational wages: conduct of surveys----------------------------------------------------------------------Limitations of surveys---------------------------------------------------------------------------------------------------Study methods and sources______________________________________________________________ Compilation procedures-------------------------------------------------------------------------------------------------- vn 1 1 2 3 4 5 8 8 9 9 13 14 14 15 17 20 21 22 22 25 25 28 29 30 30 31 33 33 34 34 37 37 38 40 42 43 43 46 50 51 52 54 59 00 00 02 64 04 05 08 70 71 71 72 Techniques of Preparing Major BLS Statistical Series I. Construction of Consumers’ Price Index Changes in prices paid for goods and services usually bought by moderate-income families in large urban centers are reflected by the Consum ers’ Price Index,1 which the U. S. Labor Depart ment’s Bureau of Labor Statistics issues from month to month. Such changes are measured by the rate of price movement of a representative list of items of specified quality. The components of the index and the weights assigned to each of them remain constant for considerable periods. The rate of price change is one of the most impor tant factors affecting the cost of living, and over short spans of time, the Bureau’s index gives an acceptable approximation of changes in the cost of living for urban workers. The index was initiated during World War I, when prices rose rapidly, for use in wage negotia tions, particularly in shipbuilding centers. Cover age was gradually extended to include industrial cities throughout the country, and estimates of Nation-wide changes in living costs were published at intervals, beginning in October 1919. Begular publication was established in February 1921. Weights used in these early indexes were based on surveys of family expenditures conducted during i The title, Consumers’ Price Index for Moderate Income Families in Large Cities, was adopted in 1945. Previously this index had been precisely designated, Changes in the Cost of Goods and Services Purchased by Wage Earners and Lower-Salaried Clerical Workers in 1934-36. In popular usage, this title was later shortened to Cost-of-Living Index. The latter designation gave rise to some misunderstanding of the scope of the series, and therefore the current term, Consumers’ Price Index, was introduced. 884616—50----- 2 1917-19. In the fall of 1935, the Bureau in troduced improved methods of calculating the index and in 1940 completed revision of the weights to correspond with 1934-36 family ex penditure patterns, as determined by another extensive study of family consumption. In addition to its long-term use as a basis for wage adjustments, the index is used as a measure of changes in the purchasing power of the dollar, and as a guide to broad economic policy. Limitations of the CPI Amounts that urban families spend for living are not shown in the index. To develop such measures, information reflecting changes in in come and in the manner of living would be re quired, as well as statistics of price changes for consumer goods and services. The index does not represent price changes affecting other population groups such as single individuals, families living in rural areas, families of business and professional men, and families deriving a major portion of their income from sources other than their earnings, whose buying habits may differ radically from those of moder ate income urban families. Nor does it take into account changes brought about by migration of families to large cities from rural communities or from other cities. Individual city indexes may not be used to compare living costs between cities. A higher index for one city than for another is no indication that prices are higher in that city than in the other. 1 2 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES It means only that prices have advanced more rapidly in one city than in the other subsequent to the base period. For example, assume that the dollar cost of a specific list of goods was $1,100 in City A and $750 in City B during the base period. Since these costs are taken as 100 for each city, an increase to $1,250 in City A would result in an index of 113.6, but an increase to $1,000 in City B would produce an index of 133.3. Thus, even though City B has a higher index, the level of prices is still lower than in City A. Basis jor Selection of Items. A study made by the Bureau in 1934-36 is the basis of the selection of items and determination of weights for the index. This survey covered the incomes and expenditures of about 14,500 families of wage earners and lower-salaried clerical workers whose average income was $1,524 a year at that time. Expenditures for food, apparel, rent, fuel, utilities, housefurnishings, and miscellaneous goods and services purchased were ascertained in detail. The items selected to represent all goods and services purchased, on the basis of the 1934-36 study, were those which were relatively important in family spending, which had distinctive price movements, and which were highly representative of larger groups of related items. Specifications of items to be priced were written to describe qualities, the retail-store prices of which would correspond with prices paid by families included in the survey. The sample of items priced con sists of 49 foods, 58 articles of clothing, 10 fuels, 23 housefurnishings, 49 miscellaneous goods and services, and rent. In order that the items selected for pricing would represent all goods and services bought by moderate-income families, expenditures for the items not priced were combined with those for the selected items. The weight for a priced item includes weights for similar items known to have the same price movement and a proportionate share of the weights for other items in the same consumption group for which price movements cannot be imputed directly to a specific article. Methods of Pricing Since all of the more than 1,400 different articles and services bought by wage-earners’ families need not be priced to determine changes in average prices paid, the Bureau prices about 190 of them. (For a listing of these items, see table 1, p. 6.) Two or more qualities of many of the 190 articles are covered and consequently the aggregate num ber of articles and services included is 270. Specifications of Goods To Be Priced. The Bureau attempts to price goods of constant quality from period to period, so that the index will reflect price changes only. To accomplish this, rents are compared on identical units from period to period; for other groups, detailed specifications have been written for the items for which prices are obtained. Each specification is for an article that experts in industry and trade judge to be most frequently purchased in the price lines in which wage earners and clerical workers concentrated their purchases in 1934-36. The specification for a man’s work shirt is typical. Shirt, work, cotton chambray: 3.90 yards per pound before sanforizing, about 3.60 yards per pound after sanforizing, based on 36-inch fabric, sanforized shrunk; Full cut, clean workmanship, good quality buttons, collar interlined with chambray or equal grade of fabric, continuous nonrip sleeve facing, double- or triple-stitched seam, 2 plain pockets with or without flap, 30 to 31 yards per dozen based on 36-inch fabric and neckband size scale 14 to 17 inches. (Specify whether double- or triple-stitched). In addition to the detailed specifications, records of brands, lot number or grade (where available), and other identifying information are also supplied to the Bureau’s representatives who collect the prices. Prices are obtained for identical articles as long as they are available in retail stores. When the Bureau’s representatives can no longer obtain prices for a given article, they must substitute another. Substitutions are of two types: (1) Substitu tion of another article which is adequately de scribed by the same specification, and (2) substi tution of an article serving the same purpose, but not of the same quality, and described by a new specification. In the first type, any difference in price between the old and new article is shown as a price change in the index calculation. For example, if one brand of men’s shirts is no longer available and another brand of substantially the same quality CONSUMERS’ is substituted, the difference in price is allowed to affect the level of the index. In the second type of substitution, the level of the index is not affected, for the new article is introduced by a linking proc ess. An example of this type of substitution is the replacement of silk hose by rayon hose during World War II. Substitute specifications are al ways made to adhere as closely as possible to those supplanted, i. e., with respect to utility of goods, materials, designs, and price movements. Methods and Cycle of Pricing.— Prices for the Bu reau’s index are those actually charged customers in retail stores. Part-time Bureau agents (usually housewives, school teachers, and ex-Government employees) collect food prices monthly in their communities, according to the written specifica tions. In food stores, prices are posted in full view of the customer and can be written down by these agents. The prices are checked if necessary with the proprietors or managers. Most of the price collection for other groups is done by full-time Bureau representatives who are specially trained and who are guided by the speci fications described. These agents obtain the price quotations for most apparel, housefurnishings, and miscellaneous items, in personal interviews with store managers and buyers. They collect rent information, by personal visit once a year, directly from a sample of renting families in each city. For subsequent quarters the rent collection is done by mail. A few prices, such as for fuel, are ob tained directly from dealers, by questionnaire. Electric-power rates are obtained from the Federal Power Commission. Food prices are collected monthly in 56 cities 2 during the first 3 days of the week containing the 15th of the month; prices of fuels in effect on the 15th of the month are obtained in 34 cities monthly; apparel, housefurnishings, and miscel laneous items are priced over a longer period (carried on as near the 15th of the month as pos sible), in 10 key cities monthly and in the remain ing 24 cities according to a rotating quarterly cycle, with 8 cities surveyed each month in addition to the 10 key cities. This cycle was carefully deter mined on the basis of historical price movements * These 56 cities account for about 60 percent of the total population in cities over 50,000 population in the United States. p r ic e 3 in d e x for individual cities, in order to approximate the national trend as closely as possible, and was coordinated with the rent cycle. Every month Birmingham Boston Chicago Cincinnati Detroit Houston Los Angeles New York Philadelphia Pittsburgh January, April, Julyt and October Buffalo Denver Indianapolis Kansas City Manchester Portland (Oreg.) Richmond Savannah February, M ay, August, November Atlanta Cleveland Milwaukee New Orleans Norfolk Scranton Seattle Washington March, Junet September, December Baltimore Jacksonville Memphis Minneapolis Mobile Portland (Maine) St. Louis San Francisco The quarterly cycle for pricing rents3 was developed from 3 groups of cities, each of which represents a good cross-section of the 34 large cities included in the index. Rents are obtained for each of these city groups quarterly as follows: J anuary, Aprils February, M ayf Julyt and Oc August, and tober November Buffalo Atlanta Denver Birmingham Cleveland Detroit Indianapolis Houston Kansas City Los Angeles Manchester Milwaukee New York New Orleans Pittsburgh Norfolk Portland (Oreg.) Philadelphia Scranton Richmond Savannah Seattle Washington March, June, Se tember, ai December Baltimore Boston Chicago Cincinnati Jacksonville Memphis Minneapolis Mobile Portland (Maine) St. Louis San Francisco Sources of Price Quotations Quotations are obtained from retail stores and service establishments that wage earners and lower-salaried workers patronize widely. Insofar as possible, scientific sampling procedures are em ployed in selecting retail outlets from which prices * For methods used in estimating the national rent index, see The Rent Index: Part 2, Methodology of Measurement, Monthly Labor Review, January 1949 (reprinted as Serial No. R. 1947). 4 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES are to be obtained; if necessary, local authorities are consulted. For food price collection, independent outlets are chosen by random sampling within geographic areas of the city. Representation of the indi vidual types of stores is based on the relation of their sales to the total food store sales in the city. All important grocery chains within the city’s corporate limits are included.4 In all, 1,129 inde pendent grocery stores and markets, 208 chain organizations (having 8,640 stores), 152 dairies, and 340 bakeries are covered. For the pricing of other items included in the CPI,5 outlets were selected by the Bureau on the basis of size, type of operation, quality of com modities sold or services rendered, location, and clientele. Representation is given to department and specialty stores, to national, sectional, and local chains, and to independent stores. Cashand-carry outlets and those granting regular credit and delivery service or installment credit are covered. In cities having stores operated by mail-order houses, such outlets are represented. Laundry and dry-cleaning establishments, beauty and barber shops, automobile-repair shops, ap pliance stores, doctors, dentists, etc., are also in cluded. Apparel, housefurnishings, and miscel laneous goods and services prices are obtained from 3,500 stores and service establishments. Fuel prices are reported by 300 fuel dealers and utility companies. A comprehensive housing survey in each city is the basis for the master dwelling sample from which rents are collected. All city blocks are stratified by size, in such a survey. Rents are collected quarterly from subsamples of rental dwellings selected at random from the master sample. Both the master sample and the sub samples of rental dwellings cover suburban areas which are an integral part of the city’s housing market. Rents are supplied by 600 to 3,000 tenants in a city, depending upon population of the city surveyed. 4 The number of food quotations obtained in a city may vary considerably. Fewer quotations are necessary for staples, such as sugar and bread, the prices of which differ little from store to store and from time to time than for perish ables, such as lettuce and round steak, which may vary considerably in a few days and from store to store at a given time. 8 For groups other than food, prices for each item are obtained from at least 5 stores or service establishments in New York City and at least 4 stores in the other cities surveyed. Few stores can supply prices for all of the articles in a commodity grouping. It is usually necessary to visit at least 10 stores in order to obtain a minimum of 4 quotations for each article priced in the clothing group. Calculation Procedures The current base period, 1935-39, was adopted in 1940 on the recommendation of the Division of Statistical Standards and indexes previously published on a 1913 = 100 base were linked to the new series. Formula for Index. The index is based on the formula of Laspeyres: p _ Z)goP i " < -S<Zo2>o where the (q0)’s are the average quantities of each item used by families in the wage earner and clerical groups in the base period, the (Po/s are the prices for these items in the base period, and the (Pf)’s the prices in a current period. In this form, the formula is used only in calculating the food index. For groups other than food, the Bureau calcu lates the index on a variation of this formula, as a weighted average of price relatives (ratio of the price in one period to that in the preceding period) for each item. where the (q0P i-i ) ’s are the “ cost weights” in the pre vious period and the are the price relatives for each item, and Bi-\ is the index for the previous period. The two formula forms yield identical results. Steps in Calculation of Different Indexes. Average prices of foods in the 56 cities surveyed each month are calculated separately for chain and independent stores and combined according to relative sales volume of the two types of stores. Prices are then multiplied by fixed quantity weights to give current value factors. For each city, the food index is calculated as a fixed-base weighted aggregative index. For those 11 or 12 cities in which rents are sur veyed in a given month, the rents in the current period are compared with those of identical units in the previous quarter, after adjustments have been made for any changes in the facilities in consum ers’ eluded in the rentals. The relative change is based on the sum of the rental rates, and this relative is applied to the previous index to obtain the index for the current date. Thus the resultant figure is a simple link-relative index. Weighting is implicit in the sample selection. For the remaining four groups of commodities and services, the indexes are calculated as weighted averages of price relatives, as indicated above. Prices used in the index for. a given specification are simple averages of the quotations in identical outlets from period to period. In each group, the sum of the value factors or “ cost weights” (price times weight) is related to the sum of the value factors for the previous period and this weighted relative is multiplied by the index for the previous period to obtain the index for the current period. The individual city indexes for all items are then computed on the basis of group totals. This entails adding the value factors for the six major groups of goods and services and relating them to the aggregate value factors for the same city in the previous period and calculating the current index by the same method described above for the group indexes other than food. For those cities in which rents are not priced but other groups are, the procedure is to hold the value factors for rents constant between pricing dates. Then the “ all-items” index is computed in the same way as for cities in which all groups of items are priced. National indexes are calculated each month for all items and the six major groups on the basis of the cities surveyed, with estimates for unpriced cities. Each month the coverage is complete for the food group for which 56 cities are included in the national average and for the fuel, electricity, and refrigeration group which covers 34 cities. In the calculation of the group indexes for all cities combined, cost weights for individual cities are weighted by population, 56 cities for food, and 34 cities for other groups. The basis is the popu lation of the metropolitan area of the particular city and of other cities in the same region and size class. For those cities in which group aggregates for rent, apparel, housefurnishings, and miscellaneous goods have not been calculated in a given month, the total value factors for these groups are esti mated for purposes of the national index on the p r ic e in d e x 5 basis of the price trend in a city which has demon strated similar price movements in earlier periods. Special Adjustments. During World War II, it was necessary to make some temporary adjust ments in the weights of items affected by rationing and shortages. Weights for items which were not available for civilian consumption, and for which no substitute could be readily priced, were taken out of the group indexes, and assigned to a group of unpriced items. Prices of these items were assumed to have the same movement as the aver age of all priced items. When these goods were again available their weights were reintroduced into the group indexes with an adjustment for the difference between the actual and estimated price movement while the goods were off the market. Adjustments of this type were made for automo biles, tires, tubes, refrigerators, and other con sumer durable goods. To reflect the effect of gasoline rationing, part of the weight for gasoline was assigned temporarily to public transportation and automobile repairs. Relative Importance of Items 8 To meet public demand, the Bureau once a year calculates the relative importance of the individual items included in the index. These relative im portance figures should not be confused with the quantity weights, which for the most part have been held constant since the base period. The relative importance figures are percentage dis tributions of the value factors which result in the index calculation when 1934-36 average family expenditures for groups of items are multiplied by price relatives that measure average price changes of the items in the group. It should be recognized that these percentage distributions change from period to period, according to the relative price changes for the individual items. All of the items priced for the CPI as of December 1948 and their relative importance within their respective groups and in the total are listed in the accompanying tabulation. The emphasis placed on each price for each city depends on the importance of that particular article in the actual spending of moderate-income families in that locality (as shown in the 1934-36* * For a more detailed discussion of relative importance, see “ Consumers’ Price Index: Relative Importance of Components,” Monthly Labor Review, August 1948 (reprinted as Serial No. R. 1933). 6 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES survey). A comprehensive revision of the index within the next few years is contemplated by the T able Bureau. It is to include establishment of new weights based on current expenditure data. 1.— C PI items and their relative importance in the major groups and in the total index, December 1948 Percentage distribu tion of index value factors in December 1948 Item Group total Cereals and bakery products. Cereals: Flour, w heat............. Com flakes................ Com meal................. R ice............................ Rolled oats................ Bakery products: Bread, w hite............ . Vanilla cookies.......... All items total 100.0 40.6 13.9 5.6 5 pounds. . .11 ounces.. — pound— .........do---.20 ounces.. 2.1 .5 .4 .3 .7 .8 .2 .2 .1 .3 ___pound— ----- do — 8.1 1.8 3.3 .7 Food. 32.8 M eats, poultry, and fish................................................... Beef: 4.7 Round steak.............................................. pound— 4.3 R ib roast.................................... do_____ 2.0 Chuck roast....................................................do__ 1.9 Hamburger.....................................................do__ 2.2 Veal, cutlets....................................................... do— Pork: 3.4 Chops..............................................................do__ 2.0 Bacon, sliced............................................ -do____ 2.3 Ham, whole....................................................do__ .4 Salt pork.........................................................do__ Lamb, leg.................................................... do______2.9 Poultry, roasting chickens..................................do__ 3.2 Fish: Fish (fresh frozen)................................ do2.2 1.3 Salmon, pink.................................16 ounce can.. Dairy products............................. .................................... . Butter...................................... -.....................pound.. Cheese.............................................. .............. —do___ M ilk, fresh (delivered)......................................quart. M ilk, fresh (grocery)............................................do__ M ilk, evaporated............................. 14J4 ounce can— 18.8 5.6 1.8 6.1 . 4.2 1.1 Item 13.3 1.9 1.7 .8 .8 .9 1.4 .8 .9 .2 1.2 1.3 .9 .5 7.6 2.3 .7 2.5 1.7 .4 Percentage distribu tion of index value factors in December 1948 Group total All items total Apparel............................. ...... .......................................... Wnnl Men’s: Overcoats............................. ................... Topcoats................................................... Suits.....................................................—. Trousers................................................... Sweaters................................................. . Women’s: Coats, heavy, fur-trim....................... Coats, heavy, plain............. .............. Coats, light, plain............................... Suits.................................................... Dresses................................................ Girls’ : Coats......................................................... Boys’ : Overcoats.................................................. Mackinaws............................................... Suits.......................................................... Slacks.................................... ................... Cotton.......................................................................... Men’s: Suits......................................................... Trousers................................................... Overalls.................................................... Shirts, w ork............................................ Shirts, business...................................... Pajamas.................................................... Shorts.............................................. ........ Undershirts................... ...................... Unionsuits............................................... Socks............................... ...... ................ Women’s: Dresses, street........................... ......... Dresses, house___ ______ __________ Girls’ : Dresses.................................... ................. Slips................................................... ...... Panties.............................. ....................... Anklets....................................... .............. Boys’ : Shirts, polo......... ..................................... Shirts, convertible collar. .............. ......... Shorts. ---------- ---------------------------------Yard goods............................................... ............. Diapers-.................... .............. ........................... 100.0 25.3 1.7 1.2 9.3 .9 .6 2.8 1.8 1.3 1.4 .6 1.2 .4 .3 1.1 .7 17.6 .1 .5 1.2 .8 2.7 .7 1.0 .5 .8 1.0 1.5 2.0 1.5 .2 .3 .4 .4 .4 .5 1.0 .1 12.4 3.1 .2 .1 1.2 .1 .1 .3 .2 .2 .2 .1 .1 .1 (*) .1 .1 2.2 (0 .1 .1 .1 .3 .1 .1 .1 .1 .1 .2 .2 .2 0) (0 .1 .1 .1 .1 .1 0) 14.6 .6 5.1 1.8 .6 .5 5.2 .8 16.6 4.6 1.1 .4 6.1 2.1 2.3 3.9 .5 1.1 1.0 .9 .4 4.5 2.3 1.3 .9 17.5 1.8 .1 .6 .2 .1 .1 .6 .1 2.1 .6 .1 Eggs, fresh.................................................................dozen . 5.8 2.4 Fruits and vegetables................................. Fresh fruits and vegetables........................ Fresh fruits: Apples............................................. __ pound.. Bananas....................................— .........do— Oranges........... .... ......................... --- dozen.. Fresh vegetables: Beans, green.................................. — pound.. Cabbage........................................ ----- do---Carrots..................—..................... ___bunch.. Lettuce........................................... ____ head— Onions........................................... ___pou ndPotatoes.......................................... 1 5 pounds. _ Spinach.......................................... — pound.. Sweetpotatoes............................... .........do---Canned fruits and vegetables.................... Canned fruits: Peaches......................................... No. 2 H can— Pineapple...................................... .............do— Canned vegetables: Com............................................... ..N o . 2 can.. Peas............................................... .............d o— Tomatoes....................................... .............do___ Dried fruits and vegetables: Dried fruits, prunes...................... ......... pound.. Dried vegetables, navy beans___ .............do___ 19.6 15.2 8.0 6.2 2.4 1.8 2.3 1.0 .7 .9 .7 .5 .9 1.4 .8 3.2 .7 .5 3.2 .3 .2 .4 .6 .3 1.3 .3 .2 1.3 .5 .4 .2 .2 .6 .4 1.3 1.2 .7 .5 .2 .2 .5 .5 .3 .2 Silk, rayon, and nylon....... ............... ...................... . Men’s: Socks...... ..................... .................... ...... Women’s: Dresses.................................... ........... Slips................................................. . Panties---------------------------------------Nightgowns...................... .................. Hose..................................... .............. Yard goods................................. ........................ Footwear....................... ....................................... ...... Men’s: Shoes, street........................... ................. Shoes, work........................ .................... Rubbers................................................. . Women’s: Shoes, street________ ____ _________ Children’s: Shoes, street, boys’ ........................ . Shoes, street, girls’ ........................... Other garments........................................................ . Men’s: Jackets, leather..................................... . Hats, felt....... .......................................... Women’s: Coats, fur................ ........... .............. . Girdles................................................ Gloves, leather.............. .................... Services............................ ........... .............................. Men’s: Dry cleaning............ .................... ......... Shoe repairs............. ............ ............ . Women’s: Shoe repairs...................... ................. Other apparel.............................................. .............. Beverages, coffee....................................................... do— 3.0 1.2 Rent.................................................................................... 100.0 Fats and oils......................................................................... Lard..................................... -.........................pound— Shortening, hydrogenated................................ do___ Salad dressing..................................................... pint— Oleomargarine................................................. pound-. 3.2 1.0 .6 .9 .7 1.3 .4 .2 .4 .3 Sugar and sweets, sugar. 2.9 1.2 Fuel, electricity, and refrigeration.................................. . Electricity...................................................... ............ Gas..................... .................................. ..................... Ice................................................. ..................... ........ Kerosene....................................... ................... ......... . Fuel oil.......................... .............................................. Anthracite coal, Pennsylvania.................................. 100.0 16.6 17.7 11.7 .9 5.9 15.3 i * 0.05 percent or less do. 0) .8 .3 .3 .5 .1 .1 .1 .1 .1 .5 .3 .1 .1 2.2 12.5 5.1 .9 .9 .6 .1 .3 .8 consum ers’ T able p r ic e 7 in d e x 1.— C PI items and their relative importance in the major groups and in the total index, December 1948— Continued Item Percentage distribu tion of index value factors in December 1948 Group total Fuel, electricity, and refrigeration—Continued B ituminous coal................... .......... ......... ................ W ood.................................................... ..................... Housefumishings___ ________ _______________________ Blankets.................................................. ................... Living rnnm set, inexpensive "Dining room set, medium Bedroom set, medium _ ____________ "Redrnnm set. inaypensive Sofa beds "Bedsprings __________ Mattresses Bad in-phnnngrapbs Sewing machines, electric Washing machines, electric_____________ _________ Vacuum cleaners, electric "Refrigerators, electric Stoves, cook Dinnerware, plate _ _______ Broom __________ _ Other housefumishings Miscellaneous.......................................... .............. ........... Transportation _ ____________ Automobiles _ _ _ _ ____________ Tires Gasoline Motor oil___ ________ Auto repairs........................................ .................. TaYes _ _ . Automobile insurance ____________ Streetcar fares....................................................... Bus fares Railroad fares Medical care Physicians: __ ___ _____ Office visit House visit................................... - ................ Obstetrical care............................................... 10.05 percent or less. Item Group total All items total 1.1 .4 22.3 8.6 .1 .5 .4 (i) 100.0 1.8 4.2 3.4 1.7 6.8 1.7 43 98 5.2 4.0 6.1 2.1 1.4 2.9 9.6 1.4 6.8 2.2 14.6 6.5 1.9 1.1 .5 (i) 0) (i) (I) u 4.7 .1 .2 .2 .1 .3 .1 2 6 .2 .2 .3 .1 .1 .1 .1 .3 .1 .4 .7 .3 .1 0) (0 100.0 27.8 9.4 .6 5.8 .6 .6 .5 1.3 7.6 1.1 .3 13.1 24.7 6.9 2.3 .1 1.4 .2 .2 .1 .3 1.9 .3 .1 3.2 1.9 1.7 .5 .5 .4 .1 Percentage distribu tion of index value factors in December 1948 Miscellaneous—Continued Medical care—Continued Surgeons: Appendectomy.... ............................. . Specialist: Tonsillectomy.................... ............... Dentist: Filling............................................................Extraction....................... .............................. Hospitals: Men’s pay ward............................................. Room................................................ ............. Optometrist: Glasses.................................... ...... Medicines and drugs: Prescriptions..................................... .......... Aspirin___________________ _______ ______ Quinine...............................................- ......... Antiseptic, iodine............. ............................. Milk of magnesia........................................... Accident and health insurance............................ Household operation____________ ________ _______ Laundry services........................... ...................... Telephone services........................ ................. . Domestic services.............................. ................ Postal services.......... ................................ .......... Water rent_________________ _____ ___________ Laundry soap: Bar.................................................................. Granulated..................................................... Toilet tissue......................................................... Other household supplies____________________ Recreation......................................... .......................... Newspapers_______ ______________ _____ _____ Motion pictures: Adults----------- -------- --------Tobacco: Cigars.................................. ......................... Cigarettes............................ ......... ......... ...... Pipe tobacco...... ................................... ........ Personal care...................... ..................................... . Barber service: Haircuts, men.......... ................. Beauty shop service: Wave set....... ................................................. Permanent wave.... .............................. ........ Toilet articles: Toilet soap....... ............................................. Toothpaste.................................................... Face powder. ................................................. Sanitary napkins............................................ Razor blades.................................................. Gifts, contributions, and other unallocated items...... All items total 0.4 .4 0.1 .1 1.9 .7 .5 .2 1.0 1.5 .6 .3 .4 .1 .9 .2 .1 .2 .5 .6 13.1 3.7 2.2 .6 .4 .8 .2 (*) 0) .1 .1 .1 3.2 .9 .6 .1 .1 .2 1.2 1.8 1.1 1.3 19.6 4.9 6.5 .3 .4 .3 .3 4.8 1.2 1.6 1.1 6.3 .8 9.9 4.1 .3 1.5 .2 2.4 1.0 1.1 .8 .2 .2 1.6 1.2 .6 .3 .2 16.5 .4 .3 .1 .1 .1 4.2 II. Collection and Compilation of Work Stoppage S tatistics7 Estimates showing the number of stoppages, workers involved, and man-days idle in the United States are issued monthly by the U. S. Department of Labor’s Bureau of Labor Statistics. Annually, totals are compiled and the statistics are also classified by industry, State, city, major issue, duration, etc. Strike statistics are a broad indica tor of industrial unrest. In this series an attempt is made to measure quantitatively the extent to which labor-management disputes result in stop pages of work. In 1880, the United States Bureau of the Census made the first exhaustive survey of labor disputes and published detailed information on 762 work stoppages. Subsequently the method of collect ing the information varied, and the statistical series on work stoppages automatically thus fall into several historical groupings. During 18811905, the Bureau of Labor (then in the Depart ment of the Interior) collected data on stoppages excluding those that involved fewer than six work ers or lasting less than 1 day— a practice that the Bureau follows currently. No Federal agency collected national information on stoppages in 1906-13. The Bureau compiled data on the num ber of stoppages only, during 1914-15. Infor mation on the number of workers involved was subsequently added for approximately two-thirds of the known stoppages in the 1916-26 period. Beginning with 1927, a fairly uniform procedure has been followed in obtaining detailed informa tion from the parties involved in work stoppages. Series have been computed on the amount of idle ness during work stoppages each month as well as on the number of stoppages and number of workers involved. Coverage of the series extends to all known strikes and lock-outs within the continental United States which involve 6 or more workers and last a full day or shift. Stoppages of American seamen 7 Prepared by Don Q. Crowther and Ann J. Herlihy of the Bureau’s Divi sion of Industrial Relations. 8 or other workers in foreign ports are not included, nor are strikes of foreign crews on foreign ships occurring in American ports. All employees made idle in the establishment are counted as “ in volved,” even though they may not be active participants or supporters of the controversy. All man-days in which work was scheduled are in cluded in the calculation. The Bureau defines a strike as a temporary stoppage of work by a group of employees to express a grievance or enforce a demand. Usu ally the issue in dispute is directly between the employer (s) and the striking employees, but there are significant exceptions. For example, in juris dictional, as well as in rival union or representa tion strikes, the major elements of dispute may be between two unions rather than directly with the employer. In a sympathy strike, usually no dispute exists between the striking workers and their immediate employer but the purpose is to give union support or broaden group pressure for the benefit of another group of workers. Some protest strikes are intended to register the dissat isfaction of workers with action (or the lack of action) by local, State, or Federal Government agencies on matters affecting their interests. So-called slow-downs, where employees continue at work, but at reduced production speed are not included, nor are those instances in which workers report an hour or two late each day as a protest gesture or quit work several hours before closing time to attend rallies or mass meetings. Limitations of the Series This series cannot be used as an accurate basis for the measurement of the cost of strikes, in terms of the amount of production and wages lost. The calculation of such items involves many factors for which information is not available, including, for example, production schedules before and after the stoppage, flow of raw materials, amount of overtime worked by employees, etc. Within the limits that the Bureau places on the series, a number of work stoppages involving few workers, or lasting short periods (i. e., less than six workers or lasting less than a full shift) are WORK STOPPAGE STATISTICS omitted from the count. Such disputes usually are of little importance in the over-all count, and frequently cause no significant idleness or inter ruption to production. Indirect or secondary effects of stoppages are not measured. The figures do not cover those employees made idle in other establishments or industries as a result of material or service short ages, resulting from a work stoppage. For ex ample, a prolonged coal strike may cause wide spread closing of industrial plants and a crippled transportation system, as fuel supplies are ex hausted. At times, the idleness of employees directly in volved in a strike may be considerably less than the idleness of other workers brought about in directly. No satisfactory measurement, however, has been evolved to gage or even reasonably to estimate such indirect effects of work stoppages. Therefore, the Bureau's work stoppage series is limited to the establishments directly involved. No attempt is made to distinguish between strikes and lock-outs because of the difficulty of determining the true facts. Stoppages are in cluded in the series regardless of who may be deemed “ responsible,” or which party takes the initiative. Survey Methods and Sources The Bureau seeks to obtain complete coverage. It does not base the series upon a sample but covers all stoppages of the specified size and dura tion for which information is obtained from any trustworthy source. Information on the existence of a stoppage is currently obtained from various sources, including (1) press clippings on labor disputes from daily and weekly newspapers throughout the country; (2) notices received directly from the Federal Mediation and Conciliation Service as well as from agencies concerned with labor-management disputes in over 30 States (such as, State media tion boards, research divisions of State Labor Departments, State Employment Service Offices, and Unemployment Compensation Offices); (3) various employer associations and some corpora tions; and (4) international unions. The im portance of the different sources has changed from time to time. If the Bureau has any indication that a work stoppage exists, questionnaires are sent by mail 884616— 50-------3 9 to both parties stated to be involved in order to secure first-hand knowledge as to the number of workers involved, the dates and duration of the stoppage, major issues involved, method of settle ment, etc. In some instances, field agents of the Bureau secure the necessary data. Strikes, by their very nature, are a matter of public knowledge and newspaper reporting. In formation as to the existence of a stoppage, its size, and major issues, therefore, is sometimes summarized on a case-by-case basis. The Bureau, of course, holds confidential the individual reports submitted by employers and unions, as well as supplementary data collected through State or Federal agencies. Calculation Procedures The Bureau's monthly strike series are based on estimates in large part. Those compiled annually are the result of an actual compilation of the figures from individual reports of work stoppages. Work stoppage series are always subject to some interpretation and rationalization. Monthly Estimates. Estimates are prepared and published monthly on the three specified measures of work stoppages: (1) number of stoppages, (2) number of workers involved, (3) man-days of idle ness. Such estimates are compiled, about 4 weeks after the end of the month of reference, from the most accurate information on all stoppages which have come to the attention of the Bureau. As the Bureau's experience shows a lag between the occurrence and reporting of a number of relatively small strikes, allowance is made (depending upon several variables) for these smaller stoppages in preparing the estimates of disputes occurring within the month. Estimates of the number of workers involved and total idleness are based up on known information on stoppages of 500 or more workers and/or 5,000 or more man-days of idle ness; allowance is made, based on the Bureau's existing information and past experience, for the smaller stoppages. The total working time lost during the month is compared with the estimated working time and published as a percentage. “ Total employed workers,” as used in making these computa tions, refers to all workers except those in occupations and professions in which there is little if any union organization or in which strikes rarely occur. In most industries it includes all 10 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES wage and salary workers except those in executive, managerial, or high supervisory positions or those performing professional work the nature of which makes union organization or group action im practicable. It excludes all self-employed, domes tic workers, agricultural wage workers on farms employing less than six, all Federal and State government employees, and officials (both elected and appointed) in local governments. Estimated working time is computed by multiplying the average number of employed workers in the month by the number of days worked per em ployee in the period. Annual Statistics. The annual series are totals of the number of stoppages, workers involved, and man-days of idleness. Compilation of such statistics is essentially a process of assembling the necessary information on individual cases, followed by analysis, evaluation, and classification into groups. Application of technical statistical formulae is not involved. The statistical unit is the individual strike or lock-out, irrespective of size. If groups of em ployees (regardless of their number or how widely scattered) join in a work stoppage for a common objective their action is classed as a single strike. The count of workers involved in a strike or lock out is the number actually made idle in the establishment directly involved. As already indi cated, no distinction is made between the actual participants in a strike and those respecting or kept idle by picket lines or those sent home by the employer when a stoppage in one department closes the plant. Man-days of idleness, like the number of workers involved, are based on the idleness at the establishments directly involved. Workers in volved multiplied by days of idleness equals total man-days idle. In this calculation, holidays and days not normally worked are omitted from the count of days of idleness. In addition, the annual statistics are classified according to a number of significant factors which are here described briefly. An industrial classification is made of each strike in accordance with the Standard Industrial Classi fication Manual published by the United States Bureau of the Budget. In a few stoppages, workers in more than one industry are directly affected. Small stoppages which fall in this category are classified in the industry having the majority of workers involved; in large interindus try stoppages, an allocation is made. The duration (length) of each stoppage is com puted on the basis of calendar days, rather than working days, i. e., the lapse of time in terms of calendar days from the beginning until the end of the stoppage. For stoppages which begin at a definite time and are terminated by a formal agree ment at a definite time, no problem arises in determining the duration. However, some strikes are never formally settled, although the workers may gradually go back to their jobs or find other employment; employers may be able to resume production with new recruits or may close their plants permanently. In such cases, the stop pages are terminated, for statistical purposes, when a majority of the vacancies are filled and production begins to approach normal. On oc casion, an actual settlement is later reached and the statistical record of the stoppage is then reopened, and the figures are adjusted correspondingly. Establishment involved is actually a single work place, e. g., a factory, mine, or store. In a wide spread strike of intercity bus drivers, truck drivers, or railroad workers, the establishment is regarded as the terminal out of which the employees work; in a strike of seamen, the ship is the establishment; and in a strike of dock workers the individual dock or loading place is regarded as the establishment or place of work. Geographical classification of stoppages follows State and city lines. In interstate stoppages, the workers involved and man-days idle are allocated to their respective States. Data are also com piled each year for 150 separate cities (excluding suburban areas outside the corporate limits). In general, all cities having a population of 100,000 or more in 1940 are covered. The causes of most strikes are multiple and varied, and do not always lend themselves readily to immediate and exact classification. After evaluation of the information available, the stop pages are classified by issues into four broad cate gories: (1) wages and/or hours; (2) union organi zation matters (representation, union security, etc.); (3) other working conditions, such as job security, physical working conditions, administra- 11 WORK STOPPAGE STATISTICS tive policies, work load, etc.; or (4) inter- or intra union matters. Within these groups they are further subdivided into more specific categories. Union involved is another major classification of the series. For this purpose the union involved is the union which has taken active leadership in the stoppage. In disputes involving more than one union (jurisdictional or rival union disputes as well as those of cooperating unions) classification is made accordingly. If unorganized workers strike independently, a separate classification is used. Method of termination of stoppages is the classi fication according to the means of termination. For example: (1) disputes in which the parties agree directly to terminate the stoppage without any third-party assistance; (2) those terminated with the assistance of private or nongovernment mediators; (3) those terminated with the assistance of government agencies; (4) those ending without formal settlements; and (5) those in which the employers discontinued business. Disposition of issues is the classification in which information regarding the settlement or disposition of issues is presented. In most strikes the issues are usually settled or disposed of before the return to work is effected, but provision is made to present data for the cases in which ad justment of issues after resumption of work is effected (1) by direct negotiations between the employers and the union (or workers); (2) by negotiation with the aid of Government agencies; (3) by arbitration; and (4) by other means (cases referred to NLRB union boards, tribunals, etc., where method is other than negotiation). The following questionnaire is used in collecting detailed information from both employers and unions. B. L. S. 817—(Rev. 1-1-48) Budget Bureau No. 44-R210.8. Approval expires 3-31-50. U. S. D E P A R T M E N T OF LABOR F ile BUREAU OF LABOR STATISTICS Washington 25, D. C. CONFIDENTIAL—Not for public inspection D ear S i r : The Bureau of Labor Statistics has received information Kindly furnish for official statistical purposes the information indicated below in connection with this work stoppage. Please return the report within 2 days, if possible, in the enclosed envelope which requires no postage. If you do not have the information, kindly forward the blank to the proper official or give us his name and address. Your cooperation will be greatly appreciated. Very truly yours, E w a n C la g u e , Commissioner of Labor Statistics. 1. Name of company____________________________________________________________________________ 2. Address of central office_______________________________________________________________________ 3. Principal products or services of plant (s) involved in work stoppage (list in order of importance) 4. Number of establishments (or work places) involved [AFL 5. Union(s) involved___________________________________________________________(C h e ck )iciO . (Name) [ I n d_ Local N o____________________ Address___________________________________________________________________________ Address___________________________________________________ 6. Dates and number of workers idle. (Please show separate data for each establishment if available; if not, give estimate for entire stoppage and show the total number of establishments involved.) Establishment involved and location (City and State) First day of stoppage Date Hour Workers idle* Date greatest number of workers were idle Date Workers idle* Last day of stoppage Date Workers idle* Number of workers on payroll (before stoppage) (Use additional sheets if more space is needed) the total number of workers idle in each plant or establishment reported—those concerned directly and those made idle because of dispute, Digitized for ♦Show FRASER 12 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES 7. If stoppage lasted only 1 or 2 days, how many workers were idle at least one full shift?___ 8. Number of days worked per week by most employees before stoppage--------------------------------9. Causes of dispute. (Specify, in order of importance, the issues involved or demands made.) 10. How long before stoppage were principal issues listed in item 9 a matter of dispute? Number of days. 11. Relation of stoppage to union-management contract: Was contract in effect when stoppage occurred? Y e s ______ N o ______ Dispute was due primarily to: (а) Attempt to obtain union recognition ( ) or establish first contract_________________________ (б) Grievances________________________________________________________________________________ (c) Failure to agree on renewal of contract____________________________________________________ (Date old contract expired or was scheduled to expire)_____________________________________ (d) Attempt to alter contract terms during life of the agreement_______________________________ (e) Matters not involving the question of a contract___________________________________________ 12. Did a Federal, State, or local government agency participate in negotiations before the stoppage began? N o __________ If “ yes” give name of a g e n cy ________________________________ Y e s __________ 13. Method of terminating stoppage: Agreement or understanding for return to work was reached: (а) By employer (s) and union directly_________________________________________________________ (б) With assistance of government agency: Federal (Name__________________________________________________________________ ) ______ State (Name____________________________________________________________________ ) ______ Local (Name____________________________________________________________________ ) ______ (c) By other means (explain)__________________________________________________________________ Cd) Workers returned without formal arrangement or settlement_________________________________ 14. Date settlement was reached___________________________________ Date most employees resumed work 15. Were all issues completely settled at termination of stoppage? Y e s ______ N o ______ If not, please indicate how the remaining issues were to be finally adjusted: (a) By direct negotiations between employer (s) and union_______________________________________ (b) By negotiations with the aid of a government agency. N a m e______________________________ (c) By arbitration (Name of arbitrator)________________________________________________________ (d) By other means (indicate)__________________________________________________________________ (Check) (Check) (Check) 16. I f no agreement to terminate stoppage has been reached, have the majority of the vacancies been filled either by old or new employees? Y e s ______ N o ______ If “ yes” give date by which majority of vacancies were filled_______ (Date) 17. Was there any violence in connection with the stoppage? Y e s ______ N o ______ Were there any deaths? Y e s ______ N o ______ N u m ber_____________________ Any injuries (necessitating medical attention)? Y e s ______ N o ______ N um ber____________________________ E xplain ______________________________________________________________________________________________________ 1 8 18. R emarks: (Signature of person making report) (Company or organization) (Position or office) (Date) (Address) PLEASE ENCLOSE A COPY OF ANY NEW AGREEMENT OR AMENDMENT TO OLD AGREEMENT SIGNED AT CONCLUSION OF DISPUTE III. Estimating National Housing Volume8 The housing statistics series, prepared by the U. S. Department of Labor's Bureau of Labor Statistics, measures the number of new nonfarm dwelling units started nationally each month. It is used generally as an indicator of building activity and related economic trends and by housing agen cies as a guide in national housing policy and State and local administrative decisions. The statistics are available on a monthly basis beginning in 1939 and on an annual basis from 1910. Over the years, the chief source of information about home-building activity has been the building permit. The Bureau began collecting buildingpermit information in 1920, with reports from 207 large cities. Coverage has expanded annually, but the most important strides were taken between 1933 and 1940. The dwelling unit, the unit of measurement of the volume of housing construction, is defined by the Bureau of Labor Statistics 9 as a permanent dwelling place containing permanent cooking facil ities, i. e., the minimum built-in facilities essential to housekeeping. The dwelling-unit count repre sents the number of families planned for in the construction of new permanent-type houskeeping dwellings and reflects the extent of new housing activity. Prefabricated houses are included, if permanent and made of new materials. Temporary units and units without housekeep ing facilities and such dwellings as trailers, house boats, sheds, and shacks, are not included. Excluded also are the temporary dwellings built 8 Prepared by Dorothy K. Newman in the Bureau’s Division of Construc tion Statistics. 9 See Census of Housing, 1940, Part I, United States Summary (p. 2) for Census definition of a dwelling unit. See also Housing and the Increase in Population, Bureau of Labor Statistics Serial No. R. 1421 (pp. 14-16) for differences between Bureau of Labor Statistics and Census definitions. The Bureau of the Census studies families as a unit of population and measures the number and kind of family accommodations, new and old, regardless of structural permanency or the significance of the housing in the volume of residential construction. The Bureau of Labor Statistics prepares current housing statistics and, from the results of building-permit reports, interviews with builders, and the like, measures the number of new perma nent dwelling units started in structures designed and built for residential purposes. during the period of defense and World War II, and the Federal temporary re-use units erected during the Veterans' Emergency Housing Program of 1946-47. Accommodations in transient hotels, dormito ries, and clubhouses are not counted in the dwellingunit figures. These are usually nonhousekeeping quarters and the buildings containing them are defined as “ nonhouskeeping residential." Farm dwellings are likewise excluded from coverage. Restriction of coverage to new units automati cally excludes units provided by the remodeling of existing residential structures or the conversion of nonresidential buildings into housing. And since the Bureau's housing statistics are designed to reflect the extent of new house-building activity, and not necessarily all additions to the housing inventory, living quarters provided for superin tendents in public buildings, warehouses, and factories are excluded also. Construction of the residence in these cases is quite incidental to the nonresidential building. On the other hand, the Bureau's totals do include housekeeping dwelling units in buddings that also contain stores. In such cases the housing accommodations are at least as important as the stores and usually account for a major part of both the physical volume and value of the construction job. The new permanent nonfarm dwelling units included in the Bureau of Labor Statistics series are classified as urban or rural nonfarm; private or public; in one-family, two-family, and multifamily structures. Urban units are those in urban areas, which, according to Census definition, are all incorpo rated places which had 2,500 population or over at the time of the latest census and, by special rule, a small number of unincorporated civil divi sions essentially urban in character. Rural non farm units are defined as those in incorporated places with less than 2,500 population, and all units in unincorporated areas that are not among those just mentioned nor are they farm homes. Thus, urban housing is related to definite geo graphic areas, while rural nonfarm housing is 13 14 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES defined largely according to the intended use of the dwelling units. Dwelling units financed by Federal, State, or local government funds are public units; all others are private. The fact that private units are financed by mortgages insured by the Federal Housing Administration or the Veterans Admin istration does not mean that they are publicly financed. A one-family structure may be detached, semi detached, or one of a solid row. A semidetached one-family structure has a common wall with another structure containing a single dwelling unit. Each unit in both semidetached and row houses is counted as a separate structure, because each has a separate entrance and separate heating facilities and utility connections. Two-family structures are those which are built so that one unit is above the other or two units on the same floor have a common entrance. In the multifamily structure, heating facilities and utilities are usually centrally controlled, and a single entrance leads to the various apartments. In apartments with individual entrances, the units are defined as being in multifamily structures because the heating, the plumbing, and, in some cases, other facilities, such as electricity and gas, may be controlled at a central location. Limitations of the Series Statistics on the number of dwelling units started do not measure the number completed in any given month. Construction on units started usually continues for several months before the dwellings are ready for occupancy. Furthermore, the Bureau's totals of starts can not be added to the number of units standing as shown in the Census of Housing (allowing for demolitions and the number of units destroyed by natural or other causes) to form an all-inclusive housing inventory. The reasons are the limita tions placed upon coverage of the series, already partially explained. Methods and Sources of Survey A questionnaire form (BLS 404) is mailed by the Bureau 10 each month to the building-permitw With the exception that the Department of Labor or like acency in 8 States (Illinois, Massachusetts, Michigan, New York, New Jersey, North Carolina, Pennsylvania, Texas) send questionnaire forms directly to build ing inspectors in their State and then assemble and publish the State data. Copies of the permit reports are sent to the Bureau of Labor Statistics in Washington for use in preparing summaries and national estimates. issuing officers in about 2,500 urban and 2,600 rural places throughout the country, including over 500 counties and townships. Forms are sent to practically all localities having building-permit systems, and returns are received monthly from about 9 in 10 of them. Information is requested on this questionnaire as to the number and value of the new dwelling units for which permits were issued, as well as certain details about nonresidential building. The portion of BLS 404 relating to housing is re produced on p. 19. Forms are mailed on the twenty-fourth of every month. Returns are suffi cient for estimating purposes by the fifteenth of the following month. But editing and tabulating of the data delay the actual preparation of the estimate by about 2 weeks. To obtain an early preliminary estimate, brief telegraphic forms are mailed on the same day as the questionnaire to a sample of the buildingpermit officials (about 550) who also report on the longer form. On the telegraphic forms, they are asked only the number of new family dwelling units for which permits were issued during the month. Returns, made by wire, are usually com plete by the eighth of the month following the month of reference, and the preliminary estimate is published about the fifteenth. Field surveys conducted to supplement the mailed questionnaire are limited to the nonpermit issuing parts of a sample of 96 rural counties. Each of the 96 counties is visited once each quarter, but at each visit the number of dwelling units started in each of the 3 previous months is obtained. The 96-county sample, thus, is divided into 3 groups of 32 counties each. One group is visited in January, April, July, and October; another in February, May, August, and November; and the last, in March, June, September, and December. Field investigators obtain leads to new homebuilding from local builders, utility companies, building-supply companies, real-estate agents, and a variety of other sources. The next step is to secure information directly from builder or owner as to the date construction was begun and the number of units in the project. In addition, each Bureau investigator inspects his territory in order to complete the canvass of all new homebuilding begun in the three previous months. The work of Bureau field agents is carefully reviewed in the Bureau's five regional offices, and NATIONAL HOUSING VOLUME an on-the-spot check is made of the completeness and accuracy of field investigations on the average of once every 6 months. 15 between the previous and the current month for identical cities is applied in each estimating cell (i. e., in this particular instance, data reported for places of given size and given locations) to the Calculation Procedures previous month’s estimate for all the cities repre Two separate calculations are made covering sented by that cell. By this procedure a prelim inary estimate is obtained of the total number of housing volume each month. These result in dwelling units for which building permits were the preliminary and revised figures issued by the Bureau. Both estimates are based upon samples, issued or work was about to begin in urban areas. and, as explained below, the sample utilized in the It is not an estimate of the amount of housing revised estimate is considerably broader than that actually started. An adjustment is then made to translate building-permit volume into dwelling for the preliminary estimate. units started. The Preliminary Estimate. In the preparation of Factors for this adjustment are based on the preliminary estimate, the telegraphic replies periodic field studies in sample localities in which used cover the number of new nonfarm dwelling the Bureau investigates the elapsed time between units started (1) in all of the 199 cities with 50,000 issuance of a building permit and the start of population or more in 1940; (2) in 45 rural non construction, and the extent to which permits are farm localities known to be active in homenot used. Compared with 1945, studies show building; and (3) in a sample of 256 cities of less that in 1948 the rate of lapsed permits has de clined from over 7 percent to only 1 percent of than 50,000 population chosen and stratified the dwelling units reported on permits. It is according to geographic division, location within estimated that in urban areas nearly 60 percent or outside of a metropolitan area, and size. Data are also included for a selection of 230 cities of the units are started in the month of permit issuance; and 94 percent by the end of the second which consistently submit their mail question month afterward. Adjustments are made each naires to the Bureau before the eighth of the month for such delays and lapses: an addition is month. The entire urban segment of reporting made for units left over from the estimated permit places which supply information in time for the preliminary estimate usually provides complete volume for the previous month; subtractions are made for the proportion to be started in later coverage for the 412 cities of 25,000 population or months, and for those abandoned, or, as in a few more, as well as for a representative sample of cases, started before the permit was issued. smaller urban places. To the telegraphic replies for 45 rural-nonfarm (2) The estimating method for the rural non localities are added all the questionnaire returns farm permit-issuing group resembles that for from rural nonfarm places which have been urban places. However, the reported permit received in time for the compilation of the pre data are stratified at this stage only by permit issuing jurisdiction, i. e., for incorporated places, liminary housing estimate. The total usually represents about 80 percent of the housing volume townships, and counties. For each classification, in rural places issuing permits, but only about 50 a total is made of all of the dwelling units for which building-permit reports have been received, percent of the places. and the percent of change between the previous Utilizing the foregoing basic figures, the pri and current month’s reports for identical localities vately financed segment of the estimate is made is applied to the previous month’s estimate. in three parts— (1) for urban places, (2) for rural Separate treatment is given areas of significant nonfarm places issuing building permits, and (3) housing volume that show trends widely variant for rural nonfarm places without permit systems. (1) To obtain the urban estimate, permit data from the general trend. The sum of the data for incorporated places, townships, and counties for the current month are grouped according to the yields the estimated total number of dwelling geographic division of the places reporting, the units for which permits were issued in rurallocation of reporting places within or outside a nonfarm permit-issuing places in the month. This metropolitan area, and their size. The percent Digitized of for change FRASER in the number of dwelling units reported aggregate is then adjusted to reflect the number 16 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES of dwelling units started, in accordance with the information for rural areas revealed in the Bu reau’s building-permit surveys. On the whole, these surveys show somewhat less lag in rural than in urban places between permit issuance and the start of construction. (3) The preliminary estimate covering the num ber of new nonfarm dwelling units started in ruralnonfarm places that do not issue permits is derived at this point by projecting the previous month’s figure, using the trend shown for the rural-nonfarm permit-issuing places. To the figure thus obtained for privately financed housing the Bureau adds the number of publicly financed units started. Information on public housing is received directly from the sponsoring Federal, State, and local agencies. The resultant total (public plus private) yields the preliminary estimate of the number of new nonfarm dwelling units started nationally for the month. The Revised Estimate. Revision of the preliminary monthly estimate is usually made at the end of every quarter after results are available from the Bureau’s field surveys in the nonpermit-issuing segments of 96 rural counties. In selecting the sample of 96 counties for survey, the Bureau in 1947 eliminated 86 counties of the country’s total of 3,103 from the list either because they were completely urban or were served entirely by building-permit systems. The nonpermit issuing rural nonfarm universe was determined according to the number of rural nonfarm dwelling units standing in 1940 in that part of each of the remaining counties where building permits are not issued. The universe was stratified according to whether the counties were metropolitan or nonmetropoli tan,*11 and whether more urban or more rural in character, as defined by the percentage of urban to total dwelling units standing in 1940. Thus classified, the metropolitan counties had 1.6 million or 40 percent of all rural nonfarm dwelling units standing in 1940 in nonpermit-issuing areas; the nonmetropolitan counties had 4.8 million or 60 percent of such units. A further division into 4 temperature zones was made on the basis 11For this purpose a county was metropolitan if any part was located within a metropolitan area as defined by the 1940 Census. of winter temperature.12 These classifications re sulted altogether in 15 cells. In order to avoid selecting a sparsely populated county to represent an estimating cell having a large number of dwelling units, an array was made for each cell according to 1940 housing inventory. The counties at the lowest end of the array, representing 10 percent of the 1940 inventory, were set aside.13 These small counties are, how ever, included in the cell totals in determining the estimating weights. Further classification of the universe was then made according to the extent of housing activity, as measured by the number of priority author izations to secure building materials for housing that were issued in each county in April 1946 under the Veterans’ Emergency Housing Program. In the southern temperature zone, this step in stratification followed classification by race (white, nonwhite) in the largely rural counties, but classification was made by race and not according to housing activity in the more urbanized counties. By this stage, the universe had been classified into 32 cells. Each of the 32 cells was further divided into 3 subcells, in such a manner that each subcell would represent as nearly as possible the same number of 1940 dwelling units. Within each cell, the counties were arrayed according to the number of dwelling units standing in 1940. The ftth county in each cell was selected by using a table of random numbers. The revised estimate, like the preliminary, is prepared in three parts—for urban, rural nonfarm permit-issuing, and rural nonfarm nonpermit issuing places. The estimating procedure for the rural nonfarm nonpermit-issuing places is to apply the weight for each county to the reported number of dwelling units for the month, and to total the weighted figures. The weight for each county is the relationship of the number of dwelling units standing in 1940 in the rural nonfarm nonpermit issuing parts of the county, to the number of 1940 12 The 48 States were classified into 4 zones based on the number of “ degree days’ ' (i. e., days with temperature below 65 degrees), in November, De cember, January, and February, as reported by the U. S. Weather Bureau. Some States, such as New York, Pennsylvania, Oregon, Washington, and Illinois were divided into 2 parts because of the wide range of winter tem perature. 11 Studies of variance in nonfarm units started in rural and small urban counties covered by area housing surveys in 1946 and 1947 indicated that bias resulting from the elimination of small counties in selecting the sample would be much less important than the error that might result from inclusion of one of the small counties. NATIONAL HOUSING VOLUME rural nonfarm dwelling units in the entire cell represented by the county. The urban and rural nonfarm permit-issuing segments of the revised estimate are prepared from virtually complete building permit returns. In estimating for urban areas, stratification of the expanded data is quite detailed, in comparison with that done in the preliminary estimate. The data for the revised estimate are classified according to type of structure (i. e., in one-family, two-family, or multifamily structures), and accord ing to the location of the places reporting, i. e., by geographic division, State, metropolitan or nonmetropolitan district, and city size. This classification process may yield as many as 11 cells in a State. The volume of homebuilding during the month is tabulated for each class of place, by type of structure. The estimate by type of structure for all urban areas is prepared by applying to the volume of housing reported for each type of place, the ratio between the total number of such places in the cell and the number reporting that month. The ratio of reporting to the actual number of places is usually 1 for cities of 25,000 population or over, because reports are received monthly from virtually all of these. For places of 5,000 to 25,000 population the multiplying figure is seldom over 3, and for places of 2,500 to 5,000 population, it is seldom over 5. Totals by type of structure are added to equal the urban total for the country, unadjusted for lag between building-permit issuance and the start of construction, and for building permits allowed to lapse. The urban total is adjusted as described on page 15 to yield the estimate of housing actually started. In preparing the revised estimate for rural non farm permit-issuing places, returns from the 1,800 places usually reporting building-permit volume are classified according to the kind of reporting locality (city, township, or county) and by location in or outside a metropolitan district, by geographic division, and by size of place. A total of 54 estimating cells results. For each cell, the estimate is derived by multiplying the number of dwelling units reported in the month, by the ratio between the total number of rural non farm dwelling units standing in 1940 in all places in the cell and the number of 1940 dwelling units in the reporting places. Adjustment is, of course, 17 made to convert building-permit volume into housing started. After adding publicly financed units, the total for the three parts of the estimate just described yields the revised estimate of new permanent nonfarm dwelling units started nationally for the month. Tests of Reliability The sampling error in the revised estimate of private nonfarm dwelling units started (the public segment is based on actual enumerations) amounts to 2 percent, using December 1948 data. Thus, if the estimate were 50,000, the chances are about 19 out of 20 that an actual enumeration would produce a figure between 48,000 and 52,000. Owing to the degree of completeness of the information reported, the percent of error is least for the urban segment of the estimate (1.0 per cent), slightly greater for the estimate covering permit-issuing rural nonfarm places (1.58 percent), and greatest for the estimate representing rural nonfarm nonpermit-issuing places (9.47 percent). Study of the revisions that were required in the preliminary estimate for months prior to March 1949 shows that adjustments have seldom reached 10 percent, and for most months they have been less than 4 percent. The revisions have usually been upward. Recent substantial addi tions to the number of permit reports available for the preliminary estimate will probably reduce somewhat the difference between the preliminary and revised figure, insofar as differences relate to the permit-issuing segments of the estimate. The magnitude of the revisions, however, re sults chiefly from the difference between the esti mate for rural nonfarm non-permit-issuing places based on field survey data and the projected figure used for the preliminary estimate. As previously stated, in the preliminary figure, housing activity in rural nonfarm places which do not issue permits is carried forward on the basis of the trend shown by activity in the permit-issuing rural nonfarm places. A figure prepared in this way is reliable, of course, only because an estimate based on field surveys is always within a span of 3 months, and provides a sound base for projection. Even so, experience shows that although the trend for rural nonfarm nonpermit-issuing places cor responds well with the trend in rural nonfarm permit-issuing places during the spring, summer, 18 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES and early fall, these two trends are less alike in the winter. The reason for this phenomenon is that the non-permit-issuing group appears to be more sensitive to seasonal influence, with home-building activity falling off faster in the winter and picking up more quickly in the spring. Experience has been insufficient to adjust for this condition because the estimating technique described here has been in operation through only two winters. However, after studying data cover ing the third winter, in 1949-50, satisfactory sea sonal adjustment factors undoubtedly can be pre pared and applied to the nonpermit rural nonfarm segment for use in the preliminary estimates. 19 NATIONAL HOUSING VOLUME ?■ *• s- 4cvi (Rev. 10-1-48) XJ. S. DEPARTMENT OF LABOR Budget Bureau No. 44-R049.9. Approval expires Sept. 30, 1949 BUREAU OF LABOR STATISTICS WASHINGTON 25, D. C. FOR BLS OFFICE USE ONLY Kardex In Tab. Add Edit Made Ver. Made Post Ver. Mach. Hand Report permits issued during month of— (If above mailing address is incorrect or zone number omitted, please indicate change) D e a r S i r : Please fill out this form and return it to the Commissioner of Labor Statistics, Washington 25, D . C., in the enclosed envelope which requires no postage. It will be observed that the number of buildings covered by permits is requested, but not the number of permits. We shall appreciate your courtesy if you will give this matter your immediate attention. Very truly yours, E w a n C l a g u e , Commissioner of Labor Statistics. (1) Privately owned Classification Number of build ings Estimated cost (omit cents) Number of dwell ing units (2) Publicly owned Number of build ings Estimated cost (omit cents) Number of dwell ing units Code NEW FAMILY DWELLING UNIT STRUCTURES 1. Single-family structures. (M ay be detached, semidetached, or one of a solid row. A semi detached single-family structure has a common wall with another structure containing a single dwelling unit. Each unit is counted as a sepa rate structure because each unit has a separate entrance.) _ _ ___________ _______________ _____ 2. Two-family structures. (May have one unit over the other or two units on the same floor with a common entrance.) _____________________________ 2a. Single-family and two-family structures with a store or shop therewith. (These should not be included in the lines above.)___ ________ ______ 3. Three- and four-family structures having common facilities such a common entrance, heating, etc__ 3a. Three- and four-family structures having stores and shops therewith. (These should not be in cluded in line 3 above.)_______ ___________ 4. Five or more family structures having common facilities such as common entrance, heating, etc_ 4a. Five or more family structures having stores or shops therewith. (These should not be in cluded in line 4 above.) _ _______ 01 02 02a 03 03a 04 04a IV. Measurement of Labor Turn-Over* 1 0 24 A measure of the gross movement of workers into and out of employment status with individual firms is provided in the U. S. Labor Department’s Bureau of Labor Statistics monthly series on labor turn-over. Transfers within the employment of the same firm, as from one department or plant to another, are not considered labor turn-over. For analysis, personnel actions covered by this series are broadly divided into accessions, or addi tions to employment, and separations, or termi nations of employment. Separations are further classified according to type: Quits (or voluntary separations); and discharges, lay-offs, and mis cellaneous separations (collectively called involun tary separations). Accessions are all additions to the work force whether of new employees or of former employees after seven or more consecutive calendar days’ absence. Such absence may be either authorized (such as after a lay-off) or unauthorized. Quits are terminations initiated by employees because of acceptance of jobs elsewhere, dissatis faction, return to school, marriage, maternity, ill health, or voluntary retirement where no pensions are provided by the different companies. Unau thorized absences of seven or more consecutive calendar days also are considered quits. Discharges are terminations of employment initiated by management for such reasons as employees’ incompetence, violation of rules, dis honesty, insubordination, laziness, habitual absen teeism, or inability to meet the organization’s physical standards. Lay-offs are terminations of employment lasting seven or more calendar days which are initiated by management without prejudice to the workers, because of lack of orders, shortage of materials, conversion of plant to new product, or introduc tion of improved machinery or processes. Suspen sions of employment for less than 7 days and suspensions for inventory or vacation periods are not considered lay-offs. Miscellaneous separations are terminations for 14 Prepared by Lucile C. Ursell of the Bureau’s Division of Employment Statistics. 20 other reasons, including permanent disability, death, retirement on company pension, or en trance into the armed forces. Prior to September 1940, miscellaneous separations were included with quits. Personnel actions during a calendar month are converted to a rate per 100 employees. Separate rates are computed for total accessions, total sepa rations, and for each of the component separation items— quits, discharges, lay-offs, and miscella neous separations. A single labor turn-over rate is not provided.15 The number of personnel actions and of persons employed used in preparing labor turn-over rates cover all employees— administrative, office, and supervisory, as well as production workers— and permanent and temporary,16 full- and part-time employees on any type of pay roll (daily, weekly, monthly, or other). The employment count refers to the number of such persons who were on the pay roll in the pay period ending nearest the 15th of the month. In 1949, the Bureau prepared labor turn-over rates for 64 manufacturing industries, 19 major manufacturing industry groups, the durable and nondurable goods divisions, and for all manufac turing industries combined. In addition, rates were prepared for 7 selected nonmanufacturing industries, mainly in the mining and public utili ties groups. These together with the earliest date for which the series are available, are shown in table 2. Because of lack of facilities, publica tion of the data for men and women has not been made since July 1947, but will be resumed in January 1950. Labor turn-over rates first were obtained in response to a demand from large manufacturers who were experiencing difficulty in maintaining a stable work force after World War I. Conse quently, the subject was first studied in connec tion with the recruiting and handling of employees, and the net or replacement rate 17*was emphasized 18 Although the “ replacement rate’’ or “ net turn-over rate” was frequently referred to in the early years when the Bureau issued the labor turn-over series, it has not been published separately since November 1934. It is either the total accession or total separation rate, whichever is lower, and therefore is available from the published data for those who wish to follow it. 16 Since January 1946, employment on force-account construction has been included for all industries except telegraph. 17 The net or replacement rate is either the total accession rate or total separation rate, whichever is lower. MEASUREMENT OF LABOR TURN-OYER 21 as an index of management efficiency. Wide spread use of improved personnel methods, includ ing scientific aptitude and intelligence tests for prospective employees, exit interviews, pension plans, and employment guarantees under specified conditions in collective-bargaining agreements, have reduced turn-over rates in recent years to relatively low levels. Nevertheless, individual employers still use the rates for their particular industries as a yardstick of individual plant per formance. By this means, they determine when a particular establishment’s rates are excessive and therefore require special analysis and remedial measures. Annual labor turn-over rates, which are some times computed by totaling the 12 monthly rates or by computing an equivalent annual rate from a single month’s rate after making proper allowance for the length of the month,18 are valuable in calcu lating the cost of labor turn-over. But annual quit or separation rates computed in these ways are extremely difficult to interpret. For example, they often amount to over 50 percent and seem to indicate that over half the work force changed jobs during the course of the year; actually job chang ing probably is confined to a relatively small segment of employees. For certain purposes a more meaningful annual rate is the average of the 12 monthly rates. Increasingly, labor turn-over rates are being used in over-all economic analyses to indicate the gross worker movements which underlie the net changes reflected in the employment series. As the two series are currently prepared, however (with different-sized samples, different industry coverage, definitions, etc.), the labor turn-over rates indicate the nature of employment changes only very roughly. In fact, a relatively high proportion of large firms in the turn-over sample may make the rates somewhat lower and more stable than they would be if smaller firms had greater representation. The reason is that large firms tend to be more affected by employmentstabilizing influences— union agreements, facilities for screening applicants more scientifically, widely marketed products, and so on— than small enter prises. For this same reason, caution should be used in applying the manufacturing turn-over rates to interpret employment changes in the non- manufacturing industries. In contrast to the man ufacturing division, many major nonmanufacturing groups (e. g., trade and service industries) have a large proportion of relatively small establish ments and experience marked seasonal fluc tuations. 18 In addition to the monthly rates, equivalent annual rates were published from July 1929 through August 1931. 19 See note VI. p. 29. Limitations of the Series As indicated above, changes in the Bureau’s employment series cannot be measured precisely by the labor turn-over data. The two series are not geared into the same period; the samples are not completely representative; the employment and industry coverage are not the same; and industry classifications are not identical.19 The turn-over sample covers far fewer establish ments than the employment sample. Therefore, data are available for fewer individual industries in the labor turn-over series than in the employ ment series. Moreover, the latter covers the highly seasonal industries, as well as printing and publishing, and reflects the influence of work stoppages, which are eliminated entirely from the former. Before the Standard Industrial Classification was adopted for both series, definitions of indi vidual manufacturing industries showed little com parability. Up to that time, the employment series was based on the Census of Manufactures industrial classification system and fixed product classification of firms, whereas the labor turn-over series used this system only until 1943. There after, the Social Security Administration’s indus trial classification system and current product classification of firms were used for labor turn-over. Since January 1943, as already stated, the labor turn-over data have covered all employees. Although total employment has been available from the employment series for all manufacturing and major industry groups since 1939, it became available for individual industries only after the adoption of the Standard Industrial Classification. In both series, employment is for the week ending nearest the 15th of the month, but the labor turn-over items refer to the calendar month. Consequently, labor turn-over measures changes during a calendar month; the employment series reflects those from mid-month to mid-month. Preparation of Union Scales of Wages and Honrs Series 22 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES Survey Methods and Sources Information is collected each month on a mail questionnaire which is sent to individual establish ments. The questionnaire provides for reporting the total number of employees and the number of personnel actions occurring during the month, classified by type. In order to complete the count of employees entering and leaving the employment of the establishment, the number of transfers to and from other plants of the same firm is also re corded but they are not included in the tabulations. Separate data for women are obtained for total employment, total accessions, total separations, and quits. Bureau turn-over rates are based upon experi ence in a sample group of establishments. For the most part, the sample of respondents represents the largest establishments in each industry plus a distribution of medium- and small-sized establish ments. The approximate coverage of the sample for major groups is as follows: Number of establish ments in sample All manufacturing_______ __ Durable goods______ Nondurable goods__ . . . Metal mining____________ Coal mining Anthracite_____ ____ Bituminous_________ Communication Telephone__________ . . . Telegraph__________ . . . Employment Percent In reporting of uni verse establishments 6,900 4,000 2,900 140 4, 500, 000 2, 900, 000 1, 600, 000 59, 000 32 39 24 61 50 300 38, 000 130, 000 46 30 516, 000 41, 000 90 63 (i) (i) 1Data not available. To prevent fluctuations of employment in highly seasonal industries from obscuring the turn-over characteristics of other industries, such lines of activity as fertilizer manufacturing in the chemical group and canning and preserving in the food group are excluded from the sample. Currently, printing and publishing are also excluded. Other wise, all other manufacturing industries are represented, although samples are too small to permit separate publication of each. Calculation of Turn-Over Rates Monthly labor turn-over rates are computed for individual industries by dividing the total number of personnel actions of each kind (accessions, quits, lay-offs, etc.) reported by the respondents in the sample by the total employment reported by these firms and multiplying the result by 100. For example, in the sample for industry A, the total number of employees who worked during or received pay for the week of June 12-18 was re ported as 25,498. During the period June 1-30, in all the reporting firms in industry A, a total of 284 employees quit. From these figures, the quit rate of 1.1 per 100 employees is computed as follows: _____ X100 —1 1 25, 498X UU Through 1949, the industry classification system developed in 1942 by the Social Security Adminis tration continued in use. By January 1950, how ever, the Standard Industrial Classification being adopted by Federal statistical agencies is to be substituted in the labor turn-over series. Regard less of classification system, allocation of reporting establishments to the various industries is based upon major product or activity, as measured by sales value in the preceding calendar year. In computing labor turn-over rates for industry groups after the change to the Standard Industrial Classification in January 1950, the rates for indi vidual industries are to be weighted by total employment in each industry. Under existing procedures (1949), the labor turn-over rates for industry groups are not weighted by employment except when unusual circumstances (such as a fire causing a complete shut-down) affect only one or a few plants in an industry. Under both new and existing classification systems, the rates for all manufacturing and durable and nondurable goods are obtained similarly, by weighting the industry group rates by employment. To avoid distortion of the rates, the figures for individual plants are excluded from the computa tions for a given period if they are directly affected by a work stoppage at any time during the period. If a work stoppage is widespread and affects a substantial number of the reporting firms in an industry, rates for that industry are omitted. General comparability of the sample is insured from month to month by telegraphic follow-up of any delinquent firms (particularly large ones which would have considerable influence on the rates). The publication of revised rates for the month preceding the current month also assures com parability of the sample, as most delinquent reports are received in time for inclusion in the revised rates. 23 MEASUREMENT OF LABOR TURN-OVER T able 2.— Earliest date for which labor turn-over rates were published for industry groups and industries Industry group and industry1 Earliest date published2 January 1930.3 January 1943.* Do. Durable goods Iron and steel and their products...................... - ......... ........ Blast furnaces, steel works, and rolling mills 8.............. Gray-iron castings................ ......................................... Malleable-iron castings................................ ................. Steel castings...... ........ - ........- ........................................ Cast-iron pipe and fittings------ ------ ------------ ------ ----Tin cans and other tinware............................................ Wire products.................. ............................................. Cutlery and edge tools---------- ---------- ---------------------Tools (except edge tools, machine tools, files, and saws). Hardware.......... ...... .............. .............. ........................ Stoves, oil burners, and heating equipment............... . Steam and hot-water heating apparatus and steam fittings. Stamped and enameled ware and galvanizing............... Fabricated structural-metal products............................ Bolts, nuts, washers, and rivets.............. ...................... Forgings, iron and steel.................................................. Electrical machinery........................................................... . Electrical equipment for industrial use....... .................. Radios, radio equipment, and phonographs.................. Communication equipment, except radios.................... Machinery, except electrical.................................................. Engines and turbines..................................................... Agricultural machinery and tractors.............................. Machine tools 8......... ..................................................... Machine tool accessories8.............................................. Metalworking machinery and equipment, not else where classified. General industrial machinery, except pumps................ Pumps and pumping equipment....... .......................... Transportation equipment, except automobiles.................. Aircraft................. .............. ............................................ Aircraft parts, including engines.................................... Shipbuilding and repairs................................................ Automobiles......... ................................................................ Motor vehicles, bodies, and trailers............................... Motor-vehicle parts and accessories............................... Nonferrous metals and their products.................................. Primary smelting and refining, except aluminum and magnesium. Rolling and drawing of copper and copper alloys......... Lighting equipment........ ............................................... Nonferrous metal foundries, except aluminum and magnesium. Lumber and timber basic products................................... . Sawmills............... ...... .................................................... Planing and plywood mills............................................. Furniture and finished lumber products.............................. Furniture, including mattresses and bedsprings........... Do. January 1930.a January 1943. Do. Do. January 1939. January 1943. Do. Do. September 1940. January 1932. September 1941. October 1937. September 1941. September 1939. January 1943. Do. Do. Do. July 1937. January 1943. Do. Do. Do. 1937.7 January 1943. Do. Do. Do. Do. 1937.7 January 1943. 1937.7 January 1943. January 1930. Do. January 1943. Do. Do. September 1941. January 1943. Do. January 1930.* September 1939. January 1943. April 1930.* 1 For a comparison of the industry titles used before and starting January 1943, see June 1943 Monthly Labor Review, p. 1210. 2 Dates refer to month of reference of the data. In most cases, the series was first published in the monthly Labor Turn-over Report and the Monthly Labor Review in which data for the specified month were published. In cases where the series was prepared retroactively, a footnote indicates the earliest published source. * Rates for 1930, revised to use arithmetic mean instead of median, were first published in the July 1937 issue of the Monthly Labor Review, reprinted as Serial No. R. 608. 4 Published currently starting September 1946. Mimeographed summary sheets show data monthly from January 1943. Earliest date published2 M A N U F A C T U R IN G —Continued M AN UFACTU RIN G All manufacturing industries. Durable goods--------------Nondurable goods............ Industry group and industry1 Durable goods—Continued Stone, clay, and glass products........................................... Glass and glass products..------ ------------- ---------------Cement.......... ................. ........................... ................. Brick, tile, and terra cotta._______ _________________ Pottery and related products....................................... Nondurable goods Textile-mill products......................... ................................... Cotton.__________________ _______ ______ ________ Silk and rayon goods___________ _____________ ____ _ Woolen and worsted, except dyeing and finishing....... Hosiery, full-fashioned......... .............. ...................... . Hosiery, seamless................. ........................................ . Knitted underwear...................................................... Dyeing and finishing textiles, including woolen and worsted. Apparel and other finished textile products....................... . Men’s and boys’ suits, coats, and overcoats.... ............ Men’s and boys’ furnishings, work clothing, and allied garments. Leather and leather products............................................... Leather_____________ _______ ____________ ________ Boots and shoes..... ......... .......... ............ ................... Food and kindred products....................... ........... ............ Meat products.............................. ......... ....................... Grain-mill products............ .......................................... Bakery products..................................... ............ .......... Tobacco manufactures.............. ............................................ Paper and allied products......... ............................................ Paper and pulp....................................... ...................... Paper boxes....... ............................ .......... .................... Chemicals and allied products.____ _______ _______ _____ Paints, varnishes, and colors......... .......... .................... Rayon and allied products-------------------- ----------------Industrial chemicals, except explosives____ _________ Products of petroleum and coal.......................................... Petroleum refining.............. ................................... ...... Rubber products___________________ ____ _____________ Rubber tires and inner tubes....................................... Rubber footwear and related products...... ................... Miscellaneous rubber industries................................... Miscellaneous industries...................................................... NONM ANUFACTURING Metal mining............. Iron-ore_________ Copper-ore.......... . Lead- and zinc-ore. Coal mining: Anthracite........... Bituminous-....... . Communication: Telephone............. Telegraph.............. January 1943. December 1937. April 1937. April 1931. January 1943. January 1943. January 1930.* May 1938. October 19.36. January 1943. Do. Do. May 1939. January 1943. Do. Do. Do. September 1941. January 1930.* January 1943. January 1930.* September 1941. January 1949. January 1943. Do. July 1938. January 1941. January 1943. May 1938. August 1936. 1940.7 January 1943. May 1931. January 1943. January 1931. December 1937. January 1943. Do. March 1943. April 1943. March 1943. Do. February 1943. January 1943. June 1943. May 1943. 8 Called iron and steel prior to May 1942. 8 Prior to January 1943 “ machine-tool accessories” were included with “ machine tools.” 7 Annual rates from specified year through 1941 were published in May 1942 Monthly Labor Review, reprinted as Serial No. R. 1463. Monthly rates were published currently starting December 1937 for machine tools, from January 1939 for aircraft and shipbuilding, and from September 1940 for industrial chemicals, except explosives. Rates for industrial chemicals, except explosives are available in mimeographed form from January 1939, however. 24 TECHNIQUES OF PREPARING MAJOR RLS STATISTICAL SERIES Besides the regular series shown in table 2, others were prepared from time to time, particu larly during World War II and the immediate postwar period, in order to (a) highlight the labor changes in war industries; (b) compare the rates for men and women; (c) to measure military separations during the war; and (d) to measure the rate of absorption of veterans into manufac turing and mining employment after the war. A list of these series and the periods for which they are available are shown in table 3: T able 3.— Special industries and groups for which labor turn-over rates were published during and immediately following World War I I Period1 Group All employees: Selected war industries _ _ _ _ _ _ __ Do............................................................ Munitions______________________________ N onmunitions_________________ ______ __ Munitions ___ ______ ____ ________-______ Nonmunitions_____________ ____________ Men and women: Manufacturing_________ ___ ___________ Do_________________________________ Do ................................................ . Selected war production industries_______ D o .................................................... ...... Munitions _ Coverage Publication * Total_____________________________________ _ Selected individual industries_________________ Total.... ......... ......... ............................................. ____ do........ ............................................................. ____ do.......... ........... ............................................... ....... do......................................... ........................... ____do..................................................................... Durable and nondurable-goods divisions_______ Industry groups............. ........................ .............. Selected industry groups_______ ______ _______ Selected individual industries_________________ Total___________ __________________________ Nnnmnnitions ____ d o ______________________________________ Munitions. ____________________________ ____do............... ..................................................... Non munitions _______________________ ____ do.......................................................... ......... Veterans:5 ____do______________________________________ Manufacturing _ ___ _____________ Do................... ................. ................... - Durable and nondurable-goods divisions_______ D o__________ __________ ____________ Industry groups.................................... ............... N onmanufacturing______________________ Anthracite and bituminous-coal mining............. Manufacturing__________________________ Total...... ............................................................... D o ........................................................... Durable and nondurable-goods divisions_______ Do......................................................... . Industry groups______ ______________________ D o................... ........................................ Individual industries_________________________ i Dates refer to month of reference of the data. * M LR = Monthly Labor Review; LTOR = Monthly Labor Turn-Over Report. * Published in a special article, Labor Turn-Over in Munitions and Non munitions Industries, 1943 and 1944, in July 1945 issue of the Monthly Labor Review; also reprinted as Serial No. R. 1757. * Data for November 1945 through November 1946 also were published in a special article, Postwar Labor Turn-Over Among Women Factory Workers, From Through October 1944___ January 1942___ January 1943___ ____do................. January 1945___ ....... do....... ......... December 1944__ June 1943 __ _ December 1944_. ....... do. —........... December 1945__ ....... do................. March 1944 _ October 1945___ August 1945____ June 1943____ _ ____ do_________ January 1944___ ____ d o ______ _ August 1945____ ____ do................ December 1945 ....... do................ ____ do................ ____ do................ ____ d o...... ......... ....... do................. ....... d o ................ July 1946............ July 1947 _ _ ___ d o ____ _ _ August 1945- _ _ ____do....... .......... December 1944__ do .. . December 1945__ ____do_________ M LR ; LTO R. M LR ; LTO R. M LR .i*6 M LR .6 M LR ; LTOR. M LR ; LTOR. M LR *; LTOR. M L R *; LTOR. M LR *; LTOR. M LR ; LTOR. M LR ; LTOR. MLR.3 MLR.3 M LR ; LTOR. M LR ; LTOR. June 1948 _____ ....... do....... ......... Special release. Do. Do. ____do_________ Do. July 1946_______ MLR.® M L R .6 ____ d o _______ _ MLR.« ____ do_________ M L R .6 do in March 1947 issue of the Monthly Labor Review; also reprinted as Serial No. R. 1880. s Total accession rates were not published. Separation rates were shown as a total, and by quits and involuntary separations. Employment and accessions of veterans were shown as percentages of the respective totals. • Published in a special article, Veterans Return to the Nation’s Factories, in December 1946 issue of the Monthly Labor Review. V. Compiling Monthly and Weekly Wholesale Price Indexes Both a monthly comprehensive wholesale price index and a weekly index are computed by the United States Labor Department’s Bureau of Labor Statistics. These indexes are described separately in the discussion which follows. programs. The indexes are widely used as a base for adjusting prices in contracts extending over a period of time— “ contract escalation.” 20 Other uses are for forecasts of future pricing, marketing, and sales policies based on historical price develop ments; for analysis of the price structure of the economy and changing interrelationships among individual commodities; and in combination with other data, for formulation of long-range economic and other programs by Government officials. Comprehensive Monthly Index Since 1902, the official monthly wholesale price index has been prepared as an indicator of general price trends and average changes in commodity prices at primary market levels. The official series has been carried back to 1890; separate monthly indexes are available for major groups of commodities from January 1890, and for sub groups of commodities beginning with 1913. The composite index has been estimated, jointly by the Bureau and by non-Governmental agencies and individuals, on an annual basis back to 1749. Currently the index contains prices of nearly 900 commodities classified into 10 major groups and 49 subgroups. The commodities are also classified into 5 special economic groupings. In general, the prices covered are those quoted by manufacturers or producers or those prevailing on commodity exchanges and in organized markets. The index gives an approximation of changes in the general price levels of commercial commodity transactions. “ Wholesale” refers to large-scale or wholesale quantities and terms in contrast to “ retail.” The major uses of the index, its components, or the individual price series are numerous. They provide one tool in the analysis of general business conditions and of the business cycle. In con junction with other data, they indicate the state of balance between aggregate supply and demand at the primary market level. Individual concerns utilize the indexes in planning plant construction and production schedules, in purchasing raw materials, in evaluating inventories, in determin ing production costs and in general investment 884616— 50------- 5 Index Limitations. The index measures completely the price changes of only the 900 commodities (and others that are closely related) which are specifically defined for pricing both physically and in terms of market structure. It is not an over-all measure of the “ general price level” or of “ the purchasing power of the dollar” — it does not in clude all classes of commodities (real estate, securities, services, etc.) which are factors in the “ general price level.” Finally, it does not cover transactions at all levels of marketing— retail, jobbing, speculative trading, export, etc. It is not a measure of prices charged by wholesalers (i. e., jobbers or distributors) and cannot be used to measure the margins between cost and selling prices. All price quotations used in the index have been selected to be representative of the direction and degree of price change. They do not necessarily measure the average dollars-and-cents levels of prices of individual commodities. Except for a few commodities, the prices used in the compilation represent those prevailing in national markets and not those effective in any specific locality. Only a few commodities, such as butter, brick, and some 20 The indexes for any given month are published four times: (1) in mimeo graphed form about 25 days after the close of the month to which the index refers and in the Monthly Labor Review dated 2 months later than the month to which the index relates, (2) a month after the first printing, cor rected for late reports and revised prices, (3) 2 months after the first printing with further corrections for late reports and revised prices, and (4) the middle of the calendar year following the month of reference, including all corrections and price changes which are reported up to May 31 of the year in which the final printing is made. The Bureau suggests that if any one of the indexes is used in contract escalation, after proper consideration of the relevance of the index (or the group index) to the costs being adjusted, the figure supplied in the third printing should be used in most cases. 25 26 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES agricultural products, are priced locally in various cities. Method of Collection. A single quotation is used for the majority of the slightly less than 900 commodi ties which are priced for the monthly comprehen sive index. This is practicable because for many articles, the price movements of various sellers are closely related. However, quotations from 2 to 50 sellers are averaged for about 150 commodities. In all, a total of 1,600-2,000 quotations go into the index each period.21 The Bureau attempts to price goods of constant quality from period to period so that the index will measure only the effect of price changes. To ac complish this, detailed specifications have been written for the items priced. Each specification has been prepared, in cooperation with experts from industry and trade, to be as typical as possible of the price movements of all specifications for the commodity. Two examples of specifications are: Sheets, steel. Box annealed, #27, U. S. Standard, hotrolled, per pound, bulk, carlots, f. o. b. mill; market price, weekly (Tuesday); prior to June 30, 1931, one pass, cold-rolled. Shirts, men’s dress. White broadcloth, 136 x 60 count, with or without collar, fine combed yarn, full shrunk, cellophane wrapped. Per dozen, pack aged, no quantity specified, f. o. b. New York; manufacturer to jobber; transaction price subject to current cash discount, weekly (Tuesday); prior to 1932, 128 x 68 count. Prices are collected on each specification by mail on schedules from individual reporters or obtained directly from trade journals or organized markets each week or month. All reports are made volun tarily upon the request of the Bureau; field agents only visit new reporters or revisit current reporters to clarify questions on the schedules. When a new specification or reporter is introduced, the commodity price information sheet shown below is filled out by the field agent. This sheet contains all pertinent information regarding the product and the reporter, in order to insure comparability over a period of time. After the initial personal collection (including, when applicable, prices for previous dates), all collection is made by mail on a 21 In the event that a price for a specific commodity is not available for a given month at the time an index is computed, an estimated price is used. The estimated price is based upon the best available information. If an actual price becomes available at a later date, the indexes are revised to take account of any difference between the estimated price and the actual quotation. simple schedule form which has space for the reporter to enter his current price, any changes in the selling structure including changes in discounts, and an historical record of his previous reports. Depending upon the individual commodity, these schedules are sent out weekly, or monthly, but provision is always made for taking account of multiple price changes within the month. Sources of Data. Manufacturers (or producers or their agents), commodity exchanges or organized markets, Government agencies, and authoritative trade journals are the sources of the basic price quotations used in the index. Trade journal quotations are used only when competent authori ties recognize the individual journal as a reliable price reporter or source. As of May 1949, the approximate distribution of commodity reports by source was : Percent of total Total___________________________________ 100 Manufacturers or sales agents___________ Trade journals__________________________ Boards of trade, commodity markets, etc_ Federal or State agencies________________ 57 38 2 3 Calculation Procedures. The wholesale price index is calculated as a fixed-base weighted aggregate, using prices in 1926 as 100. Quantity weighting factors are based on market sales during the years 1929 and 1931, except for agricultural commodi ties. For agricultural commodities, average mar ket sales in the years 1929, 1930, and 1931 are used. Each price used in the index applies to 1 day each week, although the day varies by com modity. The monthly price is an average of the four or five 1-day-a-week prices which fall within the calendar month. Indexes are computed by subgroups before the comprehensive index is compiled. The first step is to compute a monthly average price for each commodity (including average prices from more than one reporter for an individual com modity). This average price is then multiplied by its appropriate quantity weighting factor to give a value aggregate for the commodity for the pricing period. All of the value aggregates in a subgroup are then totaled and divided by the sum of the value aggregates at 1926 prices. The result of this division (multiplied by 100) yields a subgroup index on a 1926 base. Similarly, the subgroup aggregates are totaled to obtain a group 27 WHOLESALE PRICE INDEXES U. S. D E P A R T M E N T OF LABOR Bureau of Labor Statistics B. L. S. 1810 (Rev. 1-1-49) Budget Bureau No. 44-R602.1 Approval expires 12-31-50 W A S H IN G T O N 25, D. C. Code No ... Commodity Commodity Price I nformation Sheet 2. a. Mfr. □ 1. a. Firm name. b. Other (Specify) b. Address (Postal zone) (Street) Title___________________ Title___________________ Title___________________ (Informant) Address______________________ . (City and State) _________________________ _________________________ _________________________ c. Information authorized by d. Information furnished by_ e. Schedule to be mailed to _ _ (Postal zone) (Street) (City and State) 3. Description of commodity: Mfr’s a. Style N o_______________ ___________ b. Lot N o_______________ d. Grade_________ _________ e. Brand name____________ c. Model N o____________________ ______ No. □ f. Additional___________________ ---------------------------------- g. Check sheet attached: Yes □ 4. a. Price quoted is from____________________________________________ t o -------------------------------------------------------------------(Class of seller) (Class of purchaser) b. Type of quotation: (1) List price___________________ __ (2) Actual transaction price-------------------------------------(3) Exchange price_________________ (4) Other (specify)______________________________________________________ c. Unit quoted____________________________________________________________________________________________________ d. Minimum and maximum size of sale to which price applies---------------------------------------------------------------------------------e. Customary delivery period______________________________________________________________________________________ 5. Discounts applicable to prices for the commodity described above when sold to the class of purchaser listed in question 4a. (Circle any of these discounts which have been deducted in arriving at the prices listed in question IS.) a. Trade__________ % : b. Quantity__________ % on purchases o f_________________________________________________ c. (1) Cash discount terms__________________________________________ (2) Extent of use_______________________ % d. Other discounts, allowances or free deals (explain in detail)______________________________________________________ 6. Duties or taxes applicable to prices listed under No. 15 which have: (a) been included (specify)________________________________________________________________________________________ (b) not been included (specify)____________________________________________________________________________________ 7. Usual method of effecting price changes: (a) Change in list price___________________________________ (b) Change in discounts--------------------------------------------(c) Other (explain)________________________________________________________________________________________________ 8. a. Delivery terms (f. o. b., freight equalized, etc.)__________________________________________________________________ b. Usual means of delivery (carrier)_______________________________________________________________________________ 9. Customary type of packaging used. a. Crate----------------------- b. carton----------------------- c. bag---------------------------d. other (specify)__________________________________________________________________________ _______________________ 10. a. Does price in No. 15 include charge or deposit for packaging? (explain)--------------------------------------------------------------_____________________________________________________________ b. Is any part of such charge or deposit refundable? (explain)__________________________________________________________________________________________________________ 11. Channels of distribution, percentage of sales made, discounts and other allowances by type of distribution for com modity described in No. 3 above. Sales and discounts Other Mfr. (assembler) Wholesaler Jobber Distribu tor Retailer Consumer Other (specify) a. Sales (percent)________________ b. Cash discount______________ c. Trade discount_______________ d . O n an t i t v d iscou n t, 12. Market area served 13. Method of selling (i. e., salesman, catalog, etc.)----------------------------------------------------------------------------------------------------14. Major products manufactured or distributed by this firm and relative dollar volume importance expressed in percent age of total sales. e. a . -------------------------- - - - % c. % f. b . _____________________ % d. % -% 15. Prices for commodity described in 3, 4 and 5. Date 16. BLS Agent. Price Remarks Price Date Date. Remarks 28 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES aggregate and index, and the group aggregates are totaled to obtain the all-commodities index. The individual commodity aggregates are retotaled to obtain the five economic group indexes. Major changes in specifications of commodities, shifts in the relative importance of sales to differ ent types of purchasers or by different types of sell ers, alterations in the distribution pattern of the industry, or changes in commodities with economic conditions are handled in each instance by “ link ing” so as to prevent movement in the index for the period in which the change is made. First, a “ link date” is selected; this is the month chosen as being the most logical in which to make the shift, taking into account the market conditions for the individual commodities. Then, in the link-month, the quantity weighing factor is adjusted so that the product of the price of the new specification times the adjusted weighting factor is the same as the product of the price of the former commodity priced times the unad justed weighting factor. Because of the many links for individual commodities made after the adoption of the 1929-31 weights, the original quantity weighting factors are no longer mean ingful as quantities and are currently called “ multipliers.” Weekly Index From 1932 to November 18, 1948, the coverage of the weekly wholesale price index was the same as for the monthly index; in November 1948, it was replaced by a new weekly series which includes only 115 commodities or a sample of about an eighth of the commodities in the monthly index. The current series was designed as a counterpart of the monthly index;22 specifically,it is intended to show week to week changes in commodity prices, for interpolating between successive monthly indexes, and to provide an estimate of the level of the comprehensive index 2 to 3 weeks in advance of its publication.23 Regular publication of this weekly index was begun on November 19, 1948, and indexes begin ning with January 1947 were published at that time. Indexes are published each Friday for the week ending the previous Tuesday for all com 22But it should not be used for escalation since it is only an estimate of the comprehensive index. 23 A detailed description of the current weekly index and its advantages over the comprehensive weekly index was presented in the Monthly Labor Review for September 1948 (p. 290). modities; all commodities except farm products and foods; and for six of the major groups in cluded in the monthly series. The four major groups of the comprehensive index not published separately are combined to form an “ all other” index. In addition, special subgroup indexes, in cluding the entire comprehensive sample, are published weekly for grains, livestock, meats, and hides and skins. In October 1949, the full sample for these latter four subgroups was included in the weekly index and other changes were made in the sample in order to obtain a closer approxi mation of the monthly index or because of demand for certain subgroup or group indexes. As a result, about 300 commodities are currently included. For the first time, a weekly index for chemicals and allied products is currently published as a major group and the “ all other” group index has been discontinued. Additional changes in the sample will be made when revised subgroups are introduced into the monthly index. Method of Collection and Sources.— Inasmuch as the weekly index is based upon a sample of the commodities and reporters used in the compre hensive index, the sources and methods of collec tion of the data are identical to those used in the comprehensive index (see p. 26). Calculation Procedures. The weekly wholesale price index is calculated as a fixed-base weighted aggregate with 1926 prices as 100. The constant weights or “ multipliers” for individual commodi ties are derived from value aggregates of the comprehensive monthly index for the year 1947. All 900 commodities of the comprehensive index are represented in the sample either directly or indirectly. Thus, each commodity in the sample is assigned its 1947 value aggregate, plus those of non-sample commodities, which have similar price trends. To obtain the “ multiplier,” the total aggregate assigued to each sample commodity is divided by the 1947 average price for the priced commodity. These multipliers are applied to the weekly prices of the sample commodities and their products are totaled for each group. Major changes in specifications for commodities are treated in the same manner as in the compre hensive monthly index (see above). However, in the weekly index, the “ link” is made on the basis of a week instead of a month so as not to distort the percentage change from week to week. VI. Preparation of Union Scales of Wages and Hours Series 24 Annual studies of union scales are conducted by the United States Department of Labor’s Bureau of Labor Statistics in five industries: baking, build ing construction, local transit, local trucking, and printing.25 Union scales are defined as minimum wage rates or maximum schedules of hours agreed upon through collective bargaining between em ployers and trade-unions. Rates in excess of the agreed minimum, which may be paid because of long service, for special qualifications, or for other reasons, are excluded from the studies. The use of union agreements or other union records in studies of occupational wages is practic able in industries that are widely organized and in which (1) defined craft groupings persist, as in building construction or printing, or (2) key occu pations can be clearly delineated, as in local transit. The Bureau’s annual union wage studies began in 1907. Originally information was obtained for 39 cities but the number has been gradually expanded until in 1949, 77 cities are being covered, ranging in population from 40,000 to over 1 million. The scope of the information for individ ual industries has also been expanded. For example, all branches of baking are covered currently in place of the bread-baking branch only, and 11 book and job and 8 newspaper occupations in printing replace 7 book and job and 4 newspaper occupations formerly surveyed. The Bureau’s union-wage series measure inter city wage differences for comparable work, and the relationships between rates applicable to workers in occupations requiring varying degrees of skill. The data are used widely in wage negotiations by both management and labor. The scales of building-trades workers are especially important 34 Prepared by Charles Kubenstein in the Bureau's Division of Wage Statistics. 25Prior to 1936, the coverage at various periods also included barbers, line men, longshoremen, and workers engaged in breweries, laundries, metal trades, millwork, restaurants, soft-drink production, and theaters. in estimating construction costs, because labor expenditures constitute an important element in the total cost of building construction. The index series derived from these studies provide barom eters of year-to-year changes in scales of wages and hours in the industries covered. The indexes of union-wage scales and weekly hours have a base period of June 1939, a year which was not marked by any unusual fluctuations. The index series for the building trades and print ing go back to 1907, for local transit to 1929, for local trucking to 1936, and for baking to 1939. Although data for the latter three industries were collected for years prior to the dates of the index series, indexes were not constructed because of inadequacies in the available data. The study of union scales in the baking industry includes all occupations in the industry, except delivery drivers and plant maintenance workers. With the exception of the indexes which relate to the industry as a whole, data are presented by industry branch or type of baking, e. g., bread and cake hand shops, bread and cake machine shops, pie and pastry shops, cracker and cooky shops, Hebrew baking, and other nationality baking such as Bohemian, French, Italian, and Polish. In the building trades, virtually all journeyman and helper and laborer classifications are covered. Indexes and other data are shown for each im portant trade as well as for all trades combined. T able 4 . — Recent Item coverage of union wage studies in 77 cities Bak ing Build Local ing trades transit Local truck ing 1.500 90 Number of local unions.............. 150 600 Number of scale quotations........ 3,700 3.500 400 3,300 Number of workers..................... 70,000 760,000 100,000 270,000 33 Percent of entire industry1____ 30 40 0) Print ing 700 2,500 100,000 30 1 Entire industry includes both union and nonunion workers. 2 Data not available. The trucking study embraces motortruck drivers and helpers engaged in local trucking. Over-theroad drivers and local city drivers paid on a mileage or commission basis are excluded. All data, including indexes, are presented for the two classifications indicated. 29 30 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES Union scales in the local-transit industry are limited to operating employees of busses, surface cars, and elevated and subway lines. Data, with the exception of indexes, are shown separately for operators of 1-man cars and busses, motormen and conductors of 2-man cars, and elevated and sub way lines. In the printing industry, besides 11 book and job trades and 8 newspaper trades, the data for newspaper trades are further broken down for day and night work. Indexes and other data are presented separately for each trade and for all trades combined. Total union and worker coverage of the series is given in table 4. Limitations of the Series The union wage series are generally designed to provide indexes of union hourly wage scales and weekly hours of work at straight-time pay, and of scale levels by trade, city, and region. The indexes measure the trend of union scales but do not portray the movement of earnings or takehome pay and actual hours of work. Because of irregularities in the work schedules of operating employees in many of the covered cities, an index of weekly hours has not been maintained for the local transit industry. In the baking industry, the occupational structure varies between cities, especially in mechanized plants which operate on a mass-production basis. Consequently, scale levels are presented by industry branch, irrespec tive of occupation or sex of workers. Scales for apprentices are not included in any of the union wage studies. Scale levels are average union rates which pro vide comparisons of wage rates between industries, trades, and cities at a given time. They are not an accurate measurement of year-to-year changes because of fluctuations in membership and other factors. Membership figures for the various trades or classifications do not remain constant and any changes have a marked effect on scale levels. For example, if organizational drives in cities having relatively lower scales of wages result in sharp increases in membership, the movement of the scale levels for the affected trades as a whole is naturally retarded. Conversely, increases in membership in cities having high wage scales accelerate the upward movement of scale levels. A similar effect can be caused by decreases in union membership. The absence of effective wage scales for local unions, because of protracted work stop pages, may have the same influence on scale levels as fluctuations in membership. In addition, the union rates are not necessarily the actual rates paid to all workers, nor are the union hours the hours actually worked. The union scale usually fixes the minimum rate of pay and the maximum hours of work at straight-time pay. Workers with above average experience and skill may be employed at rates above the union scale, especially during prosperous times when a tight labor market creates competitive bidding for the better workmen. During periods of depressed business activity, actual hours worked are often less than those specified in the union agreement. Study Methods and Sources The mail questionnaire technique is used and information is collected from central sources, such as international u-nions and regional union organi zations. Personal visits are then made to local unions which do not respond to the mail question naire or for which data are not available from central sources. Prior to 1947, all data relative to union wage studies were collected directly from local union officials (generally the secretary or business agent) by Bureau representatives and entered on schedule forms designed specifically for this purpose. Information is requested for a specific date, namely July 1, for all industries except local transit, which is surveyed as of October 1. These dates were adopted after numerous changes, be cause the large bulk of new agreements involved have been negotiated by that time each year. In order to maintain year-to-year comparability, scale and membership data for the previous year are transcribed onto the schedules before they are sent out. Union officials are requested to check the previous year’s data and revise any figures which may have been incorrectly reported, before they insert current data. Copies of union agree ments are also requested from union officials for the purpose of (1) checking the data entered on the schedules with the terms of the agreements and (2) building up the files of union agreements maintained by the Bureau of Labor Statistics. The schedule used for the building trades is here reproduced. 31 UNION WAGES AND HOURS B. L. S. 1150.1 (Rev. April 1949) Budget Bureau No. 44-R-738.1 Approval expires 6-30-1950 SCHEDULE No.................... U. S. D epartment of Labor BUREAU OF LABOR STATISTICS Union U nion Scales of W ages and H ours in the B uilding T rades (annual survey) City State Please enter in: Column 10— Number of union members working or Column 4— Wage rate for each trade listed. immediately available for work at each Column 7— Weekly hours to be worked before overrate. time rate becomes effective. Column 11— Number of Apprentices in each trade. Membership information will be kept confidential and used only to compute average wage rates. nun Trade or occupation Code (1) Title (2) Scale of wage rates in effect on 7-1-48 (3) 7-1-49 (4) Code (5) Weekly hours before overtime rate is effective 7-1-48 (6) 7-1-49 (7) Code (8) Number of Number of active union apprentices in each members at each rate trade 7-1-48 (9) 7-1-49 (10) 7-1-49 (ID Please attach a copy of your agreement in effect on July 1, 1949 and answer the following questions: 1. How many hours must be worked each day before overtime rate is effective?_________________________ 2. When did your agreement go into e ffect?_____________ When may it be reop en ed ?______ When does it e x p ir e ? ________ Do you desire agreement to be kept confidential? Y e s ____ N o ____ 3. What is the total membership of your u n ion ?________ How many are apprentices?________________ Please sign your name here 4. D o apprentices receive classroom or equivalent instruc tion in subjects related to their trade? Yes __ No __ 5. Between July 1, 1948 and July 1, 1949 how many apprentices completed their apprenticeship?________ 6. Between July 1, 1948 and July 1, 1949 how many journeymen became unavailable for work because of death, permanent disability, or retirem ent?________ D a t e __________ , 1949 A ddress____________________________________________ (Street and No.) The union wage studies are designed to include all local unions in the covered industries in the selected cities. Periodic checks are made with central labor unions, district councils, and other authoritative bodies to obtain new local unions which should be included in the studies. Computation Procedures Chain indexes are calculated for each of the five industries to portray the trend of union rates and weekly hours at straight-time pay. In calculating these indexes, the percentage change or ratio from year to year is based on aggregates computed from quotations for all identical classifications in each industry for two successive years. To obtain the aggregates, the rates and hours for both the previous and current years are weighted by the (City and State) (BLS 49-4056) membership in the particular classification for the current year. The index for the current year is computed by multiplying the index for the pre ceding year by the ratio of the aggregates. For example, in the 1948 study of building trades, the rate aggregate for all quotations amounted to 1,620,303.472 for the current year and 1,465,426.274 for the previous year. The ratio of these aggregates is 110.57 and was obtained by dividing the figure for the current year by the one for the previous year. The July 1, 1948, index of union hourly wage rates for all building trades (163.5) is the result of multiplying the July 1, 1947, index (147.9) by the ratio of the aggregates (110.57). This method of index cal culation eliminates the influence of year-to-year changes in membership. 32 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES Indexes of union hourly wage rates and weekly hours are computed for each classification as well as for all classifications combined in the building construction, printing, and trucking industries. In the baking and local transit industries, indexes are provided only for all classifications combined. Irregular hours of work for operating employees in many of the covered cities prevent the compu tation of an index for union weekly hours in the local transit industry. Average union rates are calculated by weighting each quotation for the current year by the reported membership. These averages are levels designed to provide comparisons between trades and cities at a given time. They do not measure the trend of union rates, the function served by the index series. An over-all average hourly rate is computed for each of the industries included in the union wage studies. In addition, averages are presented by industry branch and city in baking; by trade, city, and region in building construction and printing; for three broad occupational groupings by city in local transit; and by classification and city in trucking. Union scales of wages and hours in effect on the date of the survey, as reported by union officials, for both the previous and current years are published for each classification by city. This furnishes a direct comparison of union scales between the two years for each of the industries studied. The scales of wages are indicated as hourly rates and the scales of hours as the weekly hours of work before overtime rates are applicable. VH. Measurement of Industrial Employment26 For many years, the U. S. Labor Department's Bureau of Labor Statistics has collected and pub lished monthly statistics relating to industrial employment, wages, and hours of work as an integral part of its employment-statistics program. In the present article, the collection and compila tion of the employment statistics is described; in the article on the calculation of the average wages and hours is explained. These series provide a detailed and comprehensive body of information on industrial activity for the country as a whole and, on a more limited basis, for States and im portant localities. Continuous national series are published each month on the number of wage and salary workers employed in all nonagricultural establishments and in the 8 major industry divisions: Manufac turing, mining, contract construction, transporta tion and public utilities, trade, finance, service, and government. Both all-employee and produc tion-worker employment series are also presented for 21 major manufacturing groups and 104 sepa rate manufacturing industries. Within nonmanu facturing, total employment information is pub lished for 23 major groups and 13 component industries. Production-worker employment is also shown for the mining division as a whole and for the industry components. Series covering employment in manufacturing industries are also currently published for all States; employment in all nonagricultural estab lishments and in each of the major industry divi sions is available for about 30 States. Expansion of the program, now underway, is designed to produce similar series for all States and for about 100 major metropolitan areas. For private nonagricultural industries, the employment information covers all full- and parttime employees who were on the pay roll, i. e., who worked during, or received pay for, the pay 86 Prepared by Sydney S. Netreba of the Bureau's Division of Employment Statistics. period ending nearest the fifteenth of the month. For Federal establishments, the employment period relates to the pay period ending prior to the first of the month; for State and local governments, during the pay period on or just before the last of the month. Proprietors, self-employed persons, domestic servants, and unpaid family workers are excluded. Limitations of the Data Essentially, the Bureau of Labor Statistics measure of employment represents a count of persons on the pay rolls of nonfarm establishments during one pay period. For most establishment reports, the pay-roll period covers 1 week in each month, generally the week ending nearest the fifteenth of the month. Such a pay-roll count includes persons who worked during, or received pay for, any part of the reporting period. The employment series, therefore, reflect turn-over of personnel: The same person may appear on two separate establishment pay rolls in the same period. Thus, the employment information is not a measure of the number of full-time jobs that were available, nor is it an unduplicated count of paid workers. Owing to the restrictions with respect to the reporting period, the data do not refer to employment throughout the month, nor to employment at one point of time in the month. Since the guiding principle of the Bureau's employment concept is that of work during, or receipt of payment for, a given period, distinctions must be made with respect to persons who may have jobs but who are not at work for various reasons. Thus, persons on paid vacations or paid sick leave are included in the employment count, but those on leave without pay during the report ing period are not considered as employed. Work stoppages in effect throughout the reporting period result in the exclusion from employment of those so engaged; however, the employment level is not affected if persons involved in work stoppages worked or received pay in any part of the pay period covered. Similarly, persons who are laid off or who leave temporarily are excluded from the employment count if such personnel ac33 34 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES tions cover the entire reporting period and no payment is made. The problems of classification and the limita tions of the classification structure also affect the employment data. Thus, in most instances it is not possible to provide detailed employment information for specific products. The require ment that establishments shall be classified accord ing to major product or activity (as discussed in the following section) may result in the “ conceal ment” or understatement of employment in other important activities, and in a simultaneous over statement of employment in the major activity. Survey Sources and Methods Approximately 120,000 cooperating establish ments furnish monthly employment and pay-roll schedules, by mail. The number of establish ments reporting employment data and the num ber of employees covered, for each industry division, are shown in table 5. T able 5.— Approximate coverage of employment and pay roll sample Division or industry Employees Number of establish Number in Percent ments sample of total Mining____________ ____ _______ _ Contract construction......................... Manufacturing____________________ Transportation and public utilities: Interstate railroads (IC C )___ _ Rest of division (BLS) _______ _ Trade________________________ __ Finance_____ ______ ______ _________ Service: Hotels..... ................ . ___............ Laundries and cleaning and dye ing plants___________________ Government: Federal (Civil Service Commis sion)___________ _____________ State and local (Bureau of Cen- | sus. nnart.ftrlv') ' 1 2, 700 15,000 35,200 460.000 450.000 8,845,000 47 23 62 10,500 46, 300 6,000 1, 359,000 1, 056, 000 1,379, 000 281.000 98 41 15 16 1,200 115, 000 25 1,700 86,000 17 1, 885, 000 100 2, 400,000 62 With respect to employment, the following in formation is obtained: (1) For manufacturing, mining, laundries, clean ing, and dyeing: (a) all employees or the total number of wage and salary workers, i.e. all fulland part-time employees in the respective estab lishments who worked during or received pay for any part of the period reported; salaried officers of corporations are included, proprietors or firm members are excluded; (b) the number of all full- and part-time production and related workers on the pay roll who worked during or received pay for any part of the pay period reported; persons on paid sick leave, paid holidays, and paid vacations are, therefore, also included. (2) For other nonmanufacturing industries, i.e. wholesale and retail trade, public utilities, finance, hotels, and miscellaneous: (a) the total number of full- and part-time wage and salary workers as indicated above; (b) the number of full- and parttime nonsupervisory employees and working supervisors; similarly persons on paid sick leave, paid holidays, and paid vacations are included. Cooperating State Agencies. The current employ ment statistics program is an integrated FederalState project which provides industrial employ ment information on a national, State, and area basis. In accordance with authority granted in a Congressional Act of 1930 (Public Laws, Chapter 873, Yol. 46, Part I) and in order to minimize the reporting requirements for cooperating establish ments, the Bureau of Labor Statistics has entered into agreements with various State agencies. Basic uniformity in the collection of BureauState employment statistics is obtained by the use of Bureau-designed schedules for monthly report ing. The “ contract” State agencies edit these reports according to standards defined in editing instructions issued by the Bureau, and make these data available for inclusion in the national esti mates. The cooperating agencies are responsible for the preparation of State and area employ ment statistics. To maintain comparable data from State to State, statistical standards are set forth in a BLS-State procedures manual and in related instructional memoranda. A list of the cooperating State agencies, currently covering all States and the District of Columbia, is presented in the monthly mimeographed Employment and Pay Rolls Detailed Report and in a footnote to table A-10 of the Monthly Labor Review begin ning with the October 1949 issue. Calculation Procedures In the employment series (as well as those on hours and earnings), reporting establishments are classified into significant economic groups on the basis of major postwar product or activity as determined from annual sales data. The industry classification structure currently used in the em ployment statistics program is defined in the folio wing documents: (1) For manufacturing indus tries— Standard Industrial Classification Manual, MEASURING INDUSTRIAL EMPLOYMENT volume I, Manufacturing Industries, Bureau of ^he Budget, November 1945; (2) for nonmanufacturing ndustries— Industrial Classification Code, Federal Security Agency, Social Security Board, 1942. Estimation Procedures. Current employment data ire estimates based upon reports from a sample *roup of establishments, since in many industries Pull coverage would be prohibitively costly and time-consuming. To obtain the estimates the following four steps are necessary: (1) A total employment figure (bench mark) for an industry, as of a specified period, is obtained from sources which, singly or in combination, insure either a complete count of employment for the industry or an estimate of reasonable accuracy. (2) Employment data for a sample group of reporting establishments are compiled for the bench-mark month and subsequent months. (3) The ratio of employment in one month to that in the preceding month (i. e. the link relative) for identical establishments in each industry, is computed for each consecutive pair of months. (4) The link relative for the month following the bench-mark month is applied to the bench mark for the industry to derive the level of em ployment in that month. The resulting figure, in turn, becomes the base for obtaining the next month’s level. In succeeding months the same procedure is followed until new bench-mark data become available. At that time, an adjustment is made, if necessary, to eliminate possible dif ferences between series based on the old and new bench marks. An illustration of the estimation procedure used in those industries for which both all-em ployee and production-worker employment infor mation is published follows: The latest produc tion-worker employment bench mark for a given industry was 50,000 in September. According to the reporting sample, 60 establishments in that industry employed 25,000 workers in September and 26,000 in October, a 4-percent increase. To derive the October figure of 52,000, the change for identical establishments reported in the September-October sample is applied to the bench mark: 50' 000X 2 W 0 (° r 1-04)=52>000 To determine the estimated all-employee level of 65,000 for October, that month’s sample ratio of 35 production workers to total employment is used ( 528d5" ^°r mu^ P ^ ed by 1.25)=65,000^. Sim ilarly, the employment level for the next month would be prepared by the use of reports for October-November and the previously deter mined October employment. Thus, the industry employment trends reflect the fluctuations shown by establishments reporting to the Bureau of Labor Statistics; the level of employment is determined by the bench mark. Sources oj Bench-Mark Data. Complete counts or bench marks are required periodically in order to adjust for errors resulting from the use of a month-to-month sample, and further, to provide base levels of employment for subsequent pro jections during inter-bench-mark periods. For example, a source of possible error is the difficulty of introducing new firms into the sample at the time they begin operations which results in a downward bias. Since 1939, the primary source for bench-mark materials has been the employment covered under the social-security program: (1) employment in firms liable to contributions to the State unem ployment compensation funds; (2) data from the Bureau of Old-Age and Survivors Insurance on employment in firms exempt from State unemploy ment insurance laws because of their small size. Information from these two sources covers a substantial number of the persons engaged in nonfarm employment in each State. Special bench marks are used for industries not covered by the social-security program. Services for Fed eral, State, and local governments, and services performed for religious, educational, and charitable organizations are among the more important ex clusions from the social-security program. State and local government bench marks are based on data compiled by the Bureau of the Census, and most of the data on Federal Government employ ment are made available by the United States Civil Service Commission. The Interstate Com merce Commission is the source for railroads. Production-worker employment information is not available from the social-security reports and, therefore, a production-worker bench mark must be derived for each industry for which such employment information is made available. This is done by applying to the all-employee bench 36 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES in detailed industry reports for the current an< the immediately preceding month are subject t< further revision. These revisions represent th< first type mentioned above. Revisions of Series. Owing to the sampling and (2) Appropriate revisions, based on new bencl estimating procedures used, varying degrees of marks, are introduced into the employment ser error develop in the current monthly employment ies annually or less frequently, as required. Thes< levels. These errors are corrected by subsequent adjustments are necessary primarily becaus< revision of previously published data. In this reports are not immediately available from nev connection, two types of revisions periodically firms; frequently such reports are not includec incorporated into the employment series should in the particular industry sample until after th( be distinguished: (1) revisions of current monthly establishments have been in operation for som< data following the preliminary release of the time. Experience with the employment-statistics information; (2) annual revisions on a more program has shown that, without bench-mart extensive basis after more current bench-mark adjustments, the employment data tend to develop reports become available. a marked understatement which becomes largei (1) The Bureau of Labor Statistics makes avail from year to year. For manufacturing and man} able, about a month after the month of reference, nonmanufacturing industries, the data used in preliminary information relating to wage and preparing the new bench marks are available salary employment in nonfarm establishments, about 7 to 10 months after the last month of the including major industry divisions and a number bench-mark period. In general, the bench-mark of manufacturing and nonmanufacturing indus period relates to the first quarter of the year. The try groups. Such information is based upon monthly employment levels which had been made preliminary tabulations of data from those available previously for that quarter are com sample establishments that have been received pared with the new bench-mark data. The need in time for the preliminary release. Subsequently, for adjustment of the published employment revised data are made available on the basis of information is determined from this comparison. all reports in the sample. Thus, data presented mark the ratio of production-worker employment to total employment in the comparable period, as determined from the Bureau’s industry sample. VIII. Calculating Hours and Earnings of Workers in Industry27 The Bureau of Labor Statistics makes available monthly series relating to wages and hours of work in connection with its general program of industrial employment statistics. Currently, such data are issued in comprehensive and detailed form for the manufacturing industries on a national basis, and for a selected group of non manufacturing activities. The program is de signed, however, to produce similar data for all important segments of the nonfarm economy. Some progress has been made, in cooperation with State agencies, in the development of such series for States and important industrial areas. Each month, the Bureau publishes average weekly hours, average hourly earnings, and aver age weekly earnings relating to production or nonsupervisory workers in total manufacturing, the durable and nondurable goods industry sub divisions, 21 major groups of manufacturing industries, and 178 separate manufacturing indus tries. Within nonmanufacturing, 40 series are published on hours and earnings. Similar aver ages for class I railroads are also presented, based on data supplied by the Interstate Commerce Commission. Hours and earnings in manufacturing industries are also shown monthly for 26 States and 42 areas at the present time. Expansion of this program, currently under way, is expected to result in the regular publication of similar data for all States and for about 100 metropolitan areas. Survey Sources and Methods An integral part of the Bureau’s current em ployment statistics program, the hours and earnings data are based upon monthly mail re ports provided by approximately 100,000 co operating establishments. (See description in preceding article.) 27 Prepared by Sydney S. Netreba of tbe Bureau’s Division of Employment Statistics. The number of cooperating establishments reporting the above pay-roll and man-hour data is shown in table 6. T able 6.— Approximate coverage of hours and earnings sample Division or industry Mining.......... ...... ................................. Contract construction.............. .......... Manufacturing........................... ......... Transportation and public utilities: Class I railroads (IC C )................. Local railways and bus lines 1....... Telephone-..................................... Telegraph....................................... Gas and electric utilities............... Trade................... ................................ Finance................................................ Service: H otels........................................... Laundries and cleaning and dye ing plants.................................... Employees Number of establish ments Number in Percent of sample total 2,700 15,000 35,200 460, 000 450,000 8,845,000 47 23 62 400 550 4,000 3,900 46, 300 6,000 1,215,000 135,000 579,000 33,000 360, 000 1,379,000 281,000 100 86 90 60 70 15 16 1,200 115,000 25 1,700 86,000 17 1 Refers to privately operated companies. The hours and earnings data include reports for an additional 30,000 employees in Government operated establishments. In general, the establishment reports contain the following information necessary for the computa tion of the hours and earnings averages: (1) The number of all full- and part-time pro duction workers or nonsupervisory employees who worked during or received pay for any part of the period reported. For manufacturing, mining, power laundries, and cleaning and dyeing indus tries, the data cover production and related workers only. Production and related workers in clude working foremen and all nonsupervisory workers (including lead men and trainees) engaged in fabricating, processing, assembling, inspection, receiving, storing, handling, packing, warehousing, shipping, maintenance, repair, janitorial, watch man services, products development, auxiliary production for plant’s own use (e. g., power plant), and record-keeping and other services closely associated with the production operations. Data for the telephone industry through May 1949 reflect mainly the hours and earnings of employees subject to the Fair Labor Standards Act. Begin ning with June 1949, the telephone averages relate to the hours and earnings of nonsupervisory em37 38 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES ployees. For the remaining industries, unless otherwise noted, the data refer to all nonsupervisory employees and working supervisors. (2) Total gross pay rolls for such workers before deductions for old-age and unemployment insur ance, withholding tax, bonds, union dues, and special clothing allowances. The pay-roll figures also include: Pay for sick leave, holidays, and vacations taken. They exclude: Cash payments for vacations not taken, retroactive pay not earned during period reported, value of payments in kind, contributions to welfare funds and in surance or pension plans, and bonuses, unless earned and paid regularly each pay period. (3) Total man-hours actually worked or paid for by full- and part-time production or nonsupervisory workers including hours paid for holidays, sick leave, and vacations taken; if employees elect to work during a vacation period, only actual hours worked by such employees are included. The period reported generally represents a pay period of one week ending nearest the fifteenth of the month. For those establishments which use a 2-week or longer pay period, the schedules are edited to reduce the pay-roll and man-hour aggre gates to their proper equivalents for a single week. Cooperating State Agencies. The current employ ment statistics program, and hence the hours and earnings information, is an integrated FederalState project, under which the preparation of data on National, State, and area levels is facili tated. In accordance with authority granted in a Congressional Act of 1930 (Public Laws, Chapter 873, Vol. 46, Part I) and in order to minimize the reporting burdens placed upon cooperating establishments, the Bureau of Labor Statistics has entered into agreements with various State agencies, whereby basic uniformity in the collection of Bureau-State employment and hours and earnings statistics is obtained by the use of Bureau-designed schedules for monthly reporting. The “ contract” States edit these reports according to standards defined in editing instructions issued by the Bureau, and make these data available for inclusion in the National esti mates. The cooperating agencies are responsible for the preparation of State and area statistics. To maintain comparable data from State to State, statistical standards are set forth in a BLS-State procedures manual and in related instructional memoranda. A list of the cooperating State agencies is presented in the monthly mimeo graphed Hours and Earnings Industry Report and in a footnote to table A-10 of the Monthly Labor Review beginning with the October 1949 issue. Calculation Procedures The industry classification structure currently used in the hours and earnings series under the general employment statistics program follows the standards established in the documents here listed: (1) For manufacturing industries— Standard In dustrial Classification Manual, volume I, Manu facturing Industries, Bureau of the Budget, N ov ember 1945; (2) for nonmanufacturing industries— Industrial Classification Code, Federal Security Agency, Social Security Board, 1942. Average Hours and Earnings. In the hours and earnings series, it is necessary to distinguish be tween computations respecting groups of indus tries from those for individual industries. To ob tain average weekly hours for an individual in dustry— e. g. blast furnaces, steel works, and rolling mills— the sum of the man-hour totals re ported by the plants classified in the given cate gory is divided by the number of production or nonsup ervisory workers reported for the same establishments. For example: 112,000 (weeklyman-hours) Af. A , . „ 2,775 (production workers) =‘ 40-4 40'36) weekly hours Weekly hours for combinations of industries— all manufacturing, major manufacturing and non manufacturing groups— are weighted averages. The workweek average for each individual com ponent industry is weighted by employment in order to reflect the relative importance, and com bined with others in the group to derive the group average. For individual manufacturing and nonmanufac turing industries, average hourly earnings result from the division of the reported pay-roll totals by the aggregate man-hours furnished by the same establishments. Thus,* 112,000 (weekly man-hours) P'>y>,rOU)' - t l .30 Average hourly earnings for industry groupings are also weighted. For this purpose, however, CALCULATING EARNINGS OF WORKERS aggregate man-hours (employment multiplied by weekly hours) are used, because the level of pay rolls depends, in part, on both employment and hours of work. The aggregate man-hour weights required for the group averages of hourly earnings are a byproduct of the procedures utilized to obtain average weekly hours for industrial groups. For both individual components and the major industry categories, the averages of weekly earn ings are computed by multiplying the respective averages of hourly earnings and weekly hours. Hours and earnings data are based solely upon the sample aggregates. Thus, only those totals for employment, man-hours, and pay rolls as reported by the establishments in the given indus try samples are used for the computations of the averages. Both the levels and the monthly movements shown in the Bureau hours and earn ings information are, therefore, entirely dependent upon the data reported by such establishments. In the Bureau’s employment series, owing to the availability of complete counts or bench marks, adjustments of the sample data are possible but this cannot be done for the hours and earnings information. Earnings Exclusive oj Overtime Premium Payment. The pay-roll and man-hour totals on the reports from respondent employers combine both straight-time and overtime. To obtain straight-time earnings, approximations for all manufacturing and the durable and nondurable goods subdivisions have been derived from the gross average hourly earn ings provided by these totals. Such data are computed by the application of adjustment fac tors, as indicated in the Monthly Labor Review for November 1942 (pp. 1053-6) and in Serial No. R. 1496. The adjustment factors are based on a special study of the relationship between average weekly hours and the average number of weekly overtime hours in the same period. Use of these factors makes possible an approximate deduction of overtime premium payments at the rate of time and a half for hours in excess of 40 a week. The technique used in obtaining these series of average hourly earnings exclusive of over time premium payment is to apply the factors (as indicated above) separately to the gross average hourly earnings for each of the major manufac turing industry groups. The resulting adjusted average for each major industry group is then 39 weighted by the respective man-hour aggregate, as currently reported, in order to give more ade quate representation to each of the component groups in computing the averages for all manufac turing and the durable and nondurable goods in dustries. Results obtained, therefore, by direct application of the adjustment factors to published averages of gross hourly earnings for all manufac turing and the durable and nondurable goods industries differ somewhat from the figures derived by the use of the man-hour weights. Net Spendable Average Weekly Earnings. Appli cation of the term “ take-home pay” to the Bureau’s gross average weekly earnings series has never been completely accurate because of varying deductions for taxes, group insurance, occupa tional tools and supplies, and union dues. These deductions were generally not large prior to the extension of personal income taxes to the lower income brackets, and therefore average gross weekly earnings were satisfactory as a measure of the trends of weekly earnings available for spend ing purposes. Before that extension of income taxes, the only uniform Nation-wide deduction from gross weekly pay was the social-security tax, which became effective on January 1, 1937. After income taxes became a factor affecting the spendable earnings of workers, a method was de veloped for deriving so-called “ net spendable earn ings” from the gross average weekly earnings series for production workers in manufacturing industries. Net spendable average weekly earn ings are obtained by deducting from gross weekly earnings, social-security and income taxes for which the specified type of worker is liable. The amount of income tax liability depends, of course, on the number of dependents supported by the worker as well as on the level of his gross income. Net spendable earnings have, therefore, been com puted for two types of income-receivers: (1) A worker with no dependents; and (2) a worker with three dependents. Adjustment for Changes in Purchasing Power. To provide an indication of the changes in the pur chasing power of money earnings resulting from changes in the prices of consumer goods and serv ices since 1939, gross weekly earnings in both cur rent and 1939 dollars are published for selected in dustries. All manufacturing net spendable weekly 40 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES earnings are also shown in 1939 dollars. The ad justment for changes in purchasing power is made by deflating weekly earnings by the Bureau's con sumers' price index, using the year 1939 for the base period. Thus, to express gross average weekly earnings in all manufacturing industries for August 1947 in 1939 dollars the following steps are necessary: 1. Gross average weekly earnings, all manufacturing, August 1947_____ $50. 07 2. Consumers’ price index, August 1947 (1935-39= 100)__________________ 160.3 3. Rebasing of the August 1947 con sumers’ price index from 1935-39= 100 to 1939=100; 160.3 divided by the 1939 index of 99.4 equals 1.613-— the ratio between the consumers’ price index in August 1947 and year 1939— or an index for August 1947 of 161.3 (1939=100)________ 1. 613 4. Deflation of average weekly earnings in current dollars to obtain the August 1947 average in 1939 dol lars______________________________$31.04 Thus $50.07 divided by 1.613 equals $31.04. Interpretation and Limitations Average Hourly Earnings, Weekly Hours} and Weekly Earnings. The average hourly earnings series for full- and part-time production or nonsupervisory workers in manufacturing and non manufacturing industries are on a “ gross” basis, that is, they reflect not only changes in basic hourly and incentive wage rates, but also such variable factors as premium pay for overtime and late-shift work and changes in the output of workers paid on an incentive basis. Industry averages of hourly earnings are further affected by changes in the relative importance of individual companies or establishments as well as changes in the composition of the labor force and shifts in the relative importance of individual regions or localities. In addition, averages for major groups and divisions reflect changes in the relative importance of individual industries. Gross hourly earnings refer to the actual return to the worker for a stated period of time. They differ from wage rates, which represent the rates stipulated for a given unit of work or time. Owing to the exclusion of irregular bonuses, retroactive items, payments of various welfare benefits, and of earnings for those employees not covered under the production-worker or nonsupervisory-employee definitions, the average-earnings series should not be interpreted as representing total labor costs on the part of the employer. Gross average weekly earnings are also affected by changes in the length of the workweek, parttime work, stoppages for varying causes, labor turn-over, and absenteeism. The weekly earn ings, although sometimes incorrectly termed “ take-home pay,” are not the same as the amount that is actually available to workers for spending, since no deduction has been made for income and social-security taxes, group insurance, occupational supplies, and union dues. The workweek information published each month relates to average hours worked or paid for and is somewhat different from standard or scheduled hours. Thus, average weekly hours for manufacturing and nonmanufacturing industries are less than the hours of workers who are on the pay roll during the whole of the workweek because of the influence of such factors as absenteeism, labor turn-over, part-time work, and stoppages. Group averages further reflect changes in the importance of component industries. Hourly Earnings, Exclusive oj Overtime Premium Payment. The method used in approximating straight-time earnings provides only for the elimination of overtime work paid for at one and a half times the straight-time rates after 40 hours a week. Thus, no adjustment is made for other premium payment provisions, for example, holiday work, late shift work, and penalty overtime rates other than time and a half. The adjusted hourly earnings differ in only one respect from the gross average hourly earnings series; thus, except for overtime premiums the previously mentioned in fluences also apply to the estimated straight-time earnings information. Average hourly earnings, exclusive of overtime premium payment, are made available only for manufacturing as a whole and the durable and nondurable goods industry sub divisions. The set of adjustment factors, however, can be used to eliminate overtime premium pay ments from gross average hourly earnings in those manufacturing industries in which overtime for CALCULATING EARNINGS OF WORKERS individual workers typically consists of hours in excess of 40 per week paid at the rate of time and a half. As these factors yield results which are only approximate, they may not be appropriate when exact data are required. Net Spendable Weekly Earnings. Net spendable weekly earnings differ from gross weekly earnings only because of adjustment for Federal income and social-security taxes. Consequently, those factors which are reflected in the gross earnings data are also reflected in the spendable earnings series. It is necessary to note that the computations of net spendable earnings for both the factory worker with no dependents and the factory worker with three dependents are based upon the gross average weekly earnings for all production and related workers in manufacturing industries without direct regard to marital status and family com position. The gross and spendable earnings data do not reflect, therefore, actual differences in levels of earnings for workers of varying age, occupation, skill, and family composition. Thus, the primary value of the spendable series is that of measuring 41 relative changes in disposable earnings for two types of income receivers. Earnings Expressed in 1989 Dollars. The “ real earnings” data resulting from the adjustment of the selected gross and net spendable weekly earn ings averages by the Bureau’s consumers’ price index present merely a rough indication of the changes in the purchasing power of money earn ings— or earnings in current dollars—flowing from changes in living costs. The deflated series should not be considered, therefore, as representative of changes in living standards as a whole since a number of important characteristics are omitted, such as the total income of the family or spending unit, the extension and incidence of various social services and benefits, the availability of leisure time, and the duration and extent of employment and unemployment. The year 1939 is the base selected for compari son; it is, of course, possible to select other periods. Other bases may yield different results owing to differences in the levels of both earnings and prices at various times. IX. Measurement of Unit Man-Hour Requirements28 The relationship between production and man hours worked in manufacture has long been recog nized as one of the most significant indicators of economic well-being in the United States. “ Pro ductivity,” as this relationship is called, has been of growing interest owing to the general realiza tion that in the long run, a nation’s standard of living and competitive advantage in world trade depend upon continual improvement in produc tivity levels. The indexes of man-hours expended per unit (or the reciprocal— output per man-hour) published by the U. S. Department of Labor’s Bureau of Labor Statistics, measure this relation ship. During the nineteenth and twentieth centuries, especially, there has been a rapid and steady growth in production per man-hour in the United States. This has resulted in part from continued acquisition of technical knowledge and its appli cation to the jobs to be done. Of equal impor tance, continual and conscious effort has been devoted to the study of ways and means of in creasing industrial efficiency, with the expenditure of fewer man-hours. The result has been a rising material standard of living and increased leisure time to enjoy it. Productivity is also one of the major deter minants of the amount of employment that accompanies a given volume of production. In the long run, man-hour output is a major link between wage movements and price levels. An increase in output per man-hour or a decrease in man-hours per unit implies a reduction in costs; the savings may appear in the form of higher wages, higher profits, lower prices, or a combina tion of these changes. There may be little im mediate connection between wage rates and man hour output in any particular industry, but in the long run for the economy as a whole there is a connection, because consumption is limited by production. To the individual firm, continued 18 Prepared by George E. Sadler and Allan D. Searle in the Bureau’s Branch of Productivity and Technological Development. 42 improvement in output per man-hour may mean survival. . The Bureau of Labor Statistics has compiled nonrecurrent reports on productivity from time to time since 1895, providing data on productivity trends for many industries. These early studies generally included detailed analyses by process, department, or operation, and dealt with the technological changes made during the years covered. In 1940, the Bureau was given respon sibility for extending and maintaining an earlier program of the National Research Project o f the Work Projects Administration, which involved the preparation of industry-wide measures o f output per man-hour on an annual basis from readily available statistics on production, em ployment, and man-hours. In 1945, a new pro gram was begun, namely, the preparation of de tailed industry reports presenting annual data based on statistics of unit man-hour requirementscollected directly from industry. The Bureau’s current productivity program falls into two main categories— the preparation of industry-wide measures of output per man-hour from readily available secondary-source material,*4 2^ and the compilation of detailed reports on trends in man-hour requirements based on company reports for specified products which are selected in an effort to represent the production of given industries. Data in both sets of reports are pre sented in the form of indexes, with 1939 as the base year or 100. The detailed studies also pre sent product man-hour averages for specified years.30 29 Industry-wide indexes of output per man-hour are currently (December 1949), prepared for 19 manufacturing industries, 5 mining industries, 3 public utilities, and agriculture. Before World War II, a number of additional in dexes were prepared and an index to represent manufacturing in general was compiled. During the war, the absence of suitable data necessitated a re duction in the number of industries for which indexes could be computed. Industry coverage is currently being expanded, however, to include a number of basic industries not represented since 1939. 30 Late in 1949, detailed statistics were available for 16 industries, and series for additional segments were being added. These reports show data in terms of factory direct and indirect labor. The initial reports generally cover a number of years, beginning with 1939; data are kept current through annual supplements. Indexes are published for the industry as a whole; for specified products, processes, or departments; and for groups of plants classified according to size, method of production, price line, geographic area, degree of product-diversification, or extent of improvements in factory equipment and methods. These reports contain analyses of the various conditions which have favorably or unfavorably influenced unit man-hours over the period. UNIT MAN-HOUR REQUIREMENTS Limitations of Productivity Measures Labor is only one of the factors entering into production, and, therefore, the indexes are not in themselves complete measures of “ industrial ef ficiency/J Changes in the quantities of material consumed per unit of production or in the amount of capital consumed per unit, for example, are not measured by the indexes. The indexes do not measure “ worker efficiency” and cannot be used to determine what changes have occurred in average skill and effort of the work force per se. The physical quantity of pro duction attained by a given number of man-hours worked (or conversely, the number of man-hours required per unit of physical output) depends on a large number of separate, though interrelated, influences. The type of production methods in use, the age and condition of machinery, changes in product design, changes in work methods, and the relative degree of capacity utilization, all affect efficiency. In addition, the availability of parts and components, the efficiency of the man agerial and labor forces, and industrial relations directly affect output per man-hour. It is usually impossible to isolate and measure the effect of any one of these without a detailed and costly engi neering study. In some of the Bureau’s industry reports, an attempt is made to measure separately the effects on labor requirements of changes in capacity utilization and in volume of output. Since the indexes are intended to represent the experience of entire industries, it cannot be as sumed either that the trend of man-hours per unit for any one plant will conform with the industry wide trend or that the individual industry or product indexes necessarily portray the trend of output per man-hour or of unit man-hour require ments in the economy as a whole. The Bureau has refrained from averaging its series on manu facturing to obtain an “ all manufacturing” series. An all manufacturing series must await the addi tion of a considerable number of industries to the Bureau’s program of productivity measurement. In both the industry-wide indexes of output per man-hour developed from secondary source sta tistics and in the detailed indexes of unit labor requirements derived by the field collection tech nique, output is related primarily to the man-hours of wage earners or “ production workers” as de fined by the Bureau’s Division of Employment 43 Statistics.31 Other groups of employees are ex cluded. Furthermore, no distinction is made as to level of skill, either in regard to workers or hours worked. Industry Indexes: Secondary Source Data Methods and Sources. Indexes of output per worker are obtained by dividing an index of pro duction covering the output of an industry by an index of production worker employment; indexes of output per man-hour are obtained by dividing the production index by an index of production worker man-hours.32 These indexes are based on statistics for production and employ ment collected by many different agencies and the scope of industrial activity covered by each index is dictated by the nature of the data available. In general, the indexes refer to the activities of industries as defined either in the Census of Manufactures, 1939, or as reclassified in 1947.33 Many different agencies compile the production data used in the construction of the production indexes. The Bureau of the Census of the U. S. Department of Commerce is the most important source of quantity statistics for the manufacturing industries. The Bureau of Mines compiles most of the production data used in the series for min ing, as well as those for cement, coke, and nonferrous metals. Other important sources include the U. S. Department of Agriculture for a number of food processing industries, the Fish and Wild life Service of the U. S. Department of Interior, the Interstate Commerce Commission, U. S. Tariff Commission, Federal Power Commission, Federal Communication Commission, and the Bureau of Internal Revenue. Trade associations which are important sources of data include the Tanners Council, Textile Economics Bureau, National Association of Hosiery Manufacturers, National Canners Association, Millers National Federation, National Fertilizer Association, and the American Iron and Steel Institute. The production-worker employment series is 31 These include working foremen and all nonsupervisory workers associated with production operations, e. g., fabricating, processing, inspection, assem bling, repair, janitorial and watchman services, product development. 32 Beginning with the 1948 extensions, indexes for the manufacturing series will be in terms of man-hours per unit, the reciprocal of output per man-hour. 33 The indexes are being adjusted to levels indicated by the Census of Manu factures for 1947. Changes in industry definition resulting from the adoption of the Standard Industrial Classification of the Bureau of the Budget are to be taken into account. 44 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES based on the published indexes (or on unpublished tabulations) of the Bureau of Labor Statistics in most instances. The indexes of man-hours are constructed from the employment series and Bureau data on average weekly hours.34 Mere availability of production data and of employment and man-hours statistics does not assure the construction of an accurate measure of output per man-hour. The Bureau of Labor Statistics carefully evaluates these data to assure the comparability between the production and labor figures (which are collected by different agencies) and to correct for any accumulated biases. In many cases, it is neither necessary nor pos sible to include quantity statistics for all of the items produced in an industry in order to repre sent production trends in that industry. The Bureau’s indexes of production are constructed from data on physical output of items constituting a high percentage of the total value of the products of the industry. Coverage varies between 60 and 100 percent depending on the industry; it exceeds 80 percent for three-fourths of the currently pub lished series and those proposed for compilation in the near future. Each year, before bringing these statistics up to date, the Bureau attempts to determine what new products should be added and to ascertain whether the products included continue to represent production trends in the industry. In computing the indexes, it has not generally been possible to adjust for quality changes, but they have been taken into account occasionally by assignment of separate weights to the products affected. For example, in the footwear index, men’s military work shoes and civilian work shoes were placed in separate product classes during wartime and were weighted separately (giving a higher weight to the military shoes). In some industries, changes in the nature of the product during the war period made accurate measurement of production extremely difficult. For industries which changed the nature of output completely—for example, automobiles and, in large part, agricultural machinery—no indexes were prepared for the war period. Current plans u The man-hours figures represent total production worker man-hours paid for including vacation hours, waiting time, call-in time, and other hours paid for but not actually worked call for the eventual construction of indexes for a number of these industries for 1939 and 1947. Whenever problems of measurement are known to exist, changes in the nature of output are dis cussed with qualified persons in the industry affected. If possible, product classes are broken down into relatively homogeneous categories, and an appropriate weight is assigned to each. Even when a rough production indicator exists, the Bureau does not prepare an index of output per man-hour unless some assurance is given that the production statistics adequately reflect the changes in the production composite. For instance, data are available on the number of board feet of lum ber produced each year. However, an index of production and of output per man-hour based simply on the number of board feet of lumber sawed may be influenced by a shift in manu facture from hardwood to softwood or from small dimensional lumber to large timbers, or vice versa. More man-hours per board foot are re quired for smaller-dimension lumber than for large timbers, and a rough production indicator in terms of board feet alone is insufficient for computation of a productivity index. Pending development of more suitable statistics than those currently available, the Bureau has discontinued the publication of indexes for this industry. The same action has been taken for other industries that are characterized by such production shifts. Production and labor data also must be compar able and must assure a reliable indicator of general productivity trends. Labor and production data may not be strictly comparable for two reasons: (1) The production data may not include the entire output of the industry; and (2) the pro duction figures include some quantities made as secondary products by establishments classified in other industries.35 Many of the industries for which indexes are presented are well defined and comparability presents no problem—bituminous-coal mining and beet sugar, for example. In others, quantity statistics for some items are unavailable (minor products or items normally produced by estab lishments in other industries) and a part of their w Labor data, reported by establishment, include total production-worker employment and man-hours of establishments in which the major activity falls within the industry definition. The production statistics, on the other hand, are on a commodity basis and usually include the entire output of the items reported. 45 UNIT MAN-HOUR REQUIREMENTS reported output is made in industries other than those for which the particular indexes are pre pared. The resulting output per man-hour index implies the assumption, then, that (1) the propor tion of secondary products has remained un changed, and (2) the proportion of the items in cluded in the index but made outside the industry has remained unchanged, or that (3) any changes made in (1) and (2) offset each other. Such as sumptions are never accepted by the Bureau un checked; an attempt is always made to obtain either statistics or at least descriptive information to check the assumptions. Both the Census of Manufactures for 1939 and 1947 provide valuable data for evaluating the assumptions and for mak ing necessary adjustments in the manufacturing series. When the problem of comparability can not be solved within reasonable limits, no index is published. man-hour requirements; the answer to the second question cannot. The indexes of unit man-hour requirements for individual industries are based on the formula: i• Lt in which & represents the production of Calculation Procedures. In order to establish the relation between production and man-hours ex pended, it might seem to be simply a matter of choice whether a production index is divided by a man-hours index to obtain an index of output per man-hour, or whether the man-hours series is divided by the production measure to obtain man hours per unit. Either result has practical value as a measure, one being the reciprocal of the other. In a strict sense, however, it is necessary to es tablish an index of unit man-hour requirements. For example, in 1939 the cigar and cigarette in dustries produced 5.3 billion cigars and 181 billion cigarettes with the expenditure of 133 million man-hours of labor. Suppose the question is asked: How many man-hours would have been required in 1945 to produce the same number of cigars and cigarettes? Computation shows that the total is in the neighborhood of 104 million man-hours. Dividing 104 by 133 results in an index of unit man-hour requirements of 78 in 1945 (1939 = 100). The corresponding index of output per man-hour is 128. Suppose the ques tion is rephrased: How many cigars and cigarettes could have been produced in 1945 with an ex penditure of 133 million man-hours (the number actually worked in 1939)? No single figure an swers this question, for a combination of varying numbers of cigars and cigarettes would fulfill the requirements. The answer to the first question can be expressed as an index number of unit L = ^ g A ==^g ^< . Sr0g« each item in the given year and r0 and rj represent the number of man-hours required for these items in the base year and given year, respectively. The numerator shows the number of man-hours actually worked in year “ i” for all the items con sidered. The denominator shows the number of man-hours which would have been required in the base year for the identical production. The index of unit man-hour requirements can also be expressed as the quotient of an index of man-hours and an index of production, and it is in this manner that the series are actually con structed. Thus: * ’ 2 r0q0 The first term on the right-hand side of the equa tion is an index of total man-hotirs. The second term is a production index in which the quantities of the separate products are weighted by man hours per unit in the base year. For many industries, no information is avail able on man-hours per unit for the separate products for any period of time. In such cases, weights other than man-hours per unit are em ployed in constructing the output series, and the Bureau attempts to test whether the alter native weights are proportional to unit labor requirements. In the absence of data on unit man-hour requirements in the base year, labor costs per unit provide fairly good approximations. Value added per unit—i. e., the value per unit less the cost of materials per unit— usually has a rather high labor content and hence has a relation ship to man-hours per unit. Unit value weights are used when data on labor cost or value added are not available.36 In practice, similar results are generally obtained with any of these weighting systems. Indexes of unit man-hour requirements prepared by use of a production index constructed with base-year unit-value weights show the change in man hours per unit per dollar of total value, in terms of base-year value per unit for each item. They reflect the effect of any shifts from products with high man-hours per dollar of total value, or vice versa, including the effect of shifts from products with high materials costs per man-hour to those with low material costs per man-hour or vice versa. 46 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES The manufacturing industries included in the Bureau’s existing program and those being initiated in late 1949 were being adjusted to bench marks recently made available by the Census of Manufactures, 1947. The method of adjustment follows: (1) A production index based on Census data is prepared for the year 1947 (1939 = 100) in the same manner that the previously published unad justed production index was prepared. (2) An employment index for 1947 is prepared from Census production worker totals in 1939 and 1947. (3) The production and employment series pre viously published are adjusted by spreading the percentage of the 1947 adjusted to unadjusted series over the years 1939-47. (4) The Bureau’s series of average weekly hours is used in conjunction with the adjusted employ ment index to obtain the index of man-hours. Such adjustments do much to eliminate any problems of comparability which might arise as a result of deriving the production and employ ment data from separate sources. The Bureau plans to add a number of manu facturing industries to its program during 1950. Special attention is to be given to increasing the representation of industry groups for which cover age is relatively low. Industry Reports: Field-Collected Data Survey Methods and Sources. Field-collected data are necessary to prepare comprehensive industry reports for industries of primary importance in the manufacturing economy, and for those for which no unit man-hour measures can be prepared from published data. Extensive developmental re search is necessary for each proposed study to arrive at the appropriate basis for man-hour reporting (by product, process, or department). When possible, the industry’s total output is covered. If this is too costly, a sample of products is selected to provide a typical cross section of the industry’s output; all of the hundreds of products in an industry need not be covered to provide satisfactory measures of unit man-hour trends. Physical specifications are then developed to limit product man-hour reporting to generally com parable models, which experts in the industry judge to be largest in volume and most represen tative of the entire output.37 In addition to the detailed product specifications applied, reported items are further identified in exact terms. Bureau representatives who arrange for man-hour reporting obtain copies of company catalogs or other literature describing in detail the physical and operating specifications of reported product models. Man-hour and production data are obtained annually for identical models as long as they are produced in volume. WTien a given product model is discontinued, the Bureau’s representa tives must arrange for substitution of another product made in volume. Substitutions are of two types: (1) Substitution of another model which is essentially similar in characteristics, and is adequately described by the same specification; and (2) substitution of an article serving the same purpose, but not of the same quality or design, and described by a new specification. In the first type, any difference in man-hours per unit (almost invariably negligible) between the old and the substituted model is shown as a unit man-hour change in the index calculation. For example, if one model of a roller bearing, identified by a given catalog number, is no longer available, and another model, differing only in minute detail, is substituted, any difference in man-hours is allowed to affect the level of the index. In the second type of substitution, the level of the index is not affected, for the new model is introduced by linking. Any increase or decrease in man-hours for the new product in the year of substitution is therefore not reflected in the annual percent of change on which the index is based. An example of this type of substitution is the replacement of a turret lathe with manual con trols, adapted only for high-speed steel cutting tools, by a turret lathe of similar type and capacity with automatic controls, and adapted for the use of carbide-tipped cutters. Substitute specifica tions are always made to adhere as closely as 37 Typical specifications for three items follow: (a) Refrigerator, Electric, Domestic— Exterior finish, lacquer (not porce lain); interior, porcelain; storage capacity, 6-7.2 cu. ft.; shelf area, 11-15 sq. ft.; 3-5 ice trays, 6-12.6 lbs. of ice; compressor, hermetically sealed. (b) Ball Bearing—Single-row ball bearing, industrial type, SAE #208. (c) Turret Lathe—For bar stock, arranged for motor drive; 2W' round stock capacity; 34"-42" turning length; 12" scroll stock capacity; 9-12 spindle speeds, 6-18 cross feeds, ranging .002—070 to .003-103 ; 6-18 longitudinal feeds ranging .002-.103 to ,004-49; 5-10 h. p.; 1070-1800 rpm motor recommended; 4,500-6,750 lbs. net weight. UNIT MAN-HOUR REQUIREMENTS possible to those supplanted, i. e., with respect to end-use applications of the products, materials, design, and unit man-hour standards. In some cases, changes take place in the scope of operations performed in a given establishment. Either (1) the man-hour data are adjusted to correct for these changes, or (2) the company statistics are not used for the affected years unless information is available as a basis for appropriate adjustments. Keports of production and man-hours are ob tained directly from manufacturing firms. An establishment list is developed with the advice of experts in the industry. In a few industries, if the producer universe is large and a few firms do not dominate output, scientific sampling pro cedures are employed. In most industries, how ever, the Bureau attempts to establish reporting from all of the major firms in the industry, and from a balanced sample of the medium-sized and small firms. Consideration is given to quality of product, production methods, and geographic location. Survey Procedures. Field representatives of the Bureau visit each of the firms in the sample to arrange for reporting of company unit man-hour requirements.38 At each plant, the representative explains the program and discusses the company’s cost-accounting system. He agrees with the company accounting officials upon the method to be used in deriving unit man-hour data from the company records. At the plant, the agent lists detailed information on the company’s classifica tion of labor accounts, scope of operations, parts purchased, and ratio of the various indirect labor accounts to total direct man-hours. He also ob tains information on the extent and type of hours paid for but not worked;39 and the basis for re porting of capacity data. The Bureau schedule (here reproduced) em bodying the company’s statements of production and man-hour figures are forwarded to the Bureau of Labor Statistics in Washington, where the data 38 After reporting is firmly established in each industry, annual unit man hour data for succeeding years in most industries are reported by the partic ipating firms on schedules mailed to them by the Bureau. This eliminates the necessity for an additional visit by Bureau field representatives. 39 The basic instructions are to exclude such time from the man-hour reports. If the company cannot exclude such time, the agent obtains esti mates as to the relative amount of time included on the report. 47 are processed.40 In turn, a copy of the schedule is forwarded to the agent, together with any comments or questions from the Washington analyst regarding the reported data. Bureau field agents then revisit the reporting firms and discuss the data with the accounting, production, engineering, and design experts of the companies. In this way, they obtain information on all of the factors which have affected each company’s man-hour requirements— the effect of changes in factory equipment and work methods, changes in the labor force, the relative availability of materials and parts, production and employ ment trends, the utilization of plant capacity, and similar factors. Through discussions with company design engineers, the Bureau represent atives analyze in specific detail the relationship between the reported man-hours for each firm and the specifications of the products. When possible, relative standard performance records are obtained in instances where design has changed. Production and man-hour data from schedules prepared by the reporting companies are utilized in preparing summary statistics for the industry. The information embodied in the agent’s prelimi nary and final reports provides a basis for classifying company data and for the interpretative and analytical text which accompanies the statis tical analysis for each industry. Calculation Procedures. Preparation of the unit man-hour indexes involves the following general calculations: (1) Computation of product unit man-hour re quirements from company schedules, for each year reported.41 (2) Multiplication of unit man-hours for each year by a base-period weight reflecting each com pany’s relative output; this yields annual unit man-hour aggregates. (3) Summation of unit man-hour aggregates, all company reports for each product available for consecutive years of comparison. (4) Computation of year-to-year percent of change in average man-hours per unit, each product. 40 In a limited number of industries (for example, men's dress shirts) and n a few plants in almost all industries, Bureau representatives compile the ^an-hour data from company production and cost, pay-roll, or time records. 41 If reporting is on a process or department basis, process or departments man-hours are substituted for produ ct man-hours in computations described 48 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES (5) Computation of annual indexes by applying percent of change to previous year’s index (starting with base year as 100). (6) Computation of industry indexes (all prod ucts combined) by deriving weighted average of the individual product indexes. Product weights reflect the relative importance of each product in the industry’s output pattern. (7) Indexes of trends in unit man-hours for groups of firms classified according to size, produc tion method, or other characteristics (without regard to product) are computed by averaging the indexes of trends derived from individual company product reports for all firms within each category. Two basic techniques are used in computing th product indexes— either the “ link-relative” or th “ base-year relative” method. If continuous data are available for all entries in the index, the two methods yield identical results. In some in stances, however, breaks in sample may cause the results of the two methods to differ markedly. Characteristics of the particular product sample determine the method to be used in the calculation of unit man-hour indexes for individual products. The results are checked against those for the years of references in which data are available for all reporting companies (the strongest years). The link-relative method is most generally used. For each comparison, data are used which cover all firms reporting for the previous and current year. The basic formula for this method follows: in sq* rn . 2q» T n - l ln_1 where the qx’s are the production weights assigned to each company’s product report,42 the r’s are the reported unit man-hour requirements in the year of reference and in the preceding year, and in_j is the product index for the preceding year. The formula for the base-year relative method, used less frequently, is essentially that of Laspeyres. It is expressed as . = 2qxrn . Sq*r° where the qx’s are again the production weights for each firm’s product report and the r’s are unit man-hour requirements in the year of reference and in the base year. Data for a given year, of course, are totaled only for those firms which are also covered in the base year. In computing the indexes for groups of products or for all products combined, the weight assigned to each product is an expression of its relative importance in the period for which most recent production statistics are available. The weighting period may be 1 year, or an average of 2 or more years. In most industries, it relates to postwar years (e. g., 1945-47; 1946-48; or 1947, the Census year). In each instance, the weight assigned is that which yields the most accurate approximation of total factory man-hours expended during the weighting period in the production of all models of each product. In some industries, actual man hours weights are used. In others, the best available weights are value added by manufacture or value of product. The formula covering the procedure for com puting the all-products (industry) index is similar to that shown for the base-year relative method of product index computation. It is -r _ _ S W q iq n Sw0 where w0 is equivalent to Sq^o, the man-hours ex pended by the industry in the manufacture of a particular product, or the nearest approximation thereto; and in’s are the product indexes in the current year. Annual average man-hour requirements for each product are computed by adding all units pro duced and by combining all man-hours expended for such production. The sum of the man-hours is then divided by the sum of the production. The formula for this computation is Sqxrx 2qx where qx is the number of units produced by each company in the given year, and rx is the unit man hour requirement.43 In all reports in this series, the Bureau provides statistical analyses other than the indexes of unit 43 Note that if the q’s for all firms are assigned a weight of “ 1,” the result is 42 In most instances, they consist of average annual production for 6 to 10 the so-called unweighted average. Such averages have been published in some reports (for example, Men’s Dress Shirts), when series were presented years. Average production is preferable to the volume in any 1 year, be for firms classified according to plant size, price line, production method, etc. cause any one firm’s output may be either unusually high or unusually low In such cases, an average not affected by production volume was desired. in any 1 year selected as a base. 49 UNIT MAN-HOUR REQUIREMENTS man-hour trends and the unit man-hour averages. Variations are shown in the man-hour trends from year to year for the individual reporting firms. The average ratio of factory indirect (overhead) labor to total factory labor is also computed for each industry, on the basis of the ratios reported by the individual respondents. Whenever possible, the Bureau analyzes the quantitative relationship between man-hours per unit and factors such as plant size, production volume, price line, quality of product, or utiliza tion of plant capacity. These comparisons are in some instances developed as rank correlations. However, scattergrams are usually prepared, and curves, least squares regression lines, or parabola are computed and fitted to the scattergrams to indicate the relationship. B. L. S. 1792-4B In Ref. No. 304 Budget Bureau No. 44-R580. Approval expires 1-31-50 UN ITED STATES D E PA R TM E N T OF LABOR B u r e a u o f L a b o r S t a t is t ic s CODE Ind. Co. Est. SUBMIT IN DUPLICATE WASHINGTON 25, D. C. Prod Strictly Confidential REPO R T OF UNIT MAN-HOUR REQUIREM ENTS 1. Name of C om pany_______________________________________________________________________________________________ A d d ress__________________________________________________________________________________________________________ Plant N a m e______________________________________________________________________________________________________ Plant Address____________________________________________________________________________________________________ 2. Specifications for product to be rep orted :_________________________________________________________________________ (Items 3 to 7 to be completed by Company. See instructions.) 3. List firm catalog model numbers or describe product item(s) which meet the above specifications and are covered by this re p o rt:_______________________________________________________________________________________________________ 1939 1940 1941 1945 1946 1947 1948 No. of units manufactured during calendar year Direct man-hours* Indirect man-hours* ♦For 5 above, indicate basis of calculation below: (Check one) (Check one) Man-hours per unit........................ □ Man-hours are shown for full year's production.......................................... □ Aggregate m an-hours.................... □ Man-hours are shown for representative production p e r io d ......................□ (a) If man-hours are shown for representative production runs or a portion of the production of each year, indicate the number of units manufactured during the production periods selected: 1939 ______ 1940 ______ 1941______ 1945 ______ 1946 ______ 1947 ______ 1948 ______ 6. Give the number of units constituting practical plant capacity for either: (a) models reported above or (b) all related models7 1939 1940 1941 1945 1946 1947 1948 Units constituting capacity No. of units of related models manufactured 7. Name of O fficial_________________________________________________________________________________________________ T it l e ___________________________________________________________________________ D a t e ____________________________ The data submitted on this schedule will be seen only by sworn employees of the Bureau of Labor Statistics, and will not be available to anyone else with out your written permission. The data will not be. released in any form which permits identification with any specific company. When this report is completed, please submit it, together with catalog literature identifying models reported, to Commissioner of Labor Statistics, Washing ton 25, D. C. (BLS 49-2378 X. Estimating Expenditures for New Construction44 The estimates of expenditures for new construc tion prepared jointly by the Bureau of Labor Statistics and the Office of Domestic Commerce are widely used by private business analysts and government economists as measures of construc tion activity. These estimates represent the monetary value of the construction work per formed within the continental United States dur ing the stated periods of time. This monetary value is equivalent to the cost of the materials put in place or otherwise consumed, the wages of workers who placed the materials, and appropriate charges to the work for overhead and profit. Annual estimates are available beginning with 1915, and monthly figures from January 1939. This series is an extension of yearly estimates developed in the late 1930's in the Bureau of Foreign and Domestic Commerce and presented in Construction Activity in the United States, 1915-37 (out of print). For some years after the publication of the original series, several agencies prepared a number of independent, and often conflicting, estimates projecting it. In 1945, the Bureau of the Budget, through its Divi sion of Statistical Standards, assigned the prepara tion of an official Government estimate to the Bureau of Labor Statistics and the Bureau of Foreign and Domestic Commerce jointly, and delineated the responsibilities of each agency in the preparation of the figures. Primary responsibility for estimating private residential construction and all public construc tion was given to the Bureau of Labor Statistics, and for estimating private nonresidential build ing, farm construction, and privately owned publicutilities construction to the Bureau of Foreign and Domestic Commerce. Individual responsi bilities are for the work directly involved in de veloping sources and processing the data. The two agencies take joint responsibility for the over all validity of the estimates, and the work of either agency in preparing the series is at all times sub ject to the review of the other. In these estimates, “ new construction” in cludes the production of all fixed works and struc tures, whether by contract or force-account.45 Major additions and alterations are included, except for residential building,46 but maintenance and minor repair work and, insofar as possible, work relief projects are excluded.47 The estimates cover buildings; other structures, such as dams, levees, and bridges; and nonstructural works such as airfields, highways, canals, and navigation channels. They include the installed value of equipment generally considered an integral part of a structure and commonly included in the contract price, such as plumbing and heating equipment and elevators. They exclude sepa rable equipment, such as production machinery, power-generating equipment, and furnishings. Excluded also are the value of raw land and of such preconstruction costs as architectural and engineering fees and some types of land improve ments.48 Several types of activity which have some of the characteristics of construction are excluded be cause they are primarily industrial or agricultural operations. Chief of these are (1) oil and natural gas well drilling; (2) mining operations (except for the construction of mine buildings above ground); (3) shipbuilding; and (4) farm work which is an integral part of farm operations, such as terracing and individual irrigation ditches. A few types of work properly classifiable as construction under the general definition are also omitted. These may be grouped into two classes, according to the reason for their exclusion: (1) those for which sufficient information has not yet been developed, i. e., tourist courts and cabins, dormitories (except dormitories financed with Federal funds), major additions to and alterations of private residential buildings, privately owned water and sewer systems, and some public and 4« Force-account work is done, not through a contractor, but directly by a business or government agency using a separate work force to perform non maintenance construction on the agency’s own properties. 46It was planned to include major additions and alterations on residential building in the expenditures estimates upon revision of the series in the spring of 1950. 47 The Office of Domestic Commerce prepares annual estimates of total construction activity by combining the new construction activity estimates with separate estimates of (1) work relief and (2) maintenance and repair. This series is available by years from 1915. The same office also maintains a monthly series, beginning with January 1939, showing seasonally adjusted new construction activity, and a series beginning with 1939 of new construc N ote : Information on subsequent changes in this series is available on request to the BLS. tion activity by States. (See also footnote 49.) 48It was planned to include architectural and engineering fees and all land 44 By Roland V. Murray in the Bureau’s Division of Construction Statistics improvement costs in the expenditures estimates upon revision of the series and by Bruce M. Fowler in the Construction Division of the Office of in the spring of 1950. Domestic Commerce, U. S. Department of Commerce. 50 51 EXPENDITURES— NEW CONSTRUCTION privately owned toll roads and bridges; and (2) those activities which for security reasons have not been included in the estimates, chiefly the classified projects of the Department of De fense and all construction by the Atomic Energy Commission. Data for types of work in the first group are to be introduced into the series as rapidly as sources of expenditures data are de veloped, but these must provide both current and historical information in order to insure year-toyear comparability of the series. Data for the second group are to be introduced only as projects are declassified and information is made available. Outlays for construction by the Atomic Energy Commission were scheduled for inclusion in the estimates by the spring of 1950. In addition, two items, known to be relatively unimportant, (1) public park and playground con struction and (2) non-Federal conservation and development work, are not represented in the figures before 1946. They have been included under “ All other public” beginning with January 1946. The tabulation shown lists the types of new construction for which expenditures estimates are regularly published. Somewhat greater detail is available on an annual than on a monthly basis. Type of construction Period, for which available Annually Monthly Total new construction______________ x x Private construction_________________ Residential building (nonfarm)__ Nonresidential building (nonfarm) Industrial__________________ Commercial________________ Warehouse, office, and loft buildings________ Stores, restaurants, and garages______________ Other nonresidential building Religious_______________ Educational____________ Social and recreational. _ Hospital and institutional Remaining types_______ Hotel______________ Miscellaneous______ Farm construction_______________ Residential_________________ Nonresidential______________ Public utilities__________________ Railroad____________________ x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x _______ _______ x _______ _______ x x Type of construction Period for which available Annually Monthly Private construction— Continued Public utilities— Continued Telephone and telegraph____ Other public utilities____________ Local transit___________ Petroleum pipe line_____ Electric light and power. Gas____________________ Public construction______________________ Residential building_________________ Nonresidential building______________ Educational____________________ Hospital and institutional__ All other nonresidential_____ Industrial______________ Commercial____________ Public administration__ Social and recreational. _ Miscellaneous__________ Military and naval facilities_____ Highway____________________________ State_______________________ County_____________________ Municipal__________________ Federal_____________________ Sewer and water____________________ Sewage disposal_____________ Water supply_______________ Miscellaneous public service enter prises_____________________________ Conservation and development__ Bureau of Reclamation_____ Army Engineers____________ Tennessee Valley Authority. _ Other_______________________ All other public_____________________ x x x x x x x x x x x x x x x x x x x x x x ----------------------------------------x x x x x x x x x x --------------------------------------------------x x x ----------------------------------------x x --------------------x x x x x x x x ----------------------------------------x x Limitations of the Series As the following sections show, the methods by which these estimates are compiled result in meas ures in terms of dollars of a purchasing power current during the period of reference. The fig ures therefore cannot be used as indicators of the physical volume of construction placed without extensive adjustments for differences in price levels and wage rates, technological changes, and other relevant factors.49 The expenditures estimates are distinct from contract-award or building-permit data. The latter indicate the value of construction sched 49 The Office of Domestic Commerce estimates changes in the physical voP ume of construction placed by expressing current estimates in 1939 prices. This is done by deflating the estimate for each class of construction by an appropriate construction cost index. This series is available by years from 1915 and by months from January 1939 52 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES uled for early start, and are therefore useful in forecasting construction activity. The expendi tures estimates, on the other hand, indicate cur rent activity and are therefore particularly suit able for use in making comparison with related concurrent data, such as those for employment and delivery of material. The degree of error in the estimates of expen ditures cannot be measured statistically and, because of the varied basic sources and the dif ferent characteristics of the types of construction represented, the accuracy of the figures varies considerably between types. Thus, for example, the estimates for farm construction are much less reliable than those for public utilities. In gen eral, the larger the coverage of the expenditures the more reliable are the figures. For instance, annual estimates are better than monthly, and the total for any period is more nearly accurate than the data for any of the individual types of work. Relatively small month-to-month changes should be used with caution because most monthly data are based on normal construction patterns, and not on actual observed progress (see p. 53). The year-to-year changes in the estimates for total and major types of construction are correct as to direction and are substantially correct in extent. The figures by type of construction are not adapted for use in making exceedingly fine com parisons, primarily because of some unavoidable inconsistencies in classification. For example, re gardless of type of work (whether building con struction, road work, etc.), all construction by privately owned public utilities is included under “ Public utilities,” all construction for the military by the Department of Defense is represented in “ Military and naval facilities,” and all construc tion by the civil units of the Army Engineers, the Bureau of Reclamation, and other Federal con servation agencies is included in “ Conservation and development.” Sources and General Estimating Methods Ideally, construction expenditures information would be based upon monthly reports from each construction project supplying a cost accounting figure which combines the wages of the workers on the project, the cost of the materials which they placed, and appropriate charges for overhead and profit. If this were feasible, the collection cost would greatly exceed the Federal resources cur rently made available for these estimates. There fore, the estimates are made up of data from a wide variety of secondary sources as well as from primary source material originally devised to serve other purposes. Few of these sources are static. In fact, the out standing characteristic of the list of such sources is its constant change in content. This results from: the continual search for more complete, ac curate, and timely information than previously available; the undertaking and completion of pub lic works programs; and the changes in operating and reporting requirements of Federal construc tion and regulatory agencies. Three general methods are used in constructing the estimates from the various source materials. In the order of preferred methodology, they con sist of (1) summarizing physical observations of work underway, (2) summarizing financial data on additions to plant, and (3) converting data on work started to estimates of work put in place. (1) The summaries made from actual observa tions of progress on individual construction proj ects are based primarily on the operating reports of the Federal agencies supervising the construc tion of public works, such as veterans’ hospitals, flood-control dams, irrigation canals, and the like. In designing such reports, the operating agency and the Division of Statistical Standards of the Bureau of the Budget consider the Bureau of Labor Statistics’ needs for expenditures data. Many non-Federal public construction agencies make similar reports, but the systematic collec tion and use of them is not considered feasible by the Bureau. This type of report is therefore, with the exception of the New York City Housing Authority, used only with respect to Federal construction for which it is a readily available and accurate source of information. It usually reflects the observation, or in some cases the actual measurement by a Federal engineer of the status of a construction job at uniform intervals (pri marily to determine the payments due to the contractor for work accomplished during the interval). Tabulations of data for all of the individual jobs in a given program are made either by the construction agency or by the Bureau of Labor Statistics. (2) The summaries of financial data are based 53 EXPENDITURES— NEW CONSTRUCTION on the accounting records of private companies such as privately owned public utilities, and of public agencies which are required, in the public interest, to maintain financial records. The figures reflect disbursements for construction, and are adjusted roughly, whenever possible, for deferred charges, such as for large purchases of materials to be used in a later period. In a few instances, special summary tabulations must be prepared from detailed records. These are made by the trade association involved or by the two Government agencies responsible for the statistics. Generally, however, suitable basic data are in published form. For example, some are sum marized by trade associations for presentation in yearbooks or other periodical reports to the public and certain figures are summarized by Govern ment regulatory agencies for publication in their annual reports. A special case is the table on appropriations and expenditures for civil public works in the annual Budget of the United States, made available at the time of the President’s annual budget message to Congress. It consists of a tabulation of checks issued against construction appropriations. After adjustment to eliminate survey and planning proj ects, operation, maintenance, and other non construction items, these data form the basis of the estimates for construction by a few Federal agencies from which monthly data are not obtained for various reasons. (3) Conversion of work started to estimates of work put in place is the most important of the three methods of deriving construction expendi tures estimates, from the standpoint of the dollar volume of the expenditures categories for which it is used. The value of work started each month is spread over a period of time according to a predetermined pattern. This pattern is the prob able percentage of the total cost of construction T able 7.— Sources used in estimating expenditures for new construction, by ownership, type of construction, and source of public funds Ownership and type of construction Source of public funds Source of basic data P rivate Residential building (nonfarm)____________ Nonresidential building (nonfarm): all types. Farm construction: all types______________ Public utilities: Railroads____________________________ BLS building permit reports and field studies. F. W. Dodge contract award figures. Bureau of Agricultural Economics. Interstate Commerce Commission.1* Association of American Railroads. American Telephone and Telegraph Co.8 Western Union Telegraph Co. Telephone and telegraph______________ Other public utilities: Local transit ______________ ______ Petroleum pipe line___ ____________ Electric light and power___________ Qas---------- ----------------------------------- American Transit Association. Interstate Commerce Commission. Federal Power Commission. American Gas Association. P ublic Residential building. Nonresidential building: Educational_____ _____ _ Hospital and institutional. All other nonresidential. Military and naval facilities__________ Highways............................... ......... ...... Sewer and water________________ ____ Miscellaneous public service enterprises. Conservation and development_______ All other public. Federal____ Non-Federal. Public Housing Administration.8 Local housing authorities. .......do........... Federal____ ----- do--------___ do______ Non-Federal. ___ do........ . Federal____ ___ do_____ ----- do........ . Non-Federal. Federal......... ..— do.......... ___ do______ Non-Federal. ___ do.......... . ___ do______ Federal_____ ___ do______ ___ do______ ___ do______ ___ do______ ----- do.......... Non-Federal. General Services Administration.4 Office of Chief of Engineers. Public Health Service.5 Veterans Administration. General Services Administration.4 Public Health Service.5 Federal agency supervising construction.® Federal agency awarding contract.7 Budget of United States. General Services Administration.4 Office of Chief of Engineers. Bureau of Yards and Docks. Bureau of Public Roads. Do. General Services Administration.4 D o.4 Bureau of Reclamation. Office of Chief of Engineers. Tennessee Valley Authority. Budget of the United States. Civil Aeronautics Administration. Budget of the United States. General Services Administration.4 1 Publishes data on class I railroads on Form IBS. monthly. 8 Summarizes reports from member companies of the Bell System and adjusts the results to cover independent companies. * No Federal funds were being* expended at* time of writing. 4 Office of Economic Research of G3A, using as source, F. W. Dodge and Digitized FRASER otherfor contract awards data. 1Hospital Facilities Division. 4 Covers projects of the Public Buildings Administration, certain other Federal agencies, and United Nations Headquarters in New York City. r For projects such as those by National Advisory Committee for Aero nautics. 54 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES which will be performed in each month of the known or estimated duration of the job. On a given type of construction, therefore, the esti mated expenditures for any month are the sum of the estimated expenditures during that month on all projects estimated as under way, according to the length of the expenditures pattern. Obviously the use of these three different methods of estimating expenditures raises prob lems of comparability, both as to timing and content. Timing is a problem only in month-tomonth comparisons. Over longer periods, the three methods should give similar results. Com parability in coverage is a more serious problem. Therefore, all source material is carefully examined and adjustments are made in it, or in the expendi tures estimated from it, to insure that the results conform with the general concepts previously outlined. The first and third of these three general esti mating methods yield monthly results directly. The second, financial data, are usually available only on an annual basis. Therefore, the monthly figures are obtained by projecting the levels estab lished for the previous year, month by month, on the basis of the known movements of a related series. At the year’s end, an adjustment is made to the new bench mark. At the same time, all components of the estimates are reexamined for necessary adjustment, such as to include the yearend revisions of housing data. These adjust ments may include the addition of entire series, to reduce the noncovered field previously mentioned. The preceding tabulation shows, for each type of construction, the basic source of data. These basic series are adjusted in varying degrees by the agencies responsible for the construction expendi tures series. Adjustment Procedures Private Construction. The adjustment procedures for the private construction segment are as follows. For residential building, the monthly reports to the Bureau of Labor Statistics on the value of residential building authorized by local building permits are adjusted to reflect the construction cost of new permanent nonfarm dwelling units started60 in all urban places and in rural permit issuing places. Separate inflating factors are1 0 10 See also Estimating National Housing Volume, p. 13. applied to the urban and rural segments to com pensate for the understatement of cost inherent in permit valuation. These are revised periodi cally on the basis of information obtained from field surveys in which the permit valuation and the construction cost reported by builders and contractors are compared for a large sample of projects. Construction cost of units started in rural nonpermit-issuing places is based on monthly field studies. Estimated construction costs are secured from builders and contractors for a large number of dwelling units in sample counties throughout the country. From these an average construction cost for all units started in non-permit-issuing rural areas is derived. The urban and rural segments are then com bined to give a total estimated construction cost of the dwelling units started in the given period. An expenditure pattern is then applied to this construction cost figure to estimate the amount of work put in place in the months following start of construction. This pattern is derived from: (1) special studies of construction time to obtain a distribution of completions in the month of start, in the following month, and so on; and (2) studies of the progress on actual jobs to develop typical patterns for jobs of 2-months duration, 3-months duration and so on. The final expenditure pattern is an average of these patterns for different lengths of construction time weighted by the proportion of the units started which are completed in these various lengths of time. Nonresidential buildings are covered in separate estimates made for each of the following types of new private nonresidential buildings: (1) in dustrial; (2) warehouse, office, and loft buildings; (3) stores, restaurants, and garages; (4) religious; (5) educational; (6) social and recreational; (7) hospital and institutional; (8) hotels; and (9) miscellaneous. Estimates of construction activity are derived by the following six adjustments of the figures reported monthly by the F. W. Dodge Corp.51 (1) Cancellations: A project reported by Dodge may be canceled later or indefinitely postponed. 81 Dodge figures apply only to the 37 States east of the Eoeky Mountains and to the District of Columbia. They are prepared monthly by the firm’s Statistical and Research Division, as a byproduct of its daily news reporting service. Reports are obtained by a staff of individuals who interview owners, architects, engineers, contractors, financial institutions, real estate brokers, and others able to supply reliable information on the awarding of construction contracts. EXPENDITURES— NEW CONSTRUCTION In such a contingency, the Dodge reports include the correction in the month in which the cancella tion is ascertained. Thus, the contract series may be overstated in one month and understated in a subsequent month. Therefore, the Office of Domestic Commerce applies the adjustment to the figure for the month in which the original entry appeared. (2) Undercoverage in 37 Eastern States: An adjustment is made to allow for projects not included in the Dodge reports. The omissions are chiefly smaller projects and work done by a firm’s own force. The degree of adjustment has varied both over time and between the various types of construction. (3) Allowance for 11 Western States: Since the Dodge reports cover only the 37 Eastern States, they are adjusted to reflect contract awards in the entire continental United States by applying the ratio of the valuation of urban building authorized by permits in the 11 Western States to the total valuation of urban building authorized in the United States as a whole for each type of construction. This adjustment varies from month to month and from one type of construction to another. (4) Translation of contract awards to work started: The Dodge reports represent contracts awarded during the month and due to be started soon. The assumption is made that construction on contracts awarded during a given month is started during the following month. (5) Activity timing: To convert the total value of work started each month into estimates of the value of work put in place, several activity patterns have been developed for each type of building, taking into account the average size of project and the season in which work is begun. These patterns applied to the work started series yield the estimated value of work put in place each month. (6) Duplication of public utility buildings: Offices, warehouses, and other structures built by public utilities are included in the total value of construction reported by the various utilities and are also included in the Dodge reports for nonresidential buildings. To eliminate this duplica tion, estimates for buildings constructed by public utilities are subtracted from the expenditures estimates for warehouse, office, and loft buildings. For jarm construction, estimates are currently 55 made by the Bureau of Agricultural Economics annually.52 They are prepared separately for “ operators’ dwellings” and “ service buildings” which, in the expenditures estimates, are desig nated as “ residential” and “ nonresidential” farm buildings. Both new construction and mainte nance are included. Figures for new construction alone are derived on the basis of separate estimates by the Forest Service of farm consumption of lumber for new construction and for maintenance. The lumber consumption estimates are pre pared quarterly and are, therefore, used to project the estimates of farm construction expenditures beyond the latest direct estimates of the Bureau of Agricultural Economics. In the public utilities group, adjustments are made for each of five types of public utilities: (1) Railroads: An annual adjustment to a bench mark is required with respect to the ICC monthly data on class I railroads, because coverage for class I railroads is not complete. The bench mark is obtained from the annual report, A Re view of Railway Operations in 19-, of the Bureau of Railway Economics of the Association of American Railroads, which includes information on construction expenditures of all class I rail roads.53 Construction expenditures by class I railroads are adjusted upward to include outlays for all railroad construction on the basis of the relatively stable ratio of class I investment in road and structures to such investment by all railroads. (2) Local transit: The figures on construction expenditures are included in the American Transit Association’s annual publication, Transit Fact Book, and cover the operations of all transit companies in the United States. Unpublished data, also obtained from the American Transit Association, indicate the amount of expenditures made annually by municipally owned transit companies. This figure is subtracted from the total published figure to obtain the amount spent annually by privately owned companies.54* The monthly trend is based on the trend shown by other public utilities. s2 Farm construction expenditures were first published in the Department of Agriculture study, Income Parity for Agriculture, Part II, Section 5 (March 1941). Data for subsequent years and revisions of the published estimates back to 1934 have been obtained from the BAE in unpublished form. s3 Similar data compiled by the ICC are available some time subsequent to the BRE report and may be substituted. 54 Construction expenditures by municipally owned companies are included in data for the public category, “ Miscellaneous public service enterprises." 56 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES (3) Petroleum pipe lines: Expenditures from the reports filed with the Interstate Commerce Com mission are adjusted to eliminate the purchase of existing lines and to allow for companies not required to file information with it. To obtain monthly data, estimates are made on the basis of the trend shown by Dodge contract award data that have been distributed by activity patterns, and Securities and Exchange Commission quarter ly data on capital expenditures of utility companies. (4) Electric light and power: Annual reports to the Federal Power Commission by class A and B electric utilities give the total value of all new facilities completed during the current year and the value of uncompleted work in progress at the end of the year. They are transcribed and sum marized, and the results adjusted by the Office of Domestic Commerce to exclude the value of work done in earlier years and the value of property acquired through purchase, and to include data for small companies. Monthly figures are esti mated the same way as for petroleum pipe lines. (5) Gas: The annual construction expenditures figures for both manufactured and natural gas utility industries, published by the American Gas Association, are adjusted to eliminate equipment purchases. Monthly data are estimated in the same way as for the petroleum pipe line and electric light and power groups. Public Construction. Estimates of expenditures for public construction are obtained by combining separate estimates for two components: Federal construction and all other public construction (i.e., State, county, and municipal).55 This is done because monthly information currently available on Federal construction is generally more satis factory than information on State and local public5 8 58 The main division in construction expenditures is between private and public projects. It is based on the ownership of the facilities under construc tion. Within the public category, however, two types of division between Federal and other public construction is possible, because of the grants of funds under Federal-aid programs. One is on the basis of ownership. The other, which is the one used in estimating expenditures for public construc tion, is based on the source of the construction funds. To illustrate, the Federal-Aid Highway Program provides Federal funds for State highway construction. Therefore, expenditures for State highway construction presented under an ownership classification exceed those for the same con struction under a source-of-funds classification by the amount of the Federal grants. The foregoing applies only to Federal grants. Federal loans for construc tion, such as those by the Rural Electrification Administration for the ex tension of power facilities, are considered to be the funds of the agency receiving them (i.e., the owner of the facilities) and therefore do not give rise to different results under the two classification systems. construction, and also because the Bureau of Labor Statistics is required to present a separate report of monthly Federal expenditures. The following section describes for each category of public construction the methods of estimating first the Federal, and then the non-Federal component. Residential building by the Federal Government has been confined to military installations, atomic energy facilities, and to such other Federal prop erty as dam construction and control areas, since completion in 1948 of the veterans’ temporary re-use housing program under title V of the Lanham Act. Expenditures for such construction are included in the “ Military and naval facilities” and “ Conservation and development” categories listed on page 51. All data currently reported in the public residential classification, therefore, represent outlays by non-Federal public agencies. By far the most important of these local agencies has been the New York City Housing Authority from which expenditures data are collected month ly. These data are a reconciliation of estimates by the contractor and by N YCHA engineers of the amount of work placed, and are prepared primarily to determine monthly payments to the contractors. For the smaller programs in other places, estimates are made by applying typical residential construction patterns to data collected by appropriate Bureau of Labor Statistics regional offices on the cost and estimated start and dura tion of the projects. Federal nonresidential building data are derived for each type of construction as follows: (1) Educational: Since completion of the Veter ans’ Educational Facilities Program in 1948, there has been little expenditure of Federal funds for construction of educational institutions. The relatively small amount of construction of schools on Federal properties (such as at remote Army bases or sites of large conservation projects) is included in other construction categories. (2) Hospital and institutional: The two major Federal programs of hospital construction— the Veterans Administration Program and the Na tional Hospital Program 56—make up the Federal component of this category. The Office of the Chief of Engineers and the Veterans Administra tion, which supervise the veterans hospital program, obtain monthly progress reports from8 88The construction program authorized by the Hospital Survey and Construction Act of 1946, Public Law 725, 79th Congress. EXPENDITURES— NEW CONSTRUCTION the field and make these reports available to the Bureau of Labor Statistics for transcription and tabulation. The National Hospital Program is, however, a Federal-aid program, and as such presents special estimating problems. Construction expenditures for the program as a whole are summarized by the Bureau of Labor Statistics from individual project progress reports collected by the State agencies administering the program and submitted to the Hospital Facilities Division of the Public Health Service. Of these total expenditures, the portion which represents Federal funds is then added to the veterans hospital expenditures to obtain the Federal component of the total public hospital category. (3) All other nonresidential: Separate expendi tures estimates are prepared for each type of build ing in this category. The Federal expenditures for these types of construction are relatively small, and are frequently included in other major categories (recreational buildings for military personnel, for example, are reported under “ Mili tary and naval facilities” )- The Federal expendi tures represented in this group are made by (1) the Public Buildings Administration on such structures as the General Accounting Office in Washington; (2) other Federal agencies, such as the National Advisory Committee for Aeronautics; and (3) by the United Nations for its headquarters in New York City, which is included here for want of a supranational classification. For the first and third, monthly progress reports are obtained from the agency supervising construc tion ; for the second, estimates are based on annual data from the Budget of the United States, and on Federal contract award information reported by the agency to the Bureau of Labor Statistics. Non-Federal nonresidential building estimates are derived, in general, by distributing the dollar value of contracts awarded over a number of months according to a typical pattern developed from experience records for actual construction projects. The contract award series used is that compiled by the Office of Economic Research in the General Services Administration for State and local public construction, using information col lected by the F. W. Dodge Corp. in the 37 Eastern States.67 To obtain data for the noncovered West ern States and to supplement Dodge coverage in5 7 57 See footnote 51 57 other areas, the Office of Economic Research rou tinely searches some 20 to 25 construction period icals and a number of other sources of construction information. The Bureau of Labor Statistics inflates OER award figures to compensate for undercoverage. For hospital construction, a modification is made in the general method of estimating non-Federal expenditures for public nonresidential buildings. Non-Federal expenditures (as well as Federal) in the grant-in-aid National Hospital Program are obtained from the Federal Security Agency, which administers it. The OER series is therefore used to estimate only the expenditures on public hos pital construction in which the Federal Govern ment does not participate. These expenditures are added to the non-Federal contributions to the National Hospital Program to obtain the nonFederal public component of the total public hospital category. For military and navalfacilities, the expenditures reported represent the volume of all construction, regardless of type, at Federal military installations. The relatively small amount of military construc tion by the States (armories, rifle ranges, and the like) are included with other public construction categories according to type of construction. The data for both construction agencies of the Department of Defense— the Office of the Chief of Engineers and the Bureau of Yards and Docks— are based on monthly progress reports for all construction projects at the facilities of the armed services. These reports reflect the observation or the actual measurement by service engineers of the volume of work accomplished during the month. Army Engineers data are summarized by the agency and published monthly for limited distribution. Data for construction by the Bureau of Yards and Docks, on the other hand, are sum marized by the Bureau of Labor Statistics from the basic field reports. Highway construction expenditures estimates are the sum of five components— expenditures on the Federal-Aid Highway Program,58 on State highways independent of the program, on county roads, municipal streets, and on roads on Federal lands. For the Federal-Aid Program, which accounts for roughly half of total highway ex penditures, monthly data on work actually placed fi8The construction program undertaken under the terms of the FederalAid Highway Acts of 1944 and 1948, which provide Federal funds to assist the States in road construction. 58 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES are available, which are based on observations by engineers of the Bureau of Public Roads, United States Department of Commerce. The Bureau of Public Roads tabulates these data for internal use, and supplies pertinent aggregates for the expendi tures estimates, showing Federal grants separately. Annual data on highways built on Federal lands are reported by the Bureau of Public Roads and monthly estimates are derived using trends on the Federal-Aid Program. To arrive at the total for Federal expenditures on highways, the amounts of Federal grants for highways and all expendi tures for highway construction on Federal lands are added. Annual expenditures figures (by government jurisdiction) for all roads and streets, prepared by the Bureau of Public Roads from special financial reports submitted by State highway departments, establish the general level of expenditures for highway construction outside the Federal-Aid Program. The month-to-month trend is based on reported data for the Federal-Aid Program as just described. The cumulative accuracy of these estimates is checked roughly, using highway contract award statistics prepared by the Bureau of Public Roads and the Office of Economic Research in the General Services Administration. For sewer and water and miscellaneous 'public service enterprises, the small amount of Federal construction is confined to Federal property, and is included with other major types of construction. The expenditures estimate covering sewer and water facilities and miscellaneous public service enterprises, therefore, represents State and local projects only. It is derived from a distribution of contract award data compiled by the Office of Economic Research, General Services Administra tion, and inflated for undercoverage. The esti mating procedure is the same as that described for the non-Federal component of public nonresidential building (see p. 57). Conservation and development expenditures, as reported currently, represent the volume of all new construction, regardless of type, at the sites of Federal projects for the conservation, develop ment, or control of the Nation’s water resources. Expenditures by non-Federal public agencies for construction for these purposes are represented in other categories, chiefly in “ All other public,” according to type. The Bureau of Reclamation and the Civil Works Division of the Office of the Chief of Engi neers are responsible for most Federal conservation construction. The comparatively small remain ing amount is supervised by some dozen other agencies, such as the Tennessee Valley Authority and the International Boundary and Water Com mission, United States and Mexico. For the two major construction agencies, ex penditures data are based on progress reports on work placed during the month as estimated by Government engineers. In both cases, tabula tion is by the Bureau of Labor Statistics from the individual project reports. With the exception of the Tennessee Valley Authority, which pre pares for the expenditures series a monthly sum mary of actual costs, monthly expenditures data are not available for the minor agencies. Annual figures for these agencies are obtained from the Budget of the United States, and are projected month by month on the basis of the trends re ported by the two major agencies. All other public construction not elsewhere classified covers such projects as airports, parks, athletic fields, memorials, navigational aids, and the like. Airfields have lately accounted for nearly half the total activity in this category, and the expenditures are therefore separately estimated. Most such construction is under taken with Federal financial aid (under the provisions of the Federal Airport Act of 1946). Progress of construction under each grant is reported monthly by regional offices to the central office of the Civil Aeronautics Administration, which administers the program. These data are tabulated by the Bureau of Labor Statistics, and the totals thus derived are divided to represent Federal and other contributions to the program. The other contributions are added to the relatively small expenditures for airport construction fi nanced without Federal assistance, as obtained from OER contract award data, to obtain nonFederal public expenditures for airports. For types of construction other than airfields in the “ All other” category, the small amount of Federal outlay is estimated monthly from the annual figures presented in the Budget of the United States. Non-Federal expenditures are obtained by a distribution of OER contract-award figures. XI. Compilation of Industrial-Injury Statistics89 Work-injury statistics are regularly compiled by the U. S. Labor Department’s Bureau of Labor Statistics in the following categories: (1) Annual estimates of the total volume of work injuries in each major industrial activity classification; (2) current quarterly injury-frequency rates for the primary manufacturing industry classifications; (3) annual injury-frequency rates and injuryseverity measures for manufacturing and non manufacturing industry classifications; and (4) accident-cause statistics and detailed injury-rate break-downs for selected industries. Of these series, the estimates of injury volume are con tinuous from 1936 and the annual frequency rates from 1926. The quarterly series was started in 1943. Efforts to standardize the methods of compiling work-injury statistics were initiated by the Bureau of Labor Statistics in 1911. In 1914, the Bureau called a formal conference of labor and workmen’s compensation officials and others interested in this subject. The work of this conference was carried forward in later years by the International Association of Industrial Accident Boards and Commissions, culminating in the publication of the first standardized procedures in 1920.5 60 In 9 1926, a sectional committee of the American Engineering Standards Committee, later the American Standards Association, undertook a re vision of these procedures. This work led to the publication in 1937 of the first American Standard Method of Compiling Industrial Injury Rates. This standard was subsequently revised in 1945 and is continuously under review by a sectional committee of the American Standards Associa tion. A second standard, the American Recom mended Practice for Compiling Industrial Acci dent Causes, developed under the American Standards Association procedures, was published in 1941. These two standards constitute the 59 By Frank S. McElroy of the Bureau's Industrial Hazards Branch. 60 Standardization of Industrial Accident Statistics, Bulletin No. 276 of the U. S. Department of Labor, Bureau of Labor Statistics, 1920. basis for all subsequent injury and accident statistics compiled by the Bureau of Labor Statistics. Injury-frequency rates are the primary meas ures of the incidence of work injuries. They indicate the relative level of hazard prevailing in different plants or industries during a specified period of time, or in the same plant or industry during different periods. The lack of compara bility inherent in simple injury totals, arising from variations in employment and operating time, is overcome by expressing the injuries in terms of a standard unit of exposure. By defini tion, the standard comparison injury-frequency rate is the average number of disabling work injuries for each million employee-hours worked. A disabling work injury is defined as any injury incurred in the course of and arising out of em ployment, which ( 1) results in death or any degree of permanent physical impairment, or (2 ) renders the injured person unable to work at any regu larly established job, which is open and available to him, throughout the hours of his regular shift on any day after the day of injury, including Sun days, holidays, and days on which the plant is shut down. Under this definition, the reporta bility of an injury for injury-statistics purposes is in no way related to the eligibility of the injured person for workmen’s compensation payments. In case of doubt as to whether or not an injured person is able to work, the attending physician’s decision is final. The severity of temporary injury is measured by the number of days during which the injured person was unable to work. For death and per manent impairment cases, the American Standard provides a table of economic time charges. These time charges, based upon an average working-life expectancy of 20 years for the entire working population, represent the average percentage of working ability lost as the result of specified im pairments, expressed in terms of unproductive days. Death, for example, representing the com plete loss of all future production by the injured person, is assigned a time charge of 6,000 mandays (i. e., 20 years of 300 days each). The loss 59 60 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES or loss of use of a single finger is estimated as resulting in an average reduction of 5 percent in working efficiency. By applying this percentage to the 20 -year life expectancy, the time charge for this type of injury is established as 300 man-days. The standard injury-severity rate, commonly used to compare the general level of injury severity in one plant or industry with that of another, weights each disabling injury with its established time charge and expresses the aggregate in terms of the average number of days lost for each 1,000 employee-hours worked. Limitations of the Series Estimates of Injury Volume. Comprehensive and continuing injury surveys by the Bureau of Labor Statistics, the Bureau of Mines, and the Interstate Commerce Commission provide accurate data for the estimates of injuries in manufacturing, mining, railway transportation, and public utility opera tions. Estimates for construction, trade, and miscel laneous transportation are based upon small sample studies augmented by reports of injuries filed with State workmen’s compensation agencies. Differences in the coverage of the State compensa tion acts and variations in the reporting require ments limit the usefulness of the basic data and introduce the possibility of considerable error in the final estimates, particularly in respect to nonfatal injuries. Data relating to agricultural injuries are ex tremely limited and in many respects are contra dictory. In large measure, the lack of basic figures results from the exclusion of agricultural operations from workmen’s compensation cover age in most States. Confusion in the figures which are available results from the difficulty of separat ing work-produced injuries from those which should be ascribed to home, traffic, or public accidents. The estimates for this segment of industry, therefore, are subject to substantial error. Injury Severity Bates. Some question has been raised in recent years regarding the significance of the severity rate as a true measure of injury severity. Objections are directed primarily to the use of employee-hours worked as the basis for comparison. Critics of the standard severity rate have pointed out that, for any specified number of employee-hours worked, six injuries each resulting in 1 day of lost time will produce the same severity rate as one injury which causes 6 days of lost time. The contention is that although hours worked are directly related to the occur rence of injuries, they have no bearing upon the severity of the injuries. It has been proposed that a more realistic measure of injury severity would be obtained by relating the aggregate time charges directly to the injuries which produced them— that the comparative measure of injury severity should be the average time charge per case. The average time charge has not yet been made a part of the standard. It is, however, computed and presented along with the standard severity rate in the Bureau’s annual and special industry surveys Sources and Methods of Surveys Annual Estimates. Injury statistics for particular segments of the economy are regularly compiled by a number of Federal agencies, such as the Bureau of Labor Statistics, the Bureau of Mines, the Bureau of Agricultural Economics, the Bureau of Employee’s Compensation, and the Interstate Commerce Commission. Most of the State work men’s compensation agencies prepare summaries of the cases reported to them and several private agencies, such as the National Safety Council, the Portland Cement Association, and the American Petroleum Institute, also compile current injury data. Summaries of the data compiled by these agencies constitute the base for the annual esti mates of the total volume of work injuries in the United States. Quarterly Injury Surveys. At the end of each quarter, questionnaires are sent to approximately 14,000 manufacturing establishments. The coop erating plants are requested to supply the following information for each month of the quarter ( 1) the number of workers employed; (2 ) the number of employee-hours worked; and (3 ) the number of disabling work injuries experienced by their em ployees with a break-down indicating the resulting type of disability as known at the time of prepar ing the report. Generally, about 11,000 reports are received in time for the quarterly tabulations. 61 INDUSTRIAL-INJURY STATISTICS Budget Bureau No. 44-R002.5. Approval expires N ov . 30,1950. B. Ii. S . 141 (Rev. 8-26-4 INDUSTRIAL INJURIES U. S. DEPARTMENT OF LABOR B U R E A U O F L A B O R S T A T IS T IC S INJURY SUMMARY, 1949 WASHINGTON (D o n ot list any injury more th an once. on other side.) T y p e o f d isability See request and instructions on other side. rv n n ciT T n n n A r r . EXPOSURE DATA C ode 1. Fatal._____________________ 10 Permanent total 2. Both arms............................... 21 3. Both legs.................................. 22 4. Both hands.............................. 23 5. Both feet.................................. 24 6. Both eyes (sight)................... 25 7. Other (describe over)............ 26 8. Sum of items 2 to 7 ........ X 9. 1 arm........................................ 31 Average number of employees, January 1-December 3 1 ,1 9 4 9 : 10. 1 hand...................................... 32 Include all w ho w orked in any capacity—production and related workers; force-account construction workers; ad m inistrative, supervisory, sales, technical, teaching, serv ice, and office personnel; and all others.................................... 11. 1 leg.......................................... 12. 1 foot....................................... 34 13. 1 thumb................................... 35 33 14. 1 finger................................... 36 15. 2 fingers (same hand)............ 37 16. 3 fingers (same hand) 38 W as th is establishm ent in operation throughout 1 9 4 9? -------- 17. 4 fingers (same hand)............ 39 40 If not, please indicate th e num ber of days on which it operated 18. Thumb and 1 finger (same hand). 19. Thumb and 2 fingers (same hand). 41 CLASSIFICATION DATA 20. Thumb and 3 fingers (same hand). 42 21. Thumb and 4 fingers (same hand). 43 o. T he principal ty p e of a ctiv ity of th is establishm ent is (i. e., m anufac turing, w holesale, retail, construction, public u tility , e t c .) : I f m a n u fa ctu rin g , b. answ er b and c. W hat products were m ost of your em ployees m aking during 1949? (L ist first th e product on w hich th e greatest num ber of em ployees worked, then others in descending order of em ployees involved. P lease be specific. A void generalities such a s “Ordnance” or “ M a chinery.”) (R product*listed first’ accounts for less than half the employees, show the approximate percentage of employees involved) c. W hat general ty p es of operations w ere performed by m ost of your em ployees in th e m anufacture of th ese products (e. g., foundry N u m b er o f cases Permanent partial (I n c lu d e lo s s or lo s s of u s e ) (Please complete this section even though there were no injuries to be reported) T o ta l num ber of em ployee-hours w orked by all em ployees during 1 9 4 9 - ---------------------------------------- See instructions 22. 1 great toe............................... 44 23. 2 great toes.............................. 45 24. Toe (not great toe)................ 46 25. 1 eye (loss of sight)............... 47 26. 1 ear Goss of hearing)........... 48 27. Both ears (loss of hearing)... 49 28. Other (describe over)............ 50 29.................................................. 51 30. operations, stam ping, w eaving, assem bly, etc.) ?. Sum of items 9 to 29 X N u m b er o f ca ses Temporary total T o ta l days o f d isa bility (O m it all Injuries res u ltin g in d isa bility o f les s th a n 1 d a y ) I f n o n m a n u f a c t u r i n g , answer d . d. W hat were th e principal services furnished by th is establishm ent during 1 9 4 9 and w hat w ere th e m aterials handled? (L ist first the service engaging th e largest num ber of em ployees, e. g., warehousing of clothing)........................................................... .............................. ......... 31. Cases of known duration: (a) Number causing disa bilities of 1, 2, or 3 days.............................. (6) Number causing disa bilities of 4 or more days............................... (c) Filled out by 62 Total of (a) and (6).. 63 32. Cases of unknown duration- 64 33. Sum of items 31 and 32... Grand total—- A l l Position 61 34. X in ju rie s reporte d X X X XX X N u m b e r o f ca se s Sum of items 1, 8, 30, andL33_. W --3 0 7 ia -« 62 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES A cooperative program under which the Mich igan Department of Labor and the Bureau of Labor Statistics jointly are to collect the quarterly injury data from Michigan establishments is to be inaugurated in 1950. Annual Surveys. At the end of each year, annual summary reports are requested from an additional mailing list of about 50,000 employers. Some 25.000 manufacturing establishments and about 15.000 nonmanufacturing establishments usually report in this survey. A joint program for the collection of annual injury data is already in effect in Pennsylvania. Under this cooperative arrangement, all annual reports from Pennsylvania establishments are collected by the Pennsylvania Department of Labor and Industry. The State prepares its own tabulations from these reports and transmits a copy of each report to the Bureau of Labor Statis tics for inclusion in the national totals. The report form used in the annual surveys (reproduced here) is somewhat more detailed than that used in the quarterly surveys. In addition to the summary figures necessary for the computa tion of injury-frequency rates, it includes a break down of the permanent impairment cases to show the number resulting in each of several specific types of impairment, as well as a summary of the time lost by employees because of temporary injuries. These additional data are used in the computation of injury-severity rates and severity averages. Special Industry Surveys. Special surveys are made within selected industries to obtain greater coverage and greater detail than is possible in the regular surveys, and to determine the prevail ing causes of the accidents which produce work injuries. In these surveys an attempt is made to obtain a mail report from every employer in the industry. The questionnaires cover the same items included in the annual survey form, but in Frequency rate—.^ um^er Computation Procedures Annual Estimates. All available material accumu lated in the injury surveys of the Bureau and of the other agencies previously mentioned are utilized in preparing the national estimates of injury volume. The tabulated injury totals pre pared by these agencies are related to the appro priate segments of the national employment and the estimates are computed by direct expansion to represent the probable volume of injuries in the total working population. Quarterly Injury Survey. Each report received is assigned an industry classification based upon the principal product or operation of the reporting plant, and totals of the reported figures are pre pared for each industry classification. From these totals, average injury-frequency rates for each month, for each quarter, and for the year to date are computed for each industry classification. In these computations, which conform to the pro visions of the American Standard Method of Compiling Industrial Injury Rates, the following formula is applied: disabling injuries multiplied by 1,000,000 Number of employee-hours worked No severity rates are computed, inasmuch as the final outcome of many of the injuries is not known at the time the reports are submitted. addition ask for the figures in a break-down of the operating divisions of the reporting plants. From these reports, frequency and severity rates are computed for each type of operation commonly found in the industry, for plants of various size groups, and for plants in various geographic areas. In addition, representatives of the Bureau visit a number of establishments in the selected industry and ask permission to review their original accident records. If permission is given, the Bureau rep resentative examines the records and for each recorded accident prepares a transcript indicating: ( 1) how, when, and where the accident occurred; (2 ) what unsafe conditions and/or unsafe acts contributed to the accident; and (3) what type of injury resulted. Through direct comparison between the employment in the reporting group and the total estimated employment in manufacturing, esti- INDUSTRIAL-INJURY STATISTICS mates of the total volume of fatal and nonfatal work injuries in manufacturing are prepared for each period. Annual Surveys. Data used in the computation of annual injury rates consist of ( 1) information reported on the annual injury summary form, and (2 ) the accumulated totals of the information reported during the year in the quarterly surveys. All reports are classified according to the prin cipal product or operation of the reporting establistment and totals of the reported data are pre pared for each industry classification. These totals are used in the computation of injuryfrequency and severity rates following, with one exception, the procedures specified in the American Standard Method of Compiling Industrial Injury Severity rate 63 Rates. The one exception is in the use of full time charges for each permanent-partial disa bility rather than the percentage charges per mitted under the standard. The computed rates for the various industry classifications are then weighted according to the total estimated em ployment in the classification and are combined in the computation of weighted rates for the major industry groups. Average time charges per case, as described previously, are also computed in this survey to supplement the standard severity rate. The frequency-rate formula used in the com putations for this survey is the same as that shown in the discussion of the quarterly survey proce dures. The severity rate and the average time charge are computed by the following formulas: Total days lost or charged multiplied by 1,000 Number of employee-hours worked A .. , Total days lost or charged Average time charge= ------------- -— ;-----------:-----— Number of disabling injuries Special Industry Studies. The computation of injury-frequency rates and severity measures from data collected in special industry surveys follows the same procedures described in the discussion of the quarterly and annual surveys. The acci dent-cause data collected by the field staff are analyzed on an individual case basis, according to the provisions of the American Recommended Practice for Compiling Industrial Accident Causes. The accident factors indicated by this analysis are tabulated in various break-downs, such as by department, occupation, operation or process, agency involved, and accident type. XII. Occupational Wages: Establishment Sampling61 Sampling is n e c e s s a r y in making occupational wage surveys because of the large number of establishments in many industries. This need exists even when a narrow industrial classification is studied. For instance, approximately 2,000 firms, employing 8 or more workers producing women’s and misses’ dresses, are located in the New York City area alone. Obviously they must be surveyed on a sample basis, if personal visits are involved. Otherwise an unduly large pro portion of the limited funds available to the Bureau of Labor Statistics for such work would be expended on a single industry-area group. Other reasons for sampling are to insure that a survey yields something more than an informed guess, and that reliable results are obtained at minimum cost. It is possible to reduce sources of error other than those ascribed to sampling by devoting a larger proportion of time to proper occupational classification, careful editing for accuracy of data and the like, than would be possible without the use of sampling. Otherwise errors, largely un predictable in their effects, might be more serious than sampling error. Hence, sampling some times leads to increased accuracy, not less, as is often assumed. It also reduces the time necessary to collect and process data; even with unlimited resources, the smaller the sample the earlier the publication of results. Collection of Information The choice among various methods of collecting wage information depends on the type of industry surveyed, and the nature of the data sought. Each method has its appropriate sampling pro cedure, and considerations important in one case may be trivial in another. Thus, mail questionnaire surveys are suitable if detailed occupational data are not desired and if*6 4 61 B y Samuel E. Cohen of the Bureau’s Division of Wage Statistics. 64 incentive methods of pay are uncommon. In dealing with a complex occupational structure— in which strict definitions of occupations are essential and information on practices such as vacations and sick leave is sought— personal visit surveys have been found desirable. In both cases, the sampling unit is the establishment, rather than the individual. Most industries do not have central sources showing individuals’ names and, if they did, the volume of collection work would make surveys on such a basis im possible in most cases. Surveys limited to a single occupational group, such as engineers, are an exception. For this profession, lists of names can be assembled and the individuals can be approached directly, by mail questionnaire. The sampling problems involved differ greatly in each particular survey, and are not discussed here. In establishment sampling for wage studies, no particular problem of conserving resources is encountered in most mail questionnaire surveys. The number of firms in a given industry and area is generally not so great that sizable economy in time or money can be effected by use of a refined sampling technique. Such surveys are often employed in those industries the extent of which is not fully determined. In these instances, the mail questionnaire is a useful device for gaining information as to the actual number of firms. The principal sampling difficulty is created by the nonrespondents. Failure to investigate the nature of the nonrespondents may introduce biases into a mail survey, since in a good many cases, they may not have answered owing to factors allied to wages. The direction and size of such bias cannot be predicted, and hence some personal visits are generally required in mail questionnaire surveys. In personal visit surveys, careful design is necessary in establishment sampling since the cost per schedule obtained is rather high. The object of any sampling procedure employed is to secure the desired accuracy with a minimum expenditure. Field representative visits to un wisely selected establishments adversely affect the accuracy of a survey relative to its cost. WAGES— ESTABLISHMENT SAMPLING General Sampling Procedures Some general rules apply to the sampling and collection of wage data, by field agent visit, as done by the Bureau of Labor Statistics. ( 1) All visits are made according to some fixed rule. If the plant is assigned, it must be visited, and a disposition made; a schedule must be ob tained unless the firm is outside the scope of the survey, or out of business. An agent is given no discretionary powers as to which particular firms are to be visited. Ease of obtaining a schedule does not influence the choice of individual firms for study. Every reasonable effort is made to get a schedule, once a firm is assigned. All firms within the scope of the particular survey are in cluded in the industry population from which the sample is drawn. (2 ) Personal judgment as to what constitutes representative firms is not relied upon. The sys tem of sampling employed is unbiased, i. e., there is no reason to expect that any characteristic of a given sample would be more likely to be greater than to be less than the results from a complete census. The procedure, if repeated over all pos sible samples of the same size, would give the same results as a census. In addition, the pro cedure is such that if successive samples were taken, all possible samples would occur with approximately equal frequency. (3) The system employed makes possible some appraisal of the sampling error of the results. Only then can it be determined objectively whether the sample should be increased in the interest of accuracy or decreased in the interest of economy. Determination of Industry Population. In general, the industry population is completely determined before sampling is begun. That is, a list of firms is assembled in some form, and, insofar as possible, grouped according to such pertinent character istics as are known, i. e., product, location, etc. Sources used for this purpose include listings provided by the various State departments of labor or unemployment compensation bureaus, trade directories, lists provided by trade associations, unions, or regulatory government agencies. If these lists are used in combination, care is taken to eliminate duplication. The size of the industry population may be ad justed as the result of the field work, if it is found 65 that a certain proportion of the firms assigned should not be covered, i. e., are out of business or engaged in a type of business not covered in the survey. The estimated total population is ad justed in accordance with the findings in the sample study. In some cases a preliminary sample study, or even a census, may be made of an industry in order to obtain essential data on various characteristics of establishments for use in the classification of the population. No wage data are obtained in such a preliminary survey. Size Limits. A restriction is usually placed upon the minimum size of firms to be studied. De pending on the industry and the distribution of total employment by size of plant, lower limits range from 5 employees in auto repair shops to 250 in airframe manufacturing. Otherwise, it would be necessary to expand greatly both the list of firms from which the sample is drawn and the sample studied. In addition, the sampling error of the results would be greater than that which is found when the minimum size of firm studied is limited. Another reason for such limitations is that the surveys deal primarily with occupational wages, and the smaller firms frequently do not provide the occupational specialization of duties that is necessary to define an occupation clearly. Exclusion of the smallest size firms is not likely to have a very great net effect on the average wage levels. The differences in over-all averages caused by such exclusions have, in every case investigated, been less than the sampling error normally expected. After establishing the minimum size for survey, every effort is made to insure that the results represent an unbiased estimate of the segment of an industry actually studied. Thus, firms of all sizes studied have a chance of being included in the survey. They are not necessarily all repre sented in the same proportion, but the dispro portion is adjusted by the assignment of proper weights (see p. 68 ). Approximate Sample Size Needed. No advance judgment is possible to determine a sample size uniformly sufficient for all purposes. For in stance, if the workers in one of the occupations studied, such as maintenance electricians, were paid almost uniform rates throughout a city, a 66 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES sample of very few establishments would suffice. But the same size sample would be clearly de ficient for any occupation in which rates varied considerably from plant to plant. A small sample might be sufficient for determining the average rate for jobs that are found in every establishment, but inadequate for an uncommon job. There fore, some occupation or characteristic must be selected that can be measured with the desired degree of accuracy, and about which the sample size can be developed. Some items will be meas ured more and others less accurately. The par ticular research worker’s knowledge of what is important in the given industry is indispensable, and there is no purely statistical substitute for it. If break-downs of the data by some character istic are sought, sufficiently large samples must be provided for each subgroup. For instance, in the women’s dress industry, data are customarily shown for the so-called contract shops and inside shops separately. These are considered separately and the appropriate sample size is chosen for each. In such cases, the accuracy of the data for the combined group is greater than it would be if the total sample were planned to show the combined group as a unit. After selection of the items used as the standard of accuracy, a determination is made of the ap proximate sample size needed to yield estimates of such items, within the specified degree of ac curacy. The following information is employed for this purpose. ( 1) Number of establishments in the particular industry. (2 ) An estimate of the coefficient of variation of the item being measured.62 Here, the results of any previous surveys of the same industry are most helpful. If such information is lacking, the coefficient must be estimated from other surveys of a similar nature, or (lacking such data) from experience in a related field. Knowing these two quantities, it is possible to arrive at the approximate sample size by the usual formula for the sampling error of the means of a finite population. 62 The coefficient of variation as used here is the ratio of the standard devia tion of establishment averages to the average of all establishment averages. Although wage levels themselves were increased, the relative dispersion of wages from plant to plant appears to remain relatively stationary for short periods. In recent years, the practice of granting general wage increases in cents-per-hour has caused slight shrinkage of the relative variation between establishments so that some slight overestimate of the coefficient of variation results from use of previous data. The usual formula for the sampling error of the mean— (1) S Vft where S (M ) is the sampling (standard) error of the mean, M, S (X ) is the standard deviation of the population averages (estimated from the sample, as a rule, though not necessarily), and ft is the sample size in terms of establishments— should be modified in this case to allow for the fact that in occupational wage surveys the samp ling is done without replacements from moderately small populations. Instead of ( 1) write _ /N—n S (X) s m ~ylN=i where S (M ), S (X), and ft are the same as above and N is the size of the finite population (in terms of establishments). Since the distribution of plant averages is ap proximately normal, it may be assumed with safety that two-thirds of all averages, based on samples of size ft, will lie within the range M ± S (M ), or that 95 percent will lie within the range M ± 2 S (M ). Therefore, it is necessary to solve the equation where K is a factor depending on the degree of confidence with which it is expected that the sample mean falls in the allowed range. Since it is sought to obtain the relative error and not the absolute error, division by the mean Mj yields Percent of error= K + - 7=\ N —l i n V where is the coefficient of variation. Solving for ft, the sample size, yields P E —is the permitted percentage of error N —number of establishments in the population —coefficient of variation of item studied V WAGES— ESTABLISHMENT SAMPLING K —number of standard errors on either side of mean needed to determine the desired confidence interval. (2 if the sample mean should differ from population mean by no more than the per mitted error in fewer than 5 cases out of a hundred, 1.645 if sample mean should differ from popula tion mean by no more than the permitted error in no more than 10 cases out of a hundred.) The appropriate value of K can be found in the tables of the normal curve for other situations. Tables corresponding to the various values of population sizes and coefficients of variation can be computed in advance. Stratification Procedures. The preceding discus sion on sample sizes is based on the assumption that the data presented are the sample average of establishment averages. However, in actual prac tice it is more usual to publish, not the average of plant averages, but the average for the individual workers in the occupation, i. e., the average pub lished is the weighted average of the individual earnings, not simply an average of the individual establishment averages. For instance, if the averages for tool and die makers in three of the sampled establishments are as follows: Number of tool and die makers Establishment 1______ ______ Establishment 2_____ _____ Establishment 3_____ _____ 4 5 1 Hourly earnings $1. 50 1. 80 1. 20 The weighted average hourly earnings derived from this sample is: (4 X 1.50) + (5 X 1.80) + (1X 1.20)_^n 4+5 + 1 The relatively large variation from plant to plant in the number of workers in an occupation means that the sampling problem is complicated by two quantities, each of which is subject to sampling error— the plant averages themselves and the number of workers involved. In order to reduce the effect of the latter element on the sampling error of the weighted mean, the sampling must be done, not at random from the whole population but from a series of subdivisions, or strata, in which plants of approximately the same size classifications are grouped. After deciding on sample size, allocation of the sample to the various size strata follows. This allocation should be as efficient as possible, i. e., 67 it should yield a smaller sampling error than any other allocations of a sample of the same size.63 The way to achieve the most efficient distribu tion, or optimum allocation, of the sample to the various size strata has been determined to be by the assignment of schedules in proportion to the total employment in the strata. For example, if a size stratum has 20 percent of the employment of the industry, 20 percent of the sample should be taken from that stratum. Strictly, the number taken should be jointly proportional to the total employment and the standard deviation of aver age earnings within each size stratum. However, little evidence exists to show that these standard deviations differ sufficiently from size group to size group to merit the additional labor of considering them as other than equal. It may appear that the preliminary sample size could be computed by solving the equation for the sampling error of the mean of a stratified sample selected according to the principle of optimum al location. Actually, this is scarcely ever possible in occupational wage surveys because the distribu tion of plants by size is such that the number of largest plants demanded under this theory is gen erally larger than the number that actually exist in the entire population. In order to compute the sampling error of the estimated mean, the actual number of schedules selected in the formula may be substituted for the error of the mean of a stratified sample. If the plant distribution by size is such that the error appears considerably smaller than desirable, the sample size may then be reduced. In general no modification is necessary except in those cases where several comparatively large firms account for a high proportion of the total employment in the area. Often, some slight enlargement of sam ple size is made to compensate for the irregular frequency of occurrence of fairly important occu pations. Peculiar problems of occupational structure oc cur in some industries. For instance, in the cotton textile industry, the industry is quite generally divided into integrated mills (those spinning yarn and weaving) and yarn mills (those spinning m N o reference to the relative cost of the collection of large schedules and o small or medium-sized schedules has been made up to this point. Consider, ing the large proportion of overhead (travel time, and time necessary to see the official giving the information), the method outlined herein has also been f ound the most efficient from the standpoint of total cost. Only in the largest firms has it been found worthwhile to resort to intraplant sampling. 68 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES alone) in some areas. The spinning mills con tribute no information regarding the wages of weavers, and, therefore, the two groups must be considered as separate populations in selecting samples. In a good many cases, it is possible to perform further stratification of establishments if informa tion is available regarding such characteristics as unionization, type of products, etc. In cases of this kind, it is often possible to classify the estab lishments by such characteristics within each size group and to set up small cells and to select one or more sample members from these cells. Cur rently, little is known as to the reduction of sam pling error by such procedures. Very often such information is not available, and no further strati fication is possible, and attempts to use fragmen tary information may result in an unknown bias. Regardless of stratification methods, once the cells have been established, the selection of the plants actually studied is purely random. This is true even though personal judgment is used in determining the make-up of the cells themselves. Estimation From the Sample. When the method of optimum allocation is used, the sample generally consists of all very large firms, a large proportion of the next size group, a smaller proportion of the next, and so on. Obviously, straight addition of the data for all of these establishments would yield a bias toward the large firms. Therefore, smaller firms are assigned weights that are the inverse of the sampling ratio for the stratum from which they are chosen, e. g., if a third of the plants are selected from a cell, they are all given a weight of 3. By use of punch-card methods, the compu tation of individual strata averages is avoided, and the estimated total workers and aggregate earnings for all strata combined are computed simultane ously. To illustrate the use of weights, suppose that establishment 1 was drawn from a cell in which half the plants were used in a sample. It is therefore given a weight of 2 . Establishment 2 , on the other hand, was taken with certainty, i. e., it stood in a class by itself, and is given a weight of 1 . Establishment 3 was taken from a cell where a fourth of all plants were used in the sample, and hence is given a weight of 4. The calculations are, therefore: T ota l w eighted ea rn in g s Establishment 1_______ 2 X 4 X 1 . 50 Establishment 2_______ 1 X 5 X 1 . 80 Establishment 3_______ 4 X 1 X 1 . 20 Estimated universe____ 25. 80 T o ta l w eighted w ork ers 2X4 1X 5 4X1 17 The estimated average hourly earnings in this case is —y f— =$1-52. Limitations of Sampling Theory Difficulties are encountered in attempting to calculate rigorously the sampling error of weighted averages of the type published by the Bureau of Labor Statistics in its occupational wage studies. These averages are strictly the ratio of two random and highly correlated variables, the total earnings and number of workers, information on whose exact distribution is not available in its entirety. Under favorable conditions, the sampling error of such ratios can be approximated but this cannot be done too well if the relative variation of the denominators (number of workers) is large, as is often the case. The necessary computations are also quite expensive. Experimental work indicates that the relative sampling error of the average of plant averages (within any size group) is not materially different from that of the weighted average. By insuring inclusion of all large plants, the sampling error of the over-all average of any given item is likely to be reasonable, especially for specialized occupations likely to be found in large establishments. The assumption that the optimum sample de sign for one item is also optimum for others may not be strictly sound. But it is necessary to work with only one sample for all purposes. In any event, wages and wage practices are highly corre lated, and departures from optimum design can not be too great in most cases (i. e., number of workers receiving specified types of vacation priv ileges). Samples of the design used are also reasonably efficient for estimating the total em ployment in the industry being studied, as well as the total number of workers in specific occupa tions. Illustrative Example. An illustration is the selec tion of a sample for the power laundry industry WAGES— ESTABLISHMENT SAMPLING in a large city, where there were 150 establish ments of more than 20 workers. According to a previous survey, the average coefficient of varia tion for this industry in this city was approxi mately 1 to 5, corresponding to a sample size of about 30. In the total population, the employ ment and firms were distributed as follows: N u m b e r e m p lo y e d 251 and over_____________________ 101-250............. 51-100___________________________ 21-50____________________________ T otal_____________________ 150 P la n te W ork ers 2 700 426,300 63 4, 600 43 1, 400 13,000 Hence, the final sample would be distributed as follows: 69 13660 X 30 = 1.6 from the largest size group 0300 13666 X B O " 14-5 from the second group f^666X3Q= lQ-6 from the third group 1^666X 30 = 3 .2 from the fourth group In actual practice, since a whole number of firms must be taken, the final sample consisted of 2 of the largest with a weight of 1; 14 of the next group with a weight of 3; 11 of the next group with a weight of 6 ; and 4 of the smallest firms with a weight of 1 1 ; a total of 31 schedules. The departure from the numbers shown previously was made in order to simplify the weighting pro cedure. XIII. Occupational Wages: Conduct of Surveys64 The methods of making occupational wage sur veys 65 have evolved from long years of experience by the Bureau of Labor Statistics in the collec tion and analysis of facts relating to the country’s wage structure. These surveys are designed to provide the public with wage and related infor mation in a form that emphasizes the major factors that make for differences in wage levels. Typically, wage levels in the United States vary by occupation and industry, and within industry by geographic location, size of establishment, size of community, extent of unionization, and method of wage payment. Much of the planning, the sampling procedures, methods of collection, and tabulation are influenced by these factors of vari ability. Four types of wage surveys are currently con ducted by the Bureau; each is distinguished by the population unit that is considered in sampling. First, there is the Nation-wide industry survey, in which samples are selected with reference to the industry as a whole. However, these samples are so stratified that regional and locality data are also provided. The second type is the industry-by-area survey. In these studies samples are selected with reference to a particular industry in an area. The third type is the cross-industry study, in which samples are selected with refer ence to all industries in an area. This type of study is mainly practical in occupations, such as office occupations, that are found in most indus tries within an area. Fourth, in surveys of pro fessional workers, sampling is carried out with reference to all workers in a profession rather than in any establishments with which they may be associated. Each type of survey yields some what different results from the others, but essen tially the methods of making the studies, except for the sampling, have certain elements in common. The comparatively high cost of occupational wage studies makes it impracticable to study all industries annually. In all instances, it is the*6 0 7 8 84 B y Kermit B. M ohn of the Bureau’s Division of Wage Statistics. 68 This article relates to studies in which data are derived from employer pay-roll records or from individual workers. For a description of surveys based on union records, see note VI in this series, Preparation of Union Scales of Wages and Hours (p. 29). 70 Bureau’s objective to provide occupational wage data for the leading industries at least every 2 or 3 years, however. Beginning in 1947, annual studies have been made of a group of some 20 industries on an industry-by-area basis for the twofold purpose of providing information more frequently than is usual and for establishing a wage series to portray the movement of wages in these industries. In selecting industries for study in any year, the Bureau is guided by the interest of the public in information as well as by its anticipation of important developments in par ticular industries. The wage data considered of primary impor tance and adopted for occupational studies are straight-time rates, excluding overtime and shift premiums.66 Rates as such can be obtained, how ever, only for workers paid on a time rate basis. For those workers covered by piece work or other incentive systems of wage payment, a rate of earnings is obtained by dividing earnings, ex clusive of overtime and shift premiums, by the total time worked during a pay-roll period. For plant, or production, workers rates are usually obtained on an hourly basis; weekly rates are obtained for office workers; and annual or month ly rates are most frequently measured for pro fessional workers. In general, the surveys at tempt to use the measure of wages that is typical in the industry and occupations being studied. In conjunction with the wage rates, the number and sex of the workers is obtained for use in tabulations based on this characteristic. Supplementing the wage data, information on various related practices is also studied. Among these items are shift premiums, paid holidays, paid vacations, and insurance and pension plans. The findings of these studies are used extensively in collective bargaining, as well as in private wage determinations. They are also used in making interregional and intercity comparisons of wage rates, in order to facilitate decisions with respect to plant location. In addition, they provide nec essary information for the formulation of public policy on wages, as in minimum wage legislation, and for the analysis of trends in economic develop ment. MStraight-time rates more accurately reflect a worker’s payment for a uniform unit of work as compared with gross earnings which are affected by the length of time worked. WAGES— CONDUCT OF SURVEYS Limitations of Surveys The value of successive Bureau occupational wage studies of a particular industry for measuring changes occurring between the survey periods usually has certain limitations, even though the intervals between surveys may be as short as a year. These limitations arise from the fact that, while individual studies are made under the same general procedures, they are designed to meet specific industry and labor needs of the moment in terms of occupational classifications, job descrip tions, types of data, and other factors. For exam ple, surveys have been made annually for a number of industries. Consultations with management and labor are held before each study, however, and necessary changes are adopted. For that reason alone, the consecutive studies are not strictly comparable. Although this may be dis advantageous in some respects, the primary pur pose of the studies is to indicate wage levels as of a particular time rather than to attempt to portray the movement from one survey period to another. Study Methods and Sources In developing plans for the surveys, Bureau consultations are held with industry and labor representatives, through its Business and Labor Advisory Kesearch Committees and directly with management and labor representatives in each industry. Subjects dealt with generally relate to technical matters regarding time of studies, selec tion of jobs for study, preparation of job descrip tions, and kinds of special data needed by those directly interested in the study. Planning of Surveys. Industry scope in the wage surveys is practically always expressed in terms of the classification system of the Standard Industrial Classification Manual issued by the Bureau of the Budget. Varying according to the particular study, the scope may range from a part of a 4-digit category to a combination of several 3-digit clas sifications; the basic criterion is that the industry being studied should represent a fairly homogene ous unit insofar as wages and occupations are con cerned. In order to increase efficiency in the col lection of data, the scope of the industry is usually modified to exclude all establishments with fewer than a specified number of workers. This mini mum size limitation is established after a study of 71 the possible effects this limitation would have on the results. Another practical reason for the adoption of size limitations is the difficulty en countered in classifying workers in the very small establishments where the same degree of specializa tion found in large plants does not exist. It is not essential to provide data for all occupa tions in an industry in order to describe an indus try’s wage structure. In addition to the increased cost of obtaining occupational data for all occupa tions in an industry, the usefulness of such data is seriously limited because of the wide differences in occupational structure from plant to plant. In recent years, therefore, a list of key jobs has been selected which represents the total occu pational structure ranging from the least- to the highest-skilled workers. In the selection of such jobs, the following criteria have been useful: (1) numerical importance of job, measured by workers in the job; (2) definiteness of occupation; (3) stability of occupation from period to period; (4) prevalence of occupation among establish ments; and (5) historical importance of job in establishing wage rates. In addition to these five criteria for selecting individual jobs, the entire list is selected in order to represent the complete range of rates in the wage structure from high to low, since it is assumed that the rates of pay for the key jobs can be used as bench marks for measuring rates of the nonselected jobs by relating the one group to the other. Each key occupation is carefully defined in order to obtain maximum comparability of jobs from plant to plant. In preparing such defini tions, the suggestions of industry and labor repre sentatives are of great value. A job description that is to be used in a survey involving many establishments includes the major determining characteristics of the job. It is flexible enough, however, with respect to minor variations to permit interplant comparisons based primarily on the major elements. Above all, classification of workers is based on job descriptions and not on titles of jobs used in various plants. Collection of Data. The results obtained from a complete coverage of all establishments in an industry, as compared with those obtained from a carefully selected sample, would not warrant the generally large expenditure involved even if the funds were available. For this reason, 72 TECHNIQUES OF PREPARING MAJOR BLS STATISTICAL SERIES the Bureau's surveys are practically always made on a sample basis.67 Data for most of the occupational wage surveys are obtained by personal field visit to each establishment in the sample on forms of the type here reproduced. Experience with these surveys has proved personal visits are necessary in most industries to obtain a high degree of uniformity in the data, especially in the classification of workers by occupation. The field work is administered through the Bureau's five regional offices. Sur veys of professional workers require direct report ing from the workers themselves. These, as well as a limited number of industries which have highly standardized occupations, are surveyed by mail questionnaire. The basic source of information, when estab lishments are visited, is the pay roll or other com pany records. In order to classify workers by occupation, discussions are held with company personnel acquainted with job content. In addi tion, the Bureau representative usually observes the workers on duty. Information on practices related to wages is obtained from the authorized company official. Compilation Procedures Individual company reports are forwarded to the respective regional offices of the Bureau which have the primary responsibility for reviewing, editing, and preparing the schedules for tabulation. The schedules are reviewed for reasonableness and w The sampling procedure is described in the preceding note (p. 64). completeness. Various codes and weights are assigned to each schedule as a preliminary step to the preparation of punch cards. Next the sched ules are forwarded to Washington where the information is punched on cards and listings of the punched cards are made. The schedules and listings are returned to the regional office where they are again checked. The initial tabulations prepared by the regional offices, consisting of data for a wage area (usually a central city and such surrounding territory as constitutes a rela tively homogeneous area with respect to wage structure), are made directly from the listings which are set up for the facilitation of these tabulations. In Nation-wide studies, additional tabulations are prepared to reflect the differences in wage levels resulting from the various factors previously mentioned. All final summary tables, except the area reports, are prepared in Washing ton. Airea reports are issued on standardized forms. Nation-wide and regional summaries are published in bulletins, press releases, and articles in the Monthly Labor Review. Two methods of weighting are used in comput ing the various averages. In computing the aver age for any occupation, all rates are first multiplied by the number of workers receiving those rates and then dividing by the total number of workers so that the rate for each worker will have its proportionate influence on the over-all average. The other weighting factor is one assigned to each establishment so as to adjust the data to reflect the proportionate influence of the establishment in the entire population rather than in the sample. (See preceding article.) STRICTLY CONFIDENTIAL B. L s 1476 OWR-2 (Rev. 49) Page-------------Schedule number Page number Establishment n a m e _________________________ _________________________ Occupational Rates— Pay-roll period fr o m ___________________ t o ____________________ , 19____ C odes N O c c u p a t io n a n d G CO rad e Sex W Ind. (1 ) Occup. u m be r OF M .W .P . H ours E a r n in g s orkers R ate or S t r a ig h t T im e H ourly E a r n in g s M G in im u m u a r a n teed R U n io n R ate ate G r. (3) (4) (5) (6) (V (8) (9) U. S. GOVERNMENT PRINTING OFFICE: 1 9 50 (10)