The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Review Vol. 69, No. 9 Novem ber 1987 .") Is l i^ lilh D is tric t M a n u fa c tu r in g K n d a n g rrc d ? I(> T h c t i r r a t H u ll M a rk e ts 1924-29 a n d I9S2--87: S p e c u la tiv e H u b b le s o r K c o n o m ic F u n d a m e n ta ls ? The Review is published 10 times p e r year by the Research and Public Inform ation Department o f the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available to the public fre e o f charge. M ail requests f o r subscriptions, back issues, o r address changes to: Research and Public Inform ation Department, Federal Reserve Bank o f St. Louis, P.O. Bojc 442, St. Louis, M issouri 63166. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal Reserve Bank o f St. Louis o r the Federal Reserve System. A rticles herein may be reprinted provided the source is credited. Please provide the Bank’s Research and Public Information Department with a copy o f reprinted material. F ederal Reserve Bank of St. Louis Review November 1987 In This Issue . . . Manufacturing em ployment in the nation has declined since 1979, leading some to conclude that "deindustrialization” has been taking place. Other mea sures o f manufacturing, such as output and productivity, however, suggest substantial progress in the nation's manufacturing sector. Not all regions have shared in this progress; industrial activity has shifted away from northern indus trial areas to the South and West in recent years. Given these regional variations, Thomas B. Mandelbaum evaluates the performance o f manufacturing in the Eighth Federal Reserve District in ‘‘Is Eighth District Manufacturing Endan gered?” He uses em ployment data and several measures o f output to compare regional with national manufacturing activity between 1972 and 1985. M andel baum concludes that the expansion o f District manufacturing was closely similar to that o f the nation as a w hole during this period; the District experienced neither the severe manufacturing decline associated with the Rust Belt nor the sharp expansions o f the South and West. The parallel growth o f District and national manufacturing points out the similarities in industrial composition, labor productivity and unit labor costs between the two. * * * Many people attribute the 1987 stock market crash to a bursting speculative bubble, much like the one blam ed for the 1929 crash in stock prices. The belief that speculation might cause a persistent deviation in stock prices from prices consistent with the underlying fundamentals is important. At the time of the 1929 crash, this belief spawned legislative proposals designed to curb credit for speculation, amend the National Banking and Federal Reseive Acts, impose an excise tax on stock sales and regulate the activities o f stock exchanges and investment trusts. Similarly, various proposals to alter the structure o f financial markets have follow ed the recent crash in stock prices. G. J. Santoni examines the “speculation issue” in the second article in this Review, “The Great Bull Markets o f 1924—29 and 1982-87: Speculative Bubbles or Economic Fundamentals?” Santoni compares a theoiv o f stock prices based on fundamentals to one that allows for bubbles, then examines data from the 1920s and 1980s to determine which theory is supported by the evidence. He concludes that the evidence does not support the notion that speculation caused stock prices to deviate persistently from those consistent with the fundamentals. 3 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Is Eighth District Manufacturing Endanger ed? Thomas B. Mandelbaum E l MPLOYMENT in U.S. manufacturing industries has declined more than 9 percent since 1979, casting doubt about the stability of our industrial base. Other indicators o f manufacturing activity, however, suggest a more favorable evaluation. Real output in manufac turing, for example, has increased 16.5 percent since 1979. This output growth, achieved with a shrinking labor input, reflects a gain in productivity per worker. Moreover, the proportion o f the nation’s real GNP originating in manufacturing has remained remark ably stable over the past 40 years.1 Despite this stability at the national level, a major shift o f the location of manufacturing activity among regions has occurred. While declining in the "Rust Belt,” manufacturing activity has posted solid gains in the West and the "Sun Belt."- Between 1947 and 1985, the share o f the nation’s manufactured goods pro duced in the Middle Atlantic and East North Central census regions dropped from 60 to 40 percent.' Ibis decline was offset by an increase in the South and Thomas B. Mandelbaum is an economist at the Federal Reserve Bank of St. Louis. Thomas A. Poilmann provided research assistance. 'For an analysis of the nation’s manufacturing performance, see Tatom (1986aand 1986b). SeeO tt(1987)foralong-run perspective on structural changes of the U.S. economy. 2See Crandall (1986), for an analysis of regional shifts of U.S. manufacturing. 3This statement refers to the percentage of gross value added in manufacturing, published by the U.S. Bureau of the Census in Census of Manufactures and Annual Survey of Manufactures. Gross value added is described in the shaded box on the next page. The Middle Atlantic census region includes New Jersey, New York and Pennsylvania; the East North Central region includes Illinois, Indi ana, Michigan, Ohio and Wisconsin. West from 26 percent in 1947 to 46 percent in 1985 with little change in the share contributed bv New England and the West North Central states.4 This article compares the performance o f manufac turing in the Eighth Federal Reserve District with that in the nation. Its purpose is to determine whether regional shifts of manufacturing noted elsewhere have also occurred in the Eighth District, which is not entirely in either the Sun or Rust Belts.1 MANUFACTURING PERFORMANCE IN THE EIGHTH DISTRICT In this article, em ployment data and three mea sures o f manufacturing output are used to evaluate manufacturing performance in the District. These three output measures are manufacturing product IMP), gross value added (GVA), and value o f shipments IVS). Each indicator is described in the shaded insert on page 00. An appendix outlines the m ethodology used to estimate the Eighth District’s MP. Several indi cators o f manufacturing output w ere used to gauge the consistency of the analysis. 4The New England region includes Connecticut, Massachusetts, Maine, New Hampshire, Rhode Island and Vermont; the West North Central region includes Iowa, Kansas, Minnesota, Missouri, Ne braska, North Dakota and South Dakota. Except for the states in the Middle Atlantic and East North Central regions the rest of the states make up the South and the West. 5The Eighth Federal Reserve District includes Arkansas and parts of Illinois, Indiana, Kentucky, Mississippi, Missouri and Tennessee. Due to data limitations, however, only data from Arkansas, Ken tucky, Missouri and Tennessee are used in the analysis. 5 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Measures of District Manufacturing Output Three measures of District manufacturing output were used in this article. Due to data limitations, the sum of data for the four states that dominate the District’s economy —- Arkansas, Kentucky, Missouri and Tennes see — is used to represent the District. Manufacturing Product (MP) for the nation is the same as "GNP originating in manufacturing” in the U.S. Com merce Department's national income and product ac counts (NIPA). It is conceptually similar to the economic measure of value-added. This measure is not consist ently available on a state or regional basis and was estimated for the District by the author using earnings, employment, payroll and gross-value-added data. The technical appendix describes the methodology used in its construction. The Value o f Shipments (VS), published by the U.S. Bureau of the Census, is the received net selling value of products shipped from manufacturing establishments, f.o.b. plant after discounts and excluding freight charges and excise taxes. The measure includes intermediate manufactured products purchased as inputs, so that it tends to be inflated by double-counting of products made by one manufacturer and sold as inputs to an other. In addition, the value of shipments reflects the All measures are adjusted for inflation (1982 prices) using the nation's implicit price deflator for manufac turing. Due to data limitations, the District analysis focuses on the 1972-85 period. Manufacturing Growth: Eighth District vs. the United States Employment Trends. Chart 1 shows that the Dis trict's total wage and salary employment, which equals about 7 percent of U.S. total employment, closely follow ed movements in national employment since the early 1970s. The similar growth o f total em ployment in the region is not surprising; there is a close similarity between the industrial compositions o f the regional and national work forces. The largest differences between the region’s and nation’s indus trial structures are a slightly smaller proportion o f the District econom y accounted for by the services sector and a slightly larger share accounted for by manufac http://fraser.stlouisfed.org/ 6 Federal Reserve Bank of St. Louis costs of business services of the manufacturer, such as maintenance and repair, engineering, consulting, re search and advertising. These services are assigned to service-producing sectors rather than manufacturing in the NIPA measures of manufacturing output. Since some of the intermediate inputs and business services may be purchased from other areas, a region’s value of ship ments may reflect production which originated in other regions. The value of shipments also differs from the NIPA manufacturing output measure in that VS excludes the output of establishments that perform the administra tive and auxiliary functions of a manufacturing enter prise, such as manufacturing headquarters. Gross Value Added (GVA), published by the U.S. Bureau of the Census, is the value of manufacturing shipments minus the value of materials, supplies, fuel and pur chased electricity used in production. The gross-valueadded measure avoids the duplication in the value of shipments data resulting from the use of products of some manufacturing establishments as materials by oth ers. But unlike the NIPA output measure, the grossvalue-added data includes the value of business services and excludes the output of administrative establish ments. turing." In 1986, manufacturing em ployed 21.4 percent of the District’s wage and salary workers and 19.1 percent o f the nation’s. As chart 2 shows, District manufacturing em ploy ment has also followed national trends since 1972.7 The number o f manufacturing workers peaked in 1979, then declined cyclically through 1982. In the current recovery period, manufacturing em ployment rebounded sharply in 1984 before resuming its decline in recent years. District manufacturing employment 6See Mandelbaum (1987) for a more complete discussion of the similarities of the region’s and nation’s employment compositions. 7A t-test of the average difference between District and U.S. annual growth rates of manufacturing employment, 1973-85, yields a tstatistic of - 0.46, indicating the difference is not statistically signifi cant at the .05 level. The period begins in 1973 rather than 1972, because 1972 is the first observation and this observation is used in calculating the 1973 growth rate. FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 C h a rt 1 Total Employment in 1986 was 1.41 million, almost 8 percent below its 1979 peak level and roughly equal to its 1972 level. Output Growth. In contrast to employment, District manufacturing output, like that in the nation, has grown substantially. As chart 3 shows, both regional and national manufacturing output (MP) declined in recession years but increased sharply during business cycle upturns. The net result was a substantial output gain over the period. The chart also shows that the District’s manufactur ing output has closely followed national trends. The first line o f table 1 shows the close similarity between regional and national growth in various measures of output. The District’s 2.6 percent average annual growth MP during the 1973-85 period was statistically indistinguishable from the nation's 2.9 percent pace. Regardless o f the output measure used, there was little difference between annual growth rates o f re gional and national manufacturing output." The real value o f manufacturing output in the Dis trict, as measured by MP, was $50.6 billion (1982 prices) in 1985. This represents a 7.5 percent gain between 1979 and 1985, a period in which declining em ployment trends intensified concerns about the health o f the manufacturing sector. 8T-tests of the average differences between District and U.S. annual growth rates, 1973-85, of MP, GVA and VS yield t-statistics of 0.54, -0 .2 8 and -1 .5 9 , respectively. None of these is significantly different than zero, in the statistical sense, at the .05 significance level. 7 NOVEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS C h a rt 2 M anufacturing Employment M i l l i o ns A n n u a l D a ta Mi ll i o n s 1.55 19.00 18.25 Individual Industry Growth The similarity o f manufacturing output growth in the District and the United States could mask substan tial differences between the regional and national growth in individual industry groups. Similar' growth of total manufacturing output could result if stronger gr owth o f some regional subsectors offset slowerthan-national growth in others. Each o f the industry groups o f the Eighth District manufacturing sector, however, grew at near the na tional pace. Although the growth rates o f output for most o f the District industry groups differed som e what from the national rates (see table 1), none o f the these differences is larger than w ould be expected due http://fraser.stlouisfed.org/ 8 Federal Reserve Bank of St. Louis to the chance var iation o f the data." This result holds regardless o f the output measure used. Industrial Composition Even with identical regional and national growth rates for each industry, overall manufacturing could differ considerably if the industrial compositions of 9T-tests of the average differences between District and U.S. annual growth rates for each output measure for each manufacturing indus try group were conducted. None of these is statistically different from zero at the .05 level of significance. FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Chart 3 Real M anufacturing O utput 1972 73 74 75 76 77 78 the regional and national manufacturing sectors var ied substantially. For example, if regional manufactur ing were concentrated in slow-growing industries (like prim aiy metal production), then the District s overall manufacturing growth w ould tend to trail the national expansion. The diversification o f regional and national manu facturing, however, has been quite similar. Chart 4 compares the percent distribution o f District and U.S. manufacturing output in 1985 (as indicated bv MPI among all the major industiy groups. Most are of similar relative size. The sector in which the District share varied the most from the national average in 1985 was nonelectrical machineiy. This sector ac 79 80 81 82 83 84 85 1986 counted for 14.8 percent of District MP compared with 17.4 percent nationally, hardly a dramatic difference. Earlier data show that overall structural similarity between District and national manufacturing has ex isted at least since 1972. Regional Productivity Gains The increases in District manufacturing output since 1972 with little change in manufacturing em ployment imply an increase in labor productivity. In fact, labor productivity o f District manufacturing (MP per manufacturing worker I increased at a 2.5 percent 9 NOVEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS Table 1 Average Annual Growth Rates of Real Manufacturing Output by Industry: 1973-85 Manufacturing Product District Total Manufacturing Durable Goods Lumber and wood products Furniture and fixtures Primary metal industries Fabricated metal products Machinery, except electrical Electronic equipment Transportation equipment Stone, clay and glass products Instrument and related products Miscellaneous industries Nondurable Goods Food and kindred products Textile mill products Apparel Paper and allied products Printing and publishing Chemicals and allied products Petroleum and coal products Tobacco manufacturers Rubber and miscellaneous Leather and leather products 2.6% U.S. 2.9% Gross Value Added1 District 5.3% U.S. 5.1% Value of Shipments1 District 6.6% U.S. 5.5% 1.2 2.3 0.0 3.5 8.6 3.9 2.7 1.9 6.3 3.3 1.7 2.4 -1 .7 1.7 7.0 6.6 2.8 1.2 5.7 2.9 2.3 3.4 4.8 4.7 13.3 5.6 9.3 1.8 N.A. 5.1 3.8 4.5 3.0 3.6 10.8 8.3 7.9 3.8 7.5 2.4 1.0 3.6 4.8 4.6 14.5 7.5 15.5 2.2 N.A. 3.9 3.5 4.9 2.6 3.9 11.8 8.1 9.5 4.3 7.8 2.8 2.3 1.1 0.9 4.1 2.9 1.7 3.8 -3 .2 5.8 -1 .1 2.1 2.0 1.4 3.0 2.4 3.0 0.2 - 1 .5 4.1 -1 .1 3.1 1.8 3.0 3.8 4.6 5.3 N.A. N.A. 6.8 6.5 3.1 3.4 3.3 5.1 3.1 5.9 2.3 0.2 7.3 -0 .7 3.3 2.6 2.5 4.5 4.8 6.3 N.A. N.A. 7.5 6.6 2.8 3.6 2.8 5.5 3.3 7.2 7.1 -1 .7 8.9 0.0 NOTE: N.A. indicates data not available. 'Data for 1979-81 are not available, so growth rates for 1979,1980,1981 and1982 are excluded from the average growth rates. com pounded annual rate between 1972 and 1985. Table 2 shows slightly faster growth when labor pro ductivity is measured by GVA per worker and VS per worker."’ The growth of total manufacturing output and labor productivity in the region indicate that, rather than undergoing a dramatic decline or “deindustrializa- '“Because no regional data for GVA and VS are available for 1979SI, it is impossible to compute average annual growth rates for those variables that are comparable to the average annual growth rates for MP. Therefore, compounded annual rates, which require only the initial and terminal years of the periods, are used to indicate average growth. In each productivity measure, the number of manu facturing workers are from the U.S. Bureau of the Census’ Annual Survey of Manufactures and Census of Manufactures. 10 tion,” the District’s manufacturing sector — like the nation’s — is expanding and becom ing more produc tive. Operating Ratios Labor productivity and unit labor costs o f a region’s manufacturing sector relative to the rest o f the nation are related to the region’s competitive position in national markets. A comparison o f changes in the regional and national operating ratios reveals whether the District is keeping pace with improvements at the national level. Table 2 compares the 1985 levels and the com pounded annual growth rates of labor productivity FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Chart 4 Composition of District and U.S. Manufacturing Output, 1985 Percent of t o t a l Nonelectrical Machinery T r a nsp orta tio n Equipment Food and K i n d re d Products E l ec t r i c a l E q u i p m e n t Chemicals, A llied Products Fabricated Metals Printing and Publishing Rubber and Plastics P a p e r and All ied Products Primary Metals Apparel, Textile Products S t o n e , C l a y and Glass Lumber and Wood Products Furniture and F ix tu re s Tobacco Products Instruments Miscellaneous Le a th e r Products P e t r o l e u m , Coal Products Textile Mi l l Products 10 and unit labor costs using each o f the three measures o f output. Unit labor costs are measured by payroll per unit o f output." Total District payroll per dollar o f MP, "The payroll data is published by the U.S. Bureau of the Census in the Census of Manufactures and the Annual Survey of Manufactures. 15 20 measured in 1982 dollars, was $0.49, almost identical to the $0.50 national level. In addition to similar levels, It includes gross earnings paid to an employees, but excludes employer contributions for social insurance and payments to propri etors or partners of unincorporated establishments. 11 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Table 2 Manufacturing Unit Labor Costs and Labor Productivity Compounded Annual Growth Rate 1972-85 1985 Level Labor Productivity MP/worker GVA/worker VS/worker Unit Labor Costs Payroll/MP Payroll/GVA Payroll/VS Eighth District U.S. $ 38,400 50,800 124,300 $ 42,100 52,600 119,900 0.49 0.37 0.15 0.50 0.40 0.18 Eighth District 2.5% 3.0 3.8 - 2 .3 - 2 .7 -3 .4 U.S. 2.8% 3.0 3.5 -2 .7 -2 .9 -3 .4 NOTE: See text for variable definitions. Payroll and output data in constant 1982 dollars. Productivity figures are rounded to the nearest $100. table 2 shows that the decline in District and national unit labor costs between 1972 and 1985 was also simi lar; unit labor costs (pavroll/MP) declined at a com pounded annual rate of 2.3 percent in the District, and 2.7 percent rate in the nation. Similar results are found when unit labor costs are measured by payroll/GVA or payroll/VS. Table 2 also shows the similarity o f both the level and growth o f labor productivity. Whether measured by MP/worker, GVA/worker, or VS/worker, the levels and com pounded annual growth rates o f District and LI.S. labor productivity w ere quite similar. The overall resemblance in the levels and growth of these operating ratios suggest that District manufac turing is maintaining its competitive position relative to the rest o f the nation.1- This, and the fact that the competitiveness o f the nation's manufacturing sector has improved relative to its major foreign competitors, suggests that District manufacturers are maintaining their competitive position in international markets as well as in domestic ones.0 ,2ln addition to similar composition and operating ratios, District manufacturing also resembled U.S. manufacturing in the relative importance of export industries, a factor that could influence manu facturing growth. The U.S. Census Bureau’s Annual Survey of Manu factures (Origin of Exports of Manufactured Products, 1987) reported that, in 1984, exports accounted for 5.8 percent of District manufac turing's shipments, compared with 6.7 percent nationally. 13See Tatom (1986a), pp. 14-15. http://fraser.stlouisfed.org/ 12 Federal Reserve Bank of St. Louis Uneven Growth and Structural Change The declining growth o f some mature industries, especially metal production, is sometimes cited as an example o f the decline o f manufacturing. As table 1 shows, however, the growth o f primary metal produc tion is not typical o f manufacturing as a whole. While the District's total MP expanded at a 2.6 percent pace in the 1973—85 period, the average annual growth rate o f regional prim aiy metals output was zero. Nation ally, total MP grew at a 2.9 percent rate while primaiy metals output fell at a 1.7 percent rate. Because the sector produced less than 10 percent o f regional or national MP between 1973 and 1985, however, its slug gish performance was offset by the more rapid growth in other manufacturing industiy groups. For example, MP o f the nonelectrical m achineiy and electronic equipment sectors grew at 8.6 and 3.9 percent rates in the District and at 7 and 6.6 percent rates nationally. These examples and the data in table 1 point out the uneven growth among manufacturing's industiy groups. Despite this diversity among the industries' growth rates, the uneven growth led to only minor changes in the industrial composition o f manufactur ing between 1972 and 1985. Chart 5 shows the propor tion o f total District MP contributed by each of the 10 largest industiy groups. Although there were some changes in the components o f manufacturing — for example, the rapid growth o f electronic equipment output caused that industiy's share to increase, while FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 C h a rt 5 Composition of District M anufacturing Product, 1 9 7 2 -8 5 P e rc en t of total 80 P rim ary Metals P a p e r and Allied Products R u b b e r an d Plastics Printing and Publishing Fabricated M etals Chemicals Electrical E q u ip m e n t Food and Kindred Transportation Products Equipment Nonelectrical M achinery 1972 74 76 the sluggish expansion o f prim aiv metals output caused its share to shrink — overall, the composition o f District manufacturing throughout this period re mained relatively constant. 78 80 82 1984 SUMMARY In both the nation and the Eighth District, em ploy ment growth in the manufacturing sector has not kept 13 FEDERAL RESERVE BANK OF ST. LOUIS pace with the rest o f the econom y’s employment growth, leading some observers to view manufactur ing as an ailing industry. Output trends, however, provide a different picture of manufacturing perfor mance. Nationally, real manufacturing output has grown as fast as the other sectors o f the economy. Labor productivity in manufacturing has grown faster than in the rest o f the economy, allowing manufactur ing to produce a constant proportion o f national out put with a declining proportion o f its labor force. Not all regions shared in the nation’s manufacturing stability. Rapid growth in the South and West offset declines in northern industrial areas. In the Eighth District, however, the growth o f manufacturing em ployment and output w ere quite similar to the na tional expansion in the 1972-85 period. This parallel growth was made possible by similarities in com posi tion, labor productivity and unit labor costs. Although some individual manufacturing indus tries contracted sharply since the early 1970s in terms of real output, others grew briskly as the composition of manufacturing evolved in response to consumer demands and comparative advantage. The overall trends point to the stability and increased productiv ity o f the Eighth District and U.S. manufacturing sec tors. 14 NOVEMBER 1987 REFERENCES Crandall, Robert W. “The Transformation of U.S. Manufacturing,” Industrial Relations (Spring 1986), pp. 118-30. Kendrick, John W., and C. Milton Jaycox. “ The Concept and Esti mation of Gross State Product,” Southern Economic Journal (March 1965), pp. 153-68. Mandelbaum, Thomas B. “The Eighth District’s Economy: A Micro cosm of the Nation’s,” Business — An Eighth District Perspective (Summer 1987). Niemi, Albert W., Jr. “ Gross State Product and Productivity in the Southeast, 1950-80,” Growth and Change (April 1983), pp. 3-8. Ott, Mack. “ The Growing Share of Services in the U.S. Economy — Degeneration or Evolution?” this Review (June/July 1987), pp. 5 22. Tatom, John A. “ Domestic vs. International Explanations of Recent U.S. Manufacturing Developments,” this Review (April 1986a), pp. 5-18. ________ _ “ Why Has Manufacturing Employment Declined?” this Review (December 1986b), pp. 15-25. U.S. Bureau of the Census. Annual Survey of Manufactures (Geo graphic Area Statistics). (GPO, various years). ________ _ Annual Survey of Manufactures, (Origin of Exports of Manufactured Products). (GPO, 1987). ----------------Census of Manufactures (Geographic Area Statistics). (GPO, various years). Weber, Richard E. “ A Synthesis of Methods Proposed for Estimat ing Gross State Product,” Journal of Regional Science (March 1979), pp. 217-30. FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Appendix Computing District Manufacturing Product Manufacturing product (MP) data com puted by the U.S. Com m erce Department measures that portion o f the na tion's real GNP originating in manufacturing. No corres ponding measure is available at the state or regional level. While the value o f shipments and gross value added are related measures, the shaded insert explains h o w they differ from MP. T o com pute a measure o f District manufacturing output corresponding to national MP, the m ethodology developed by Kendrick and Javcox (1965) and m odified by Niem i (1983) and W eber (1979) was follow ed. District MP is an estimate o f the sum o f manufacturing output in the four states that dom inate the District econom y — Arkansas, Kentucky, M is souri and Tennessee. MP was derived by estimating output for each o f the District's 20 manufacturing industiy groups and sum m ing over all industry groups. (1) PMPiin= (MP„US/EI11,S)E11U w here M P is real GNP originating in the nation's m anufac turing industiy group i, year t, E represents earnings, and the US and D subscripts sym bolize the U.S. and the Eighth District, respectively. Earnings and U.S. M P data are pub lished by the U.S. Com m erce Department. Earnings include wages and salaries, other labor incom e and p roprietoiy income. The prelim inaiy estimates resulting from equation 1 will be accurate to the extent that the ratio o f M P to E in each industiy group is similar in the District and the nation. This assumption has been interpreted as one o f similar prod u c tivity at the regional and national levels. In the second step o f com puting District MP, the prelim inary estimates for each industiy group w ere adjusted by a measure o f that industry’s productivity in the District relative to the nation. This procedure was developed by Niem i (1983). The m ea sure o f relative productivity is the ratio o f gross value added p er dollar o f payroll for the District to gross value added per dollar o f payroll in the nation, or District MP was com puted in two steps. First, prelim inaiy estimates w ere calculated assuming that the ratio o f output to earnings in each manufacturing industiy was identical in the District and the United States. In the second step, the prelim inaiy estimates w ere adjusted to correct for produ c tivity differences betw een the District and the United States. w here GVA and P are gross value added and payroll data More specifically, the first step in estimating District MP is to m ultiply the ratio o f national output to national earn ings in each o f the industiy groups bv District earnings in that industiy. That is, the prelim inaiy estimate o f District output originating in industiy group i, year t is: published by the U.S. Bureau o f the Census’ Annual Survey o f Manufactures and the Census o f Manufactures. For each industry group, the relative productivity measure was m ul tiplied by the prelim inaiy estimates (PMP,,,,) to com pute the final estimates. Total manufacturing output is the sum of the final estimates for all industiy groups. (2) (GVAil,)/Pill))/(GVAlll1s/P,ll(s), 15 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 The Great Bull Markets 1934—29 and 1983—87: Speculative Bubbles or Economic Fundamentals? G. J. Santoni Every so often, it seems, humankind almost en masse has a com pulsion to speculate, and it yields to that com pulsion with abandon. — Robert T. Patterson, The Great Boom and Panic, p. xiii. M ANY people attribute the bull market o f 192429 and the subsequent collapse in stock prices to a “speculative bubble.” 1 According to this view, the crash was inevitable because it was only a matter of time until the bubble burst (see shaded insert on opposite page). The same theoiy of stock price formation is used to describe the bull market o f 1982-87. Recent discus sions have characterized this bull market as the prod uct o f "unexpected insanity,” subject to "trading fads and frenzies rather than econom ic fundamentals” and “out of control.”- Comparisons between the 1920s and 1980s like the one summarized in chart 1 have ap peared recently in the press.3 Chart 1, which plots quarterly data on the levels o f the Dow Jones Indus G. J. Santoni is a senior economist at the Federal Reserve Bank of St. Louis. Thomas A. Polimann provided research assistance. 'See the shaded insert on page trial Index over the two periods, shows that the behav ior o f stock prices in both periods is similar.4Both bull markets began in the second quarter o f the year; each lasted 21 quarters; each hit its peak in the third quarter with the timing o f the peaks separated by only a few days (September 3, 1929, and August 25,1987); in each case, 54 days elapsed between the peak and the crash; and each crash stripped slightly more than 20 percent from the stock market averages. The belief that speculative bubbles might cause a persistent deviation in stock prices from the price consistent with the fundamentals is important. At the time o f the 1929 crash, it spawned legislative proposals that w ould curb credit for speculation, amend the National Banking and Federal Reserve acts, impose an excise tax on stock sales and regulate the activities o f stock exchanges and investment trusts.' Furthermore, and Kindleberger (1978), p. 17. 2“ Abreast of the Market” (1987) and Jonas and Farrell (1986). 3See, for example, Koepp (1987), Powell (1987), Schwartz and Tsiantar (1987) and Wall Street Journal (1987). 16 4Scale (1982-87) = 8 x scale (1924-29). r'New York Times (October 25, 1929). NOVEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS Some Popular Notions Regarding the Cause of the 1929 Crash “Gambling in stock has become a national disease . . . Neither assets nor earnings, large as the earnings have been in many instances, warrant the market values of hundreds of stock issues. There has been an inflation (in stock prices) not free from the charge of criminality,... It was inevitable that a day of reckoning would come and the billions would be lost as the water and hot air were eliminated from hundreds of stock issues.” Senator King, New York Times (October 25, 1929). “The bull market was created by phenomenal profits in a few leading shares. Even in these shares there were not sufficient profits to justify the prices which prevailed before October 1928." Niebuhr (1930), p. 25. “This growth (in stock prices) was matched by wide spread, intense optimism which in the end deteriorated into lack of perspective and discipline. This optimism went so far in places that people began to believe that there was such a thing as 'permanent prosperity’ and that economic crises could be eliminated.” Roepke (1936). "As already so often emphasized, the collapse in the stock market in the autumn of 1929 was implicit in the speculation that went before.” Galbraith (1955), p. 174. “The most common explanation of the Crash to this day is that the market was overpriced because of specu if stock price bubbles exist, econom ic policymakers face a difficult problem because bubbles suggest that plans to save and invest may be based on irrational criteria and subject to erratic change.'1 The purpose o f this paper is to compare the im pli cations of a theoiy o f stock prices based on fundamen tals to one that allows for bubbles, then to examine 6Keynes (1935), p. 159. Keynes discussed erratic shifts in the invest ment schedule caused by changes in the "state of confidence” (pp. 153-55) and ' speculation” (p. 161). He argues that a “ . . . boom which is destined to end in a slump is caused, therefore, by the combination of a rate of interest, which in a correct state of expectation would be too high for full employment, with a misguided state of expectation which, so long as it lasts, prevents this rate of interest from being in fact deterrent. A boom is a situation in which over-optimism triumphs over a rate of interest which, in a cooler light, would be seen to be excessive” (p. 322). See, as well, Gordon (1952), p. 378 and Varian (1979). lation . . .” Wanniski (1978), p. 125. "In the end, fright may have been what turned retreat into rout. And that fright may have been partly motivated by the perception of absurdly high stock prices . . Schumpeter (1939), p. 876. “Among the immediate or precipitating causes (of the crash) were the unjustifiably high prices of common stocks . . Patterson (1965), p. 215. "The breakdown of 1929 was as nearly the result of willful mismanagement and violation of every principle of sound finance as such occurrence has ever been. It was the outcome of vulgar grasping for gain at the ex pense of the community.” Willis (1930). “It may be legitimately said that the boom and slump were caused by the alternate domination of greed and fear, and that the one was bound to resign sooner or later in favor of the other,. . Hodson (1933). “Never a boom and high prosperity without an out break of speculation. Never such an outbreak that has not ended in a financial crisis.” Snyder 11940). “Might one still suppose that this kind of stock market crash (in 1929) was a rational mistake, a forecast error that rational people might make? This paper . . . implies that the answer is no.” Shiller (1981), p. 422. evidence from the 1920s and the 1980s to determine which set o f implications is supported by the data. The behavior o f stock prices during these two periods is particularly useful in testing asset prices for the presence o f speculative bubbles. The 1924—29 experi ence is one o f the most significant bull markets in U.S. history in both its duration and rate o f advance. Though not quite as dramatic, the behavior o f stock prices in the 1980s has been similar. If stock price bubbles exist, these are likely places to look for them. THE FUNDAMENTALS OF STOCK PRICES People value comm on stocks for their expected return. Since investors may choose among broad cate gories o f stock, the expected return on any particular stock must be equal to the expected return on other 17 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Chart 1 The Bull M arketsof the 1920s and 1980s11 Dow Jones Industrial Index (Nominal Values) 1982 1 2 3 1983 4 1 2 3 1984 4 1 2 3 1985 4 1 2 3 1986 4 1 2 3 1987 4 1 2 3 4 | i i i i i i i i i i i i i i i i i i i i i i r~| D o w Industrials 1 9 2 4 - 2 9 scale D o w Industrials 1982-87 scaled |_1_ Sources: Moore (1961), pp. 109 ,145 and Economic Report of the President. Various years. [ 2 Scale 1982-87 = 8x Scale 1924-29. http://fraser.stlouisfed.org/ 18 Federal Reserve Bank of St. Louis NOVEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS stocks o f similar risk. For example, if a particular stock is expected to yield a relatively lo w return, investors w ill shun it causing its price to fall. This raises its expected return. The reverse holds for any stock with an expected return that is higher than other stocks o f similar risk. An equilibrium exists when the expected returns are equal across equally risky stocks. Econo mists call this equilibrium return the required dis count rate. Equation 1 calculates the expected return from holding a stock for one year assuming dividends are paid at year-end.7 (1) Expected Rate o f Return = Forecast o f price at year end + Forecast of dividend —Current Price Current Price Equation 2 solves equation 1 for the current price by noting that the expected return is equal to the re quired discount rate in equilibrium. (2) Current Price = Forecast o f price at year-end + Forecast o f dividend (1 + Required discount rate) The Price Depends on Forecasts o f Future Outcomes The important thing to note in equation 2 is that the current price depends on forecasts o f fu tu re out comes which, o f course, are subject to change as new information becomes available. The price does not depend on dividends that are observed in the present as Senator King and others have im plied in their comments on the behavior o f stock prices during the 1920s (see shaded insert on page 17). The current price may change even though observed dividends do not and conversely. How Far Ahead? The discussion so far indicates that investors must forecast the price o f the stock next period. What are the fundamentals for this future price? In princi ple, the future price depends on the stream o f divi dends and the required discount rate investors expect to prevail over the life o f the firm. Typically this re quires forecasts that extend into the distant future and suggests that the job o f analyzing stock prices is formi- 7See Brealey (1983), pp. 67-72, and Brealey and Meyers (1984), pp. 43-58. dable. It is sometimes possible to simplify the calcula tion, however. If dividends are expected to grow at a constant annual rate and the discount rate is con stant, the calculation can be simplified as shown in equation 3.8 (3) Current Price = Forecast o f dividend Required discount rate - Expected growth rate o f dividends The Price Fundamentals Restating the solution for the current price as in equation 3 is particularly useful for the purposes o f this paper. Equation 3 is a list o f the price fundamen tals: the forecast o f the dividend next period, the required discount rate, and the expected (forecast) growth rate o f dividends. The solution for the current price in equation 3 is called the fundamentals price. Furthermore, the equation can be used to show how relatively small changes in forecasts can account for relatively large changes in the fundamentals price. For example, suppose investors forecast a year-end divi dend o f $.60 per share, an annual dividend growth rate o f 6 percent and the required discount rate is 8 per cent. Equation 3 indicates that the fundamentals price is $30 per share [ = .6/( 08 —.06)]. N ow suppose that new information leads investors to low er the forecast o f dividend growth to 5 percent. This is a decline of about 17 percent in expected growth [ = (.01/.06)100]. The fundamentals price, however, declines to $20 [ = .6/(.08 — .05)], or more than 30 percent. Notice that a large decline in price may occur even though observed dividend payments do not change. It is even possible for the price to decline when observed dividends rise. STOCK PRICES AND MEASURES OF THE FUNDAMENTALS Table 1 shows annual average growth rates o f the D ow Jones Industrial Index in each year during the two bull markets.9 The index rose rapidly during the “Brealey (1983), p. 69. The current price is defined by equation 3 only if the expected growth rate in dividends is less than the required discount rate. 9The data on stock prices used in this paper are daily closing levels of the Dow Jones Industrial Index. Daily closing levels of this index are available on a consistent basis from January 1915. See Pierce (1982). When possible, the statistical results obtained with this data were checked against results using daily closing levels of the Standard and Poor’s Composite Index. In no case were any qualita tive differences observed. 19 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Table 1 Growth Rates in Stock Prices (annual average growth rates)1 Panel A: 11/1924-111/1929 Period Dow Industrials 11/1924 —IV/1924 IV /1 92 4- IV/1925 IV /1 92 5- IV/1926 IV /1 92 6- IV/1927 IV/1927 - IV/1928 IV /1 92 8- 111/1929 32.8% 34.6 1.5 21.5 32.7 37.7 Average 11/1924 - III/19292 25.7% Panel B: 11/1982-111/1987 Period Dow Industrials 11/1982 - IV/1982 IV/1982 - IV/1983 IV/1983 - IV/1984 IV /1 98 4- IV/1985 IV/1985 - IV/1986 IV/1986 —111/1987 40.1% 20.9 -4 .4 17.8 26.8 31.5 Average 11/1982-III/1 9 8 7 2 20.0% ’Computed from quarterly averages of seasonally adjusted data. See Moore (1961), pp. 106-09. 2ln computing this average, the growth rates for each period are weighted by the length of the period. initial phases o f the bull markets, slowed down con siderably in 1926 and 1984, then rose rapidly through the third quarters o f 1929 and 1987. A rapid advance in stock prices is not surprising if it results from changes in the fundamentals. The investi gator, however, seldom has the luxury o f direct obser vation o f the fundamentals. Instead, other variables (proxies) that are believed to provide information about the behavior o f the unobserved fundamentals must be used. For example, credit market interest rates and actual dividend payments have been used to proxy the required discount rate and the expected stream o f future dividends. It is important to recognize that, at best, the behavior o f these (or other) proxies may give only a rough approximation o f the behavior o f the fundamentals and, on occasion, they may be entirely misleading. The 1920s may be an example of the latter case. 20 Long-term rates w ere roughly constant from 1924— 29.'“ Data on actual per share dividends are very sketchy for this period. One estimate, however, indi cates that actual dividends increased at an annual rate o f about 8.8 percent from 1924—29." W hile this is a fairly rapid rate o f increase, it is far less than the growth observed in stock values. (See shaded insert on opposite page for a more precise estimate o f the rela tionship between stock prices and these proxy vari ables.) When the market crashed, people like Senator King pointed to these proxy variables and claimed that stock prices before October 1929 contained “water and hot air.” An alternative explanation is that the proxies give a misleading impression o f the behavior o f the fundamentals. FUNDAMENTALS, FOOLS AND BUBBLES In order to evaluate the notion that stock prices in the 1920s and 1980s w ere driven by psychological factors extraneous to the fundamentals, it is necessary to be clearer about the implications the alternative hypotheses have for variables that can be observed by the investigator. This paper considers three different theories that potentially explain stock prices: the ef ficient market hypothesis, the greater fool theory and the theory o f rational bubbles. Efficient Markets and Fundamentals A long-standing proposition in both economics and finance is that stock prices are form ed in efficient markets.'2This means that all o f the relevant informa tion currently known about interest rates, dividends and the future prospects for firms is contained in current stock prices. Stock prices change only when new information regarding the fundamentals is ob tained bv someone. N ew information, by definition, cannot be predicted ahead o f its arrival; because the news is just as likely to be good as it is to be bad, jumps in stock prices cannot be predicted in advance. Many present-day stock market analysts are skepti cal o f the efficient markets hypothesis.13 Similarly, 10See Friedman and Schwartz (1982), table 4.8, and Homer (1977), p. 352. "S ee Cowles (1938), p. 389. 12See Brealey and Meyers (1984), pp. 266-81; Malkiel (1981), pp. 171-79; Brealey (1983), pp. 15-18; Leroy (1982) and Fama(1970). 13See Malkiel (1981), pp. 126-79. FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 The Relationship Between Growth in Stock Prices, Dividends Per Share and the Interest Hate: 1872-1930 The following regression estimate relates first differ ences in the natural log of the Cowles Commission index of stock prices, ALnP, to first differences in the natural logs of the Cowles Commission estimate of per share dividend payments, ALnD, and the interest rate on long term bonds, ALnR. The data are annual and span the period 1872-1930. The regression estimate is intended to illustrate the results that are obtained when observed values of dividends and credit market interest rates are used to proxy expected dividends and the required discount rate. ALnP = .16 + 49ALnD-1.26ALnR. (.11) (4.54) (4.07) Rho RSQ SE = .03 (.23) = .39 = 10.40 The estimated coefficients of these proxy variables are significantly different from zero and the qualitative rela tionship between stock prices and these proxies is the same as that expected for their theoretical counterparts. traders in the 1920s generally did not subscribe to it (see shaded insert on following page). But that is not important. If the behavior o f stock prices is consistent with the implications o f the theoiy, the hypothesis helps both to understand bow stock markets work and to evaluate the claim that the bull markets were products o f price bubbles. If the efficient markets hypothesis is correct, past price changes contain no useful information about future price changes. With some added assumptions, this can be translated into useful empirical proposi tions. If the expected return to holding stock is con stant and the volatility o f stock prices does not change during the time period examined, the efficient market hypothesis implies that observed changes in stock prices should be uncorrelated and that price changes should not exhibit long sequences o f successive changes that are greater or less than the median change for the sample. There is a considerable amount of "noise” in the esti mate, however, in the sense that variation in the proxy variables explains a relatively small amount (about 40 percent) of the variation in stock prices. More importantly, the estimated equation performs veiy poorly in 1929 and 1930. For example, the percent age change in stock prices predicted by the regression estimate for 1929 is -1.24 percent. Stock prices actually rose in 1929 bv 23.86 percent. The deviation of the actual from the predicted value is 25.10 percent. This deviation exceeds two standard errors of the estimate, indicating that such a large deviation is not likely to result from chance. In short, it suggests that the large increase in stock prices in 1929 was unrelated to movements in the proxy variables. In the case of 1930, the actual decline in stock prices exceeds the predicted decline by more than two standard errors. This pattern — a significantly larger percentage increase in stock prices than predicted for 1929 and a significantly larger decrease in stock prices than predicted for 1930 — appears to be consistent with the notion that a speculative bubble was responsible for a boom in prices and a crash when the bubble burst. The above propositions should hold even if the level o f stock prices appears to drift upward or downward. These propositions concern the relationship between the sequence of price changes, not the average change over some specific period. Clearly, stock prices drifted upward during both bull markets; but that does not necessarily mean that price changes w ere correlated or that there were long runs o f positive changes that exceeded the median change for these periods. Put differently, it does not necessarily mean that market participants w ere able to predict future changes in stock prices by observing the past. Greater Fools The notion that self-feeding speculative bubbles, on occasion, can drive stock prices is known as the “greater fool theoiy.” According to this theoiy, people regard the fundamentals as irrelevant. Rather, they buy stock on the belief that some (bigger?) fool will buy 21 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 What Some Big Plungers Thought of Efficient Markets1 William C. Durant Arthur W. Cutten Durant had been acquiring a large interest in Ameri can Smelters and its share price had risen from $119 to $140. One day during this period a friend burst into his office and exclaimed, “Now look here, Billy, what are you doing with Smelters? You know it’s not worth $140." "Possibly not,” Durant said, "but take my advice and don’t sell me any more of it, because it’s going much higher.” The stock went to $390 on a split share basis. "Yes, I have taken my bit out of the market. Oh, quite a bit. But I would advise other men to stay away from it. If I had a son I wouldn’t let him touch it with a ten-foot pole. Jesse Livermore "A gambler is a man who doesn't know the market. He goes to a broker and says, 'What can I do to make a thousand dollars?’ He is only an incident. The specula tive investor buys or sells against future conditions on his knowledge of what has happened in the past under a similar set of conditions." Louis W. Zimmerman Zimmerman employed a team of experts to study the market constantly. He never purchased a stock until he received a final report from the analysts concerning the condition of the company. the shares from them at a higher price in the future. People maintain this belief because they think “that market values will rise — as they did yesterday or last week — and a profit can be made.” '4Once the specula tion begins, stock prices continue rising because peo ple, seeing the rise in the previous period, demand additional shares in the belief that prices w ill continue to rise. This pushes prices still higher. The greater fool theory is based on the presumption that there are times when past movements in stock prices matter. According to this theoiy, during the “fooling” periods, there should be positive correlation in the past sequence of price changes and long runs o f positive changes that exceed the median change for the sample period. There are too many wrecks down there in the pit. People call them brokers. They are only part of that — the broke part.” * * * The efficient markets hypothesis (EMH) suggests that Durant was lucky. He could not have known that the price of American Smelters would rise. Livermore's eval uation of the "gambler’s” strategy vs. the "investor's” contradicts the implication of EMH that the strategy of each is just as likely to succeed (or fail). Similarly, EMH suggests that hiring teams of experts, as Zimmerman did, is not expected to result in raising the return from stock purchases above a normal return. This applies to Cutten's comment regarding brokers who, according to EMH, are expected to earn a normal return on their stock trades, not a negative return as suggested by Cutten. ’ See Sparling (1930), various pages. Rational Bubbles Recently, some economists have discussed the possibility that stock prices may contain "rational” bubbles.15The theoiy o f rational price bubbles is based on the belief that some asset prices (for example, stock, gold and foreign currency prices) are too vari able to be justified by variation in the fundamentals.16 (A more formal theoiy of price bubbles is summarized in the appendix to this paper.) Briefly, the th eoiy says that there may be occasions w hen stock prices deviate from the price that is consistent with the fundamen tals. The deviation is called a bubble. ,5See Flood and Garber (1980 and 1982), Blanchard and Watson (1982), West (1986), Diba and Grossman (1985 and 1986) and the appendix to the paper. '"Galbraith (1955), p. 23. See, as well, Malkiel (1981), pp. 31-49. http://fraser.stlouisfed.org/ 22 Federal Reserve Bank of St. Louis ,6See, for example, Shiller (1981) and Mankiw, Romer and Shapiro (1985). FEDERAL RESERVE BANK OF ST. LOUIS Bubbles must possess certain characteristics if they are to have econom ic significance: Bubbles must be persistent so that a forecast o f stock prices based solely on the fundamentals is biased. This means that forecast errors (actual price minus forecast price) w ill tend to have the same sign and not average out. The persistence o f one-sided errors is important because random variation in the data gen erally w ill cause the actual price to differ from any well-constructed forecast o f the price even though a bubble is not present. If bubbles w ere only a name used to describe random variation in the data, they w ould not be very interesting. Bubbles must be explosive in the sense that they must grow at a rate that compensates the stock pur chaser for the additional amount invested in the stock due to the bubble. In addition, there may be a risk premium to compensate stockholders for the addi tional risk that the bubble may burst.17This character istic causes the price to deviate further and further from the fundamentals for as long as the bubble lasts. Bubbles can not be negative. A negative bubble means that stock prices are less than im plied by the fundamentals. The explosive characteristic o f bubbles means that the prices im plode with some chance that stock prices w ill be negative at some future date.18 Negative stock prices, however, are impossible; they are inconsistent with the liability rules associated with com m on stock which limit potential losses to the extent o f the initial investment. RATIONAL BUBBLES AND STOCK PRICE BEHAVIOR The theory o f rational bubbles has implications for the behavior o f stock prices that are different than the theory o f efficient markets.19This is shown in table 2, which makes use o f the fundamentals theory o f stock price determination discussed above. One im portant assumption o f this example is that, at each m oment in time, investors expect dividends to grow at a constant rate over the future. To keep things simple, the example assumes that subsequent events conform ,7See Diba and Grossman (1985 and 1986), Blanchard and Watson (1982), Flood and Garber (1980), West (1986) and the appendix to this paper. ,8See Diba and Grossman (1985 and 1986) and Blanchard and Watson (1982). 19See Diba and Grossman (1985) and the appendix. NOVEMBER 1987 to the expectations o f investors (perfect foresight, an extreme version o f rational expectations) and that the dividend is initially expected to be $2. The expected dividend is constant in panel A (expected growth rate is zero) but grows in panel B at an expected annual rate o f 2 percent. The required discount rate is 10 percent, and a bubble o f $1 occurs in period zero. Column 3 o f panel A computes the fundamentals price, P[. This is simply the expected dividend, E,(D, t,) = $2, (assumed constant in panel A) divided by the difference between the required discount rate, r = .10, and the expected growth rate in dividends, g = 0. The fundamentals price is $20 each period. The fourth colum n computes the bubble com p o nent o f the price. As discussed above, the bubble expands over time at the required discount rate, r. The observed price, P,, is the sum o f the fundamentals price and the bubble as in column 5. Column 6 calculates the percentage changes in the price. These are positive. More importantly, the num bers in column 6 rise over time indicating that this bubble produces a time series o f observed price changes that are positively correlated. The observed price does not follow a random walk. O f course, the real w orld is never so neat. Changes in the fundamen tals — r,g, E,(Dt+1) — may cause the observed price to change in a w ay that masks the bubble. If that occurs, however, it is not clear that the bubble is very im por tant since an investor’s behavior under the theory of rational bubbles depends on his ability to detect the presence o f bubbles. The example in panel B is similar to the example in panel A except that dividends are expected to grow at a 2 percent annual rate. Notice that this does not change the qualitative result with respect to the ob served price changes. These rise over time and w ill be positively correlated. The only difference between the two examples is that the fundamentals price in panel B rises (drifts upward) over time at a constant 2 per cent rate (see column 7). This results from the growth in dividends. W hile the fundamentals price drifts up ward at a constant rate o f 2 percent, the sequence o f changes in the fundamentals price are uncorrelated. The fundamentals price w ill follow a random walk with drift. An important thing to note is that both the greater fool theory and the theory o f price bubbles discussed in this paper im ply that stock prices behave similarly. Both reject the efficient markets hypothesis, which implies that stock prices follow a random walk. 23 NOVEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS Table 2 Fundamentals vs. Bubbles: An Example Panel A: Expected growth of dividends is zero Years Ei(Dt ,,) P! Bt=(1 +r)'B0 P ,= P |+ B t 0 1 2 3 4 5 $2.00 2.00 2.00 2.00 2.00 2.00 $20.00 20.00 20.00 20.00 20.00 20.00 $1.00 1.10 1.21 1.33 1.46 1.61 $21.00 21.10 21.21 21.33 21.46 21.61 .48% .52 .57 .61 .70 0.0% 0.0 0.0 0.0 0.0 %AP, %APi Panel B: Expected growth of dividends is 2 percent Years E,(Dti1) P| B, = (1 +r)'B0 Pt = P |+ B , %APt %APJ 0 1 2 3 4 5 $2.00 2.04 2.08 2.12 2.17 2.21 $25.00 25.50 26.01 26.53 27.06 27.60 $1.00 1.10 1.21 1.33 1.46 1.61 $26.00 26.60 27.22 27.86 28.52 29.21 2.31% 2.33 2.35 2.37 2.42 2.00% 2.00 2.00 2.00 2.00 Where: E, (D,.,) = the expected dividend next period PJ B, r P, g = = = = = the the the the the fundamentals price in period t bubble in period t and B0 is the initial bubble required discount rate observed price in period t expected growth rate in dividends P, = P! + B, = E'<D- > + B, r-9 SOME PROBLEM S WITH BUBBLES The notion that stock prices are influenced by bubbles is troublesome because it is not based on a well-specified theory. A com plete theory o f bubbles should identify the cause o f bubbles in terms o f some phenom enon that can be observed separately from bubbles themselves. On those occasions when the cause is observed, a bubble should also be observed and conversely. This allows a direct test o f the theory and explains w hy bubbles may be observed on some occasions but not others. In contrast, the greater fool and rational bubble theories do not suggest a cause o f bubbles that can be observed separately. Rather, unusual price behavior (the bubble) is attributed to “ intense optimism,” “a compulsion to speculate” and "manias.” These do not http://fraser.stlouisfed.org/ 24 Federal Reserve Bank of St. Louis identify the cause o f the bubble; they merely give the bubble a different name.211 These criticisms suggest that attributing crashes in stock prices to bursting bubbles adds nothing to our understanding o f w hy crashes occur or how to pre vent similar occurrences in the future. To illustrate, “ Brunner and Meltzer (1987) note that Some further reflections on bubbles and sunspots equilibria should make us doubt their contribution to a useful reconciliation of analysis with critical observations. The bubble term refers neither directly nor indirectly to any observable entities. It is fundamentally inconsistent with any rational exploitation of infor mation invoked by the same analysis (p. 2). See, as well, Singleton (1987), pp. 28-30. Slrkin (1975) and Sch wartz (1981), p. 25, question the bubble hypothesis as an explana tion of the 1929 crash. NOVEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS Wesley Clair Mitchell (a noted student o f business cycles) wrote that By a com bination o f various agencies such as public regulation o f the prospectuses o f n ew companies, leg islation supported by efficient administration against fraudulent prom otion, m ore rigid requirements on the part o f stock exchanges concerning the securities ad m itted to official lists, m ore efficient agencies forgivin g investors information, and m ore conservative policy on the part o f the banks toward speculative booms, w e have learned to avoid certain o f the rashest errors com m itted bv earlier generations.21 Mitchell made this statement in 1913 in reference to the legislative and regulatory precautions instituted after the Panic o f 1907. Like the crash in 1929, the 1907 crash had been attributed to a speculative bubble. EFFICIENT MARKETS VS. PRICE BUBBLES: SOME EVIDENCE The efficient markets hypothesis suggests that stock prices follow a random walk. The hypothesis has no implication for the drift in stock prices. Prices may be higher or low er at the end o f the period being examined. Neither o f these events is necessarily in consistent with the hypothesis. Rather, the hypothesis implies that the sequence o f price changes are unre lated; they behave as random variables. In contrast, the greater fool theory and the theory o f rational bubbles discussed here imply that changes in stock prices are not random but are positively related. Which explanation is better supported by the evi dence for the 1924—29 and 1982-87 bull markets? To evaluate these theories, data on the level o f the D ow Jones Industrial Index are used. Tw o periods are examined. One extends from January 3, 1928, through September 3, 1929. The second runs from January 2, 1986, through August 25, 1987. The data are first differ ences o f the log of the D ow ’s daily closing level multi plied by 100 and are approximately equal to the daily percentage change in the index. Each sample contains more than 400 observations. Stock prices advanced very rapidly in these periods. If bubbles were present, they should be apparent in these data. Were Stock Prices A Random Walk? Table 3 presents the results o f a test (called a BoxPierce test) based on the estimated autocorrelations of percentage changes in the Dow Jones Industrial In- 21Mitchell (1950), p. 172. Table 3 Autocorrelation Coefficients and Box-Pierce Statistics (first differences of logs of Dow Industrial Index) January 3,1928 - September 3,1929 To lag Autocorrelation coefficient Box-Pierce statistic 1 2 3 6 12 18 24 .0196 - .0325 - .0494 .0200 .0069 - .0521 .0213 .18 .70 1.91 10.41 16.43 21.65 29.58 Mean of series = .128* = 2.57 t-score January 2,1986 - August 25,1987 To lag Autocorrelation coefficient Box-Pierce statistic 1 2 3 6 12 18 24 .0553 -.0 1 4 0 - .0095 -.0151 - .0076 - .0044 .0024 1.28 1.36 1.40 4.66 7.86 13.14 14.24 Mean of series = .136* t-score = 2.83 'Statistically significant at the 5 percent level dex. This test is designed to determine w hether there is significant autocorrelation in the data, that is, whether current changes in the index are related to past changes. Recall that the efficient markets hypoth esis implies that past changes in stock prices are unrelated to (contain no information about) current or future changes. An empirical counterpart of this prop osition is that changes in the index are not correlated. Conversely, if the hypothesis that stock prices were influenced by self-feeding bubbles is correct, percent age changes in the index should be positively correl ated. Table 3 shows test results for the two periods dis cussed above. None of the test statistics indicate signi- 25 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 C hart 2 An Illustration of a Random Sequence Vs. Correlated O bservations11 Panel A : R a n d o m sequence http://fraser.stlouisfed.org/ 26 Federal Reserve Bank of St. Louis FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Table 4 Runs Test Sample period Number of observations Observed number of runs Expected number of runs Variance Jan. 3 ,1 9 2 8 Sept. 3,1929 495 233 248.0 123.50 Jan. 2, 1 9 8 6 A u g .2 5 ,1987 417 220 209.0 104.00 Expected number of runs = (Number of observations + 1)/2 Variance = (Number of observations - 1 )/4 ficant correlation at conventional confidence levels.22 Stock prices follow ed a random walk, which is consist ent with the efficient markets hypothesis. Table 3 also shows the mean change for each p e riod. The means are positive and significantly different from zero in a statistical sense. Today, the upward drift in stock prices during these time periods is obvi ous. At that time, however, the upward drift is not something that investors could have bet on with any confidence. Runs Test A run is the number o f sequential observations that are greater or less than the sample median (the m iddle value o f the sample). If a series o f observations exhibits too few runs relative to what is expected for independent observations, the data are positively cor related or drawn from different populations. The efficient markets hypothesis suggests that ob served changes in stock prices are uncorrelated, that is, the changes are independent o f one another. This means, for example, that there is no tendency for a large positive change to be follow ed by another large positive change. Consequently, the sequence o f ob served changes w ill move back and forth across the median change for the sample fairly frequently as shown in panel A o f chart 2. If changes in stock prices 22Daily data between October 22, 1929, and March 31, 1930, show significant autocorrelation at various lags. This is likely a statistical artifact produced by a substantial increase in the variance of the data at the time of the crash in October and November that appears to taper off over time. Consequently, the significant correlations do not suggest the presence of a bubble. Furthermore, stock prices were declining at this time and bubbles can not be negative. are correlated as im plied by the bubble hypothesis, however, a plot o f the observations in the order that they appear w ill indicate some tracking as shown in panel B. This plot crosses the sample median infre quently. The example exhibits relatively long and, con sequently, few er runs than expected o f independent observations.23 Table 4 presents the results of a runs test for the bull markets o f the 1920s and 1980s. The third column of the table shows the number o f runs observed for daily percentage changes in the D ow Jones Industrial Index during each period o f rapidly increasing stock prices. Column 4 gives the number o f runs expected for a series o f (495 and 417) independent observations and column 5 gives the variance o f this series. Since the observed number o f runs is not much different than expected, the hypothesis that percentage changes in the D ow Index behaved randomly during the sample periods is not rejected by this data. The evidence on the behavior o f stock prices (as characterized by the D ow Index) is not consistent with the notion that stock prices w ere driven by self feeding speculative bubbles during the 1920s and 1980s. CONCLUSION Many people attribute the stock market crashes of 1929 and 1987 to bursting speculative bubbles. The perception that stock prices may be driven by bubbles presents econom ic policymakers with an important problem because such bubbles suggest that plans to ?3See Wonnacott and Wonnacott (1977), pp. 486-88. 27 FEDERAL RESERVE BANK OF ST. LOUIS save and invest may be based on irrational criteria and subject to erratic behavior. This paper has examined data on stock prices around the time o f the Coolidge and Reagan bull markets. The paper provides evidence contrary to the notion that the crashes w ere the result o f bursting speculative bubbles. No evidence was found that changes in stock prices w ere autocorrelated or that the data contained long runs. Rather, the data suggest that stock prices follow ed a random walk. This evi dence is consistent w ith the efficient markets hypoth esis, which is based on the proposition that all rele vant and ascertainable information regarding stock price fundamentals (interest rates, dividends, future prospects, etc.) is contained in current stock prices. REFERENCES “ Abreast of the Market." Wall Street Journal, January 26,1987. NOVEMBER 1987 Hodson, H. V. Economics of a Changing World (Harrison Smith and Robert Haas, 1933), p. 164. Homer, Sidney. Press, 1977). A History of Interest Rates (Rutgers University Jonas, Norman, and Christopher Farrell. “ Program Trading: Let the Little Guy In,” Business Week (September 29,1986), p. 100. Keynes, John Maynard. The General Theory of Employment, Interest and Money (Harcourt, Brace and Company, 1935). Kindleberger, Charles P. Books, 1978). Koepp, Stephen. pp. 44-46. Manias, Panics and Crashes (Basic “ How Ripe for a Crash?” Time (October 5,1987), Leroy, Stephen F. “ Expectations Models of Asset Prices: A Survey of Theory,” Journal of Finance (March 1982), pp. 185-217. Malkiel, Burton G. A Random Walk Down Wall Street (W. W. Norton and Company, 1981). Mankiw, N. Gregory, David Romer, and Matthew D. Shapiro. “An Unbiased Reexamination of Stock Market Volatility," Journal of Finance (July 1985), pp. 677-87. Mitchell, Wesley Clair. Business Cycles and Their Causes (Univer sity of California Press, 1950). Blanchard, Oliver J., and Mark W. Watson. “ Bubbles, Rational Expectations, and Financial Markets,” in Paul Wachtel, ed., Crises in the Economic and Financial Structure (Lexington Books, 1982), pp. 295-315. New York Times Brealey, R. A. An Introduction to Risk and Return from Common Stocks (The MIT Press, 1983). Niebuhr, Reinhold. "The Speculation Mania," The World of Tomor row (January 1930), pp. 25-27. Brealey, Richard, and Stewart Meyers. nance (McGraw-Hill, 1984). Patterson, Robert T. Company, 1965). Principles of Corporate Fi Brunner, Karl. “ Epilogue: Understanding the Great Depression,” in Karl Brunner, ed., The Great Depression Revisited (Martinus Nijhoff, 1981), pp. 316-58. Brunner, Karl, and Allan H. Meltzer. “ Bubbles and Other Essays,” Carnegie-Rochester Conference Series on Public Policy (Spring 1987), pp. 1-8. Cowles, Alfred III, and Associates. 1937 (Principia Press, 1938). Common-Stock Indexes, 1871- Diba, Behzad T., and Herschel I. Grossman. “ Rational Bubbles in Stock Prices?” (National Bureau of Economic Research, Working Paper 1779, 1985). _________“ On the Inception of Rational Bubbles in Stock Prices,” (National Bureau of Economic Research, Working Paper 1990, 1986). Fama, Eugene F. “ Efficient Capital Markets: A Review of Theory and Empirical Work,” Journal of Finance Papers and Proceedings (May 1970), pp. 383-417. Flood, Robert P., and Peter M. Garber. “ Bubbles, Runs, and Gold Monetization,” in Paul Wachtel, ed., Crises in the Economic and Financial Structure (Lexington Books, 1982), pp. 275-93. _________“ Market Fundamentals versus Price-Level Bubbles: The First Tests,” Journal of Political Economy (August 1980), pp. 745-70. Friedman, Milton, and Anna J. Schwartz. Monetary Trends in the United States and United Kingdom, 1867-1975 (Chicago Univer sity Press, 1982). Moore, Geoffrey H. Business Cycle Indicators, vol. 2 (Princeton University Press, 1961). (September 6, October 24, and 25,1929). The Great Boom and Panic (Henry Regnery Peters, William S., and George W. Summers. Statistical Analysis for Business Decisions (Prentice-Hall, Inc., 1968). Powell, Bill. “The Prophets of Gloom ’87,” Newsweek (September 14,1987), p. 56. Pierce, Phyllis S. Irwin, 1982). The Dow Jones Averages 1885-1980 (Dow Jones- Roepke, William. Crises and Cycles (William Hodge and Company, Ltd., 1936), pp. 51-52. Schumpeter, Joseph A. Business Cycles (McGraw-Hill, 1939J, vol. i 2. Schwartz, Anna J. "Understanding 1929-1933,” in Karl Brunner, ed., The Great Depression Revisited (Martinus Nijhoff, 1981), pp. 5-48. Schwartz, John, and Dody Tsiantar. “The Market’s Latest Bull Run,” Newsweek (August 24,1987), p. 32. Shiller, Robert J. “ Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?” American Economic Re view (June 1981), pp. 421-36. Singleton, Kenneth. “ Speculation and the Volatility of Foreign Cur rency Exchange Rates,” Carnegie-Rochester Conference Series on Public Policy (Spring 1987), pp. 9-56. Sirkin, Gerald. “The Stock Market of 1929 Revisited: A Note,” Business History Review (Summer, 1975), pp. 223-31. Snyder, Carl. Capitalism the Creator (The Macmillan Company, 1940), p. 229. ________ A Monetary History of the United States, 1867-1960 (Princeton University Press, 1963). Sparling, Earl. 1930). Galbraith, John Kenneth. Varian, Hal R. “ Catastrophe Theory and the Business Cycle,” Eco nomic Inquiry (January 1979), pp. 14-28. Gordon, Robert A. 1952). http://fraser.stlouisfed.org/ 28 Federal Reserve Bank of St. Louis The Great Crash (Houghton Mifflin, 1955). Business Fluctuations (Harper and Brothers, Mystery Men of Wall Street (Blue Ribbon Books, Wall Street Journal. October 26,1987. FEDERAL RESERVE BANK OF ST. LOUIS Wanniski, Jude. 116-48. NOVEMBER 1987 The Way the World Works (Basic Books, 1978), pp. Willis, Parker H. “ Who Caused the Panic of 1929?” North American Review (February 1930), p. 183. West, Kenneth D. “ Dividend Innovations and Stock Price Volatility,” Discussion Paper #113 (Princeton University Working Paper, July 1986). Wonnacott, Thomas H., and Ronald J. Wonnacott. Introductory Statistics for Business and Economics (John Wiley and Sons, 1977). 29 FEDERAL RESERVE BANK OF ST. LOUIS NOVEMBER 1987 Appendix Price Bubbles The following assumes rational investors with infinite time horizons and a complete set of markets. With these assumptions, the solution for the expected price of a share of stock next period given the information set in t, E,(P,+1 |wt), is its price this period, P„ plus appreciation during the period at the market rate of discount, r, • P,, less the expected dividend in t + 1, E,(Xt+1 |w,l.‘ This relation ship is summarized in equation 1. ID E,(Pt*, |w,l = P, + r,P, - E.IX,,, |w,l The fundamentals price is the discounted present value of the expected future stream of dividends. This is shown in equation 2 for the price in period t.z Note that r, is the i'h period interest rate. 00 12) P, = 2 e.E.IX,., I wt) i= l e , = 1/(1 + r, I1< 1 If the expected dividend receipt is the same in each future period, E,(XI+I |w,) = E,(X,+1 |w t) for alii; and the yield curve is flat so that r, = r, for all i, equation 2 can be rewritten in the following form.3 (3) P, = E,(X1+1 |w,)/r, . Substituting (3) into (1) and collecting terms gives the solu tion that the expected price in period t +1 is the price in period t. The notion (expressed by Sen. King and others) that the Coolidge market was the product of a price bubble that eventually burst is approximated by a theory that allows share prices to deviate from the fundamentals price in period t by bubble, B„ with probability tt.4 The average duration of the bubble is 1 / ( 1 - t t ) periods before it crashes. Given the assumptions regarding expected future divi dends and the shape of the yield curve, a solution for the price that allows for bubbles, P,', is: (5) P,' = E,(X1+1 I Cj),l/r, + B, B, = t t G' B, , + U, w ith p ro b a b ility tt B, = U, w ith probability 1 —tt E,(U, I <J>, ,) = 0. Substituting (5) into (1) and collecting terms gives the solu tion that the expected price of a share next period is its price this period plus the appreciation in price due to the period t bubble. (61 E,(p;„ I <|>,l = P,'., + r,B, As long as the bubble lasts, the actual rate of return from holding the stock exceeds the market rate of discount, r. This compensates for the risk of a crash in the share price should the bubble burst. The price in t 4-1 is the sum of the expected price and a white-noise error term. (7) P,\, = E,(P1+1|()V + e, = P,' + r,B, + 6, E,(P,+1 |w,) = P,. The observed price in t +1 can be expressed as the period t expectation of the price in t +1 (which, by the above argu ment, is equal to Pt) and a white noise error term, e1+1, as in equation 4. (4) Pltl = P, + €ttl Equation 4 is consistent with the efficient markets solution for asset prices. It implies that prices follow a random walk. (81 E,(P,'+, —P,') = r,B, = r.iTe-’B,^ > 0 Notice that the expected change grows over time at rate r so the market price is expected to deviate further from the fundamentals price in each subsequent period for as long as the bubble lasts. Furthermore, as shown below, the expected percentage change in the price is not constant. (9) E,[(P,' ’ See Brealey and Meyers (1984), pp. 45-47, and Blanchard and Watson (1982), pp. 296-97. 2See Shiller (1981), Blanchard and Watson (1982), West (1986) and Mankiw, Romer and Shapiro (1985). 3The data are consistent with this assumption during the period analyzed in the shaded insert on page 21. For example, the average difference between the yield on high-grade corporate bonds and the call money rate was -30 basis points, which is not significantly different from zero (t-score = .74). Furthermore, the data are con sistent with the assumption regarding expected dividends. It is not possible to reject the hypothesis that dividends per share followed a random walk. The first differences of dividends per share are a white noise process. The Box-Pierce statistics at lags 6 ,1 2 ,1 8 and 24 are 6.94, 12.33, 14.10 and 17.47. The dividend data are from Cowles (1938). The data are annual for the period 1871-1930. http://fraser.stlouisfed.org/ 30Bank of St. Louis Federal Reserve —P.'l/P,'] = r,B/P,' Substituting for P ' from (5) and noting that the fundamen tals price, P| = E,(X1+l |<|>t)/r,, gives (10) E,[(P,\,-P,')/P,'] = r,B,/(P; + B,) = r,/(PJ/B, —1). Since B, grows at rate r, the percentage change in price is expected to rise over time. In contrast to the efficient markets solution, bubbles imply that share prices do not exhibit random walk proper ties. 4See Blanchard and Watson (1982), pp. 297-98.