The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
E C O N O M I C 1 9 8 6 2 Metropolitan Wage Differentials: Can Cleveland Still Compete? How high are Cleveland’s wages, relative to those in other cities, when differences in worker skills are held constant? Skill-adjusted wage differentials for 44 cities show that Cleve land’s wages are higher than the national average, but lower than in some “ fastgrowth” cities. In addition, wage differen tials for Cleveland and other cities have nar rowed substantially in the last decade. The Effects of Supplemental Income and Labor Productivity on Metropolitan Labor Cost Differentials. Author Thomas F. 9 Luce examines the effects that a combina tion o f supplemental income and labor productivity have on the measurement o f metropolitan labor-cost differentials in manufacturing. Using data for the 20 largest Standard Metropolitan Statistical Areas (SMSAs), he finds that controlling for these factors increases the measured labor-cost differentials among these SMSAs. He also finds that the data do not support the pro position that higher-than-average wage rates are associated with greater-than-average labor productivity. Reducing Risk in Wire Transfer Sys tems. Wire transfer provides an effi cient electronic method for moving huge sums o f money— sometimes as much as $500 billion per day— in the nation’s pay ment system. Some users have com e to rely on daylight credit generated by wire transfer systems. This practice creates risk for those extending credit as well as a systemic risk o f disrupting worldwide financial markets. Author E. J. Stevens discusses wire transfer systems, a new Federal Reserve Board of Governors’ risk control policy, and some institutional changes that might be expect ed to reduce risk in this policy environment. R E V I E W Q U A R T E R Economic Review 2 is published quar terly by the Research Department of the Federal Reserve Bank of Cleve land. Copies of the issues listed here are available through our Public Information Department, 216/579-2047. Editor: William G. Murmann. Assistant editor: Meredith Holmes. Design: Michael Galka. Typesetting: Liz Hanna. Opinions stated in Economic Review are those of the authors and not necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System . Material m ay be reprinted provided that the source is credited. Please send copies of reprinted materials to the editor. IS S N 0013-0281 Metropolitan Wage Differentials: Can Cleveland Still Compete? by Randall W. Eberts and Joe A. Stone Randall W . Eberts is Associate Pro This paper was prepared for the fessor, and Joe A . Stone is W . E. Labor Conference, which was held Miner Professor, Department of Eco M ay 2 ,1 9 8 5 , in Cleveland, and which nomics, University of Oregon. was sponsored jointly by the Regional Economic Issues (REI) Pro gram and the Federal Reserve Bank of Cleveland. The study was sup ported in part by a grant from the Cleveland Foundation through the REI Program. The authors would like to acknowledge useful comments by Michael Fogarty, director of the REI Program, and by Joseph Antos. In addition, the authors are especially indebted to Ralph Day for technical computer assistance. Introduction Labor costs are often cited as one of the primary7 reasons for the economic hardships plaguing many older industrial cities, such as Cleveland. O f course, other factors, such as local taxes, proxim ity to markets, product cycles, and energy costs may also contribute to the area’s diminished abil ity to compete with other regions in attracting and retaining businesses. Nonetheless, since labor costs represent an important part of total production costs, the initial presence of signifi cant wage differentials among metropolitan areas may have been a major factor in the economic expansion of Sunbelt cities and the relative decline of Snowbelt cities. In turn, divergent pat terns of growth resulting partly from firms relocat ing in low-wage areas may have caused wage lev els to converge. With respect to the effect of differ ential labor costs on firm location and on regional employment growth, two aspects of labor costs must be considered. First, there is more to examining labor cost differentials across regions than simply looking at regional differen ces in wage rates. Firms consider not only the amount they pay workers, but also the productiv ity of their workers. Stated simply, an employer is willing to pay a worker in Cleveland a higher hourly wage than a worker in Atlanta, for exam ple, if the Cleveland worker is more productive than the Atlanta worker. Therefore, a comparison of regional wage differentials is much more meaningful when these wages are adjusted for differences in worker skills. Second, the advantage to a firm in searching for a low-wage area is directly propor tional to the degree of regional dispersion in labor costs. A large regional variation in labor costs would make it advantageous for firms to search for low-wage areas, since the relative cost savings would be sizable. O n the other hand, if wage differentials, adjusted for worker skills, are observed to converge over time, then the competi tive disadvantage of relatively high-wage areas, such as Cleveland, would diminish over time. The purpose of this paper is three fold: to provide estimates of variations across metropolitan areas in the wage employers pay a worker of given skills and training, to compare these “skill-adjusted” wage differentials with observed differentials, and to examine how these differentials may have changed over the past decade. The Cleveland metropolitan labor market is used as a point of comparison to highlight how labor costs in a major industrial city in the Fourth Federal Reserve District fare with respect to other U. S. cities. I. Theoretical Framework Metropolitan areas in the United States are charac terized by many firms that act as price-takers when they sell to national markets and that con sider the rental prices of capital to be fixed by ex ternal conditions (see Borts and Stein [1964]; and Muth [1968 and 1983] )• This demand-side in terpretation of regional labor markets fixes local nominal wages by the horizontal labor demand curve of firms competing in national or interna- tional product markets. Long-run equilibrium levels of local wages are determined by the demand for labor, under the technical condition that the level of output changes in constant pro portion to changes in labor and capital. Shifts in labor supply have no long-run effect on local nominal wages in this model, but supply changes do cause changes in total employment and even- Estimates of Wage Equations for 1974 and 1983 ( Current Population Survey data) Variable 1974 1983 Intercept 1.26 (39.08) 1.58 (115.27) Schooling 0.12 (9.15) 0.13 (30.70) Schooling squared 0.007 (2.17) 0.004 (3.18) 0.024 (31.39) 0.026 (114.22) Potential experience Potential experience squared Employment status (full time = 1) Gender (female =1) Race (nonwhite =1) Occupation dummy variables (omitted for brevity) R-square Number of observations -0.0004 -0.0004 (-25.05) (-86.03) 0.14 (14.25) 0.16 (58.34) -0.31 (-35.15) -0.23 (-96.52) -0.05 (-3-90) -0.02 (-6.82) ------ ------ 0.49 13,733 0.49 175,268 NOTES: Coefficients are follow ed by /-statistics in parentheses. The 1983 re gression also contains quarterly dum my variables to control for variations during 1983. See text for definition o f variables and further explanation o f data. All coefficients are statistically significant at the 0.01 percent level, except for schooling squared in 1974, which is statistically significant at the 0.05 percent level. TABLE 1 tually in total population. O f course, other influ ences on local wages are possible in short-run dis equilibrium and even in long-run equilibrium, if local products are relatively unique or sold in geo graphically limited markets, if local natural re sources are a significant input into the production of exportable goods, or if any of the other condi tions of the demand-side model above are violated. Johnson (1983) provides an extensive theoretical and empirical analysis of many of these factors, including local costs of living, environmental amenities important to workers, taxes, income transfers, moving costs, unionization, transporta tion costs, discrimination, and various human capital and skill variables. Most of the previous studies of geographical wage differentials have allowed a dominant role for labor supply in determining local wages (see Coelho and Ghali [1971 ]; Bellante [1979]; Sahling and Smith [1983]; Scully [1969]; and Johnson [1983]). Without necessarily denying a role for nondemand factors, the purpose of our study is to obtain estimates of metropolitan wage dif ferentials relevant for identifying demand-side effects and to explore the possible significance of such effects over the past decade. To do this, we first estimate the demand-side differentials for 1974 and 1983, and then examine the trends in the differentials between the two periods. Under the demand-side model, the change in skilladjusted wage differentials during this period is expected (all else the same) to be inversely related to subsequent rates of economic growth via firm locations, expansions, and contractions. We have found in Eberts and Stone (1985), for example, a significant inverse relationship between metropolitan wage differentials in the 1970s and subsequent firm locations. Therefore, one would expect wage differentials measured in 1974 to narrow by 1983. II. Data and Empirical Results The data used to estimate the metropolitan wage differentials are obtained from 1974 and 1983 (CPS) compiled by the Bureau of Labor Statistics. The 1974 data come from the May survey, which contains sup plementary questions regarding employment. The 1983 information is derived from questions asked of one-quarter of the individuals in each of the 12 monthly surveys. Because of this differ ence (and also because of other changes in the CPS between 1974 and 1983), the total number of workers with sufficiently complete records for analysis is much smaller in 1974 than in 1983 (13,733 workers in 1974 versus 175,268 in 1983). The sample allows us to identify 43 of the largest metropolitan areas— Standard Metropolitan Statis tical Areas (SMSAs)— for both years of data. Our first step in obtaining skilladjusted wage differentials is to specify estimable wage equations that reflect appropriate demand determinants of the wages of individual workers. This approach follows the human capital specifi cation of individual wages set forth by Hanoch (1967) and Mincer (1974). Thus, we specify indi vidual wages (expressed in logarithms) as a func tion of observed determinants of individual productivity— education level (entered as a quad ratic), potential experience (age, minus years of education, minus six, also entered as a quad- Current Population Surveys ratic), a binary7dummy variable indicating full time employment status, and 46 binary occupa tion dummy variables (with one of these omitted as a constant). Binary dummy variables are also 1974 Metropolitan Wage Differentials (percentage difference from national average) Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 SMSA New York Paterson San Francisco Detroit Chicago Nassau-Suffolk Rochester San Jose Portland Gary San Diego Anaheim Seattle Los Angeles Albany Akron Cleveland Atlanta Denver New Orleans Baltimore Sacramento Indianapolis Minneapolis-St. Paul Milwaukee Columbus Boston San Bernardino Houston Newark Philadelphia St. Louis Pittsburgh Cincinnati Miami Kansas City Dallas Ft. Worth Birmingham NonSMSAs and other SMSAs Buffalo Norfolk Greensboro Tampa Skilladjusted 18.6 17.9 17.5 17.1 15.7 15.5 14.5 13.4 13-3 12.9 12.9 12.2 9.1 8.4 8.3 7.9 7.5 6.5 5.9 5.9 5.8 5.7 5.5 5.1 4.9 4.3 4.1 3.9 3.8 3.7 3.1 1.4 0.6 0.5 -0.6 -1.8 -2.9 -4.4 -4.7 -5.8 -6.9 -7.1 -8.1 -15.9 Actual 21.6 18.6 19.8 23.6 16.1 24.8 19.9 23.7 16.8 10.5 21.2 27.3 24.4 10.8 18.7 3-8 14.4 2.8 11.4 -0.8 5.4 9.0 8.9 9.8 8.0 3.9 9.4 5.0 10.4 3.6 6.3 1.7 -1.6 -0.3 0.2 3.6 -0.9 -0.5 0.1 -8.7 -4.9 -7.6 -8.6 -17.9 entered to control for race and gender differences in wages. Under the assumptions of the demand model, the separate wage regressions for 1974 and 1983 yield coefficients that reflect national average marginal productivities in specific occu pations and for particular human capital compo nents. Industry dummy variables and union mem bership status are not included, because these variables are not viewed as productive attributes. Detailed information on other components of labor compensation (pensions, health insurance, and the like) is not available in the data. The predicted wage level for each worker in the sample is obtained by multiplying the estimated coefficients by each worker’s char acteristics. The predicted wage can be interpreted as the compensation a worker could expect to receive, given his or her characteristics, regardless of geographic location. Subtracting the predicted wage from the actual wage, then, nets out the portion of the actual wage that is related to the worker’s skills. The skill-adjusted metropolitan wage differentials are then obtained by averaging the wage residuals (actual, minus predicted wage) in each year for all workers in a particular metropolitan area. Average wage differentials are calculated for each of the 43 SMSAs for each year. The national average wage differential is, of course, equal to zero by the property of leastsquares regression. For purposes of comparison, an additional average is calculated jointly for nonSMSAs and other excluded SMSAs. Wage regressions. The estimated (log) wage equations for both 1974 and 1983 are presented in except that the 45 estimated coeffi cients for the occupation dummy variables are omitted for brevity. These equations are pres ented only to document the results of our demand-oriented wage regressions. Except for the absence of nondemand factors (for example, controls for union membership), these are famil iar regressions (with minor variations) in the labor literature. The estimated coefficients in are as expected in both years. Schooling (with a value equal to 1 for eight to 11 years, a value of 2 for 12 to 15 years, a value of 3 for 16 to 17 years, and a value of 4 for more than 18 years) enters with a significantly positive coefficient.1 Schooling squared also enters with a significantly positive coefficient; years of potential experience table 1, table 1 NOTE: Wage differentials are derived from C urrent Population Survey’ files, using the technique described in the text. I This specification of education permits greater nonlinearity in the effects of different education levels than the use of individ ual years of education, although the difference is trivial for our estimated wage differentials. enters with a positive coefficient; experience squared enters with a negative coefficient; a dummy variable for full-time employment enters with a positive coefficient; and dummy variables 1983 Metropolitan Wage Differentials (percentage difference from national average) Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 SMSA San Francisco San Jose Anaheim Seattle Minneapolis-St. Paul Nassau-Suffolk Houston Los Angeles Chicago San Bernardino Detroit Gary Dallas Portland Paterson Sacramento Denver Newark Milwaukee New York San Diego Cleveland Rochester New Orleans St. Louis Ft. Worth Pittsburgh Atlanta Boston Kansas City Baltimore Philadelphia Cincinnati Akron Greensboro Columbus Indianapolis Buffalo NonSMSAs and other SMSAs Albany Birmingham Miami Norfolk Tampa Skilladjusted 18.1 18.1 15.5 14.9 12.0 11.3 10.8 10.7 10.4 10.0 9.3 8.4 8.0 7.9 7.8 7.8 7.6 7.3 7.1 7.1 5.7 5.1 5.1 4.8 3.9 3-4 2.7 2.7 2.2 2.1 1.6 1.5 1.3 -1.3 -1.6 -2.2 -2.5 -2.6 -4.8 -6.0 -6.9 -6.9 -7.3 -10.7 Actual 25.4 28.4 23.0 22.1 12.6 16.0 14.5 13.0 14.5 7.1 9.1 3.6 12.2 9.7 13.5 7.0 12.7 12.7 7.9 11.4 4.5 7.0 11.0 8.8 4.0 3.0 5.4 6.2 5.7 4.5 4.8 4.7 1.4 1.4 -3.5 -2.1 -2.6 -4.5 -7.1 -5.4 -5.1 -11.4 -7.3 -11.7 NOTE: Wage differentials are derived from C urrent P opulation Survey files, using the technique described in the text. for race (with nonwhite equal to 1) and gender (with female equal to 1) enter with negative coef ficients. All listed coefficients are significant at the 5 percent level. With the exception of the decline in absolute value for the race and gender coefficients in 1983, the 1974 and 1983 regressions are basic ally the same. The similarity extends, by and large, to the 45 occupation dummy variables as well, although a few of these coefficients do change. Intercepts in the two equations, of course, differ significantly, due to both nominal and real wage growth between 1974 and 1983 for the Unit ed States as a whole. Both regressions explain 49 percent of the variation in actual wages. Metropolitan wage differentials. Skill-adjusted and actual metropolitan wage differentials (expressed as the percentage deviation from the national average) are presented in for 1974 and in for 1983- The SMSAs are ranked according to the size of the skill-adjusted differential. Because of the semilogarithmic speci fication of the wage equation, residuals are expo nentiated to obtain percentage differentials. The rankings offer a perspective on how Cleveland’s wages compare with regions against which the area might compete for eco nomic development. In 1974, Cleveland’s skilladjusted wage was 7.5 percent above the national average, which put Cleveland in seventeenth place among the cities considered. A number of cities usually associated with rapid growth, such as San Jose, San Diego, and Anaheim, had wage differen tials that were higher than Cleveland’s. O n the other hand, Cleveland’s skill-adjusted wages are consistently higher than they are in southeastern cities. About one-quarter of the cities with wage rates below Cleveland’s level were in the South east, and no southeastern city had a skill-adjusted wage differential higher than Cleveland’s. More over, small SMSAs and nonSMSA regions showed much lower skill-adjusted wage differentials than Cleveland’s— over 12 percent lower. In 1983, Cleveland’s skill-adjusted wage fell to only slightly above 5 percent of the national average, which brought its ranking down to twenty-second place. All the southern cities in the sample still had wage differentials below Cleveland’s. A few additional cities, such as San Bernardino and Sacramento, were added to the 1974 list of west coast cities that surpassed Cleve land in the skill-adjusted wage differential. Wage differences between metro politan areas can be broken down into two compo nents: differences in the skill-adjusted wages and differences in the value of skills (measured in dollars). Consider the difference in actual wages between two SMSAs ( and ). Recall that: table 2 table 3 w\ w0 5 (1) log 0 0) = log ( u\ ) = bS0 + ?0, bS, + e x, b where is the regression coefficient associated with the skill-related variables (5), and denotes the residual or skill-adjusted wages (actual wage, minus predicted wage). We assume that the appropriate aggregation has been done, so that each equation represents wages in a specific met ropolitan area. The difference in the actual (log) wages between the two metropolitan areas is: e ( u \ ) - log ( w 0) = b(Sx - S 0) + (e v e0) (2) log The first component on the right-hand side is the difference in levels of skills normalized in wage units between the two areas. The second is the difference in skill-adjusted wages between the two metropolitan areas. If, for example, the actual wage differential is greater than the skill-adjusted differential, then the skill level is necessarily greater in area 1 than in area o. Consider the wage differentials displayed for San Francisco in 1983. The actual wage in San Francisco is 17.2 percent higher than the actual wage in Cleveland, but the skill-adjusted wage is only 12.4 percent higher.2 The difference of 4.8 percentage points is due to the higher skill levels of San Francisco workers relative to Cleveland workers. Since employers are willing to pay workers the value of their contribution to the production of each unit of output, the higher wages associated with higher productivity do not affect the relative competitiveness of the two areas. Rather, it is the difference in wages over and above the differen tial associated with higher labor productivity that affects competitiveness among regions. In the case of San Francisco, a 12.4 percent wage differ ential exists, which is not accounted for by skill differentials. On the other hand, Rochester’s 3-2 percent wage differential relative to Cleveland is due entirely to higher skill levels in Rochester. 2 The percentage difference in wages between any tw o metropol itan areas can be easily calculated from tables I and 2, by using the following formula: WvW° = I WyWus - \ ^us wus J (1/(1+(w0-wus)/wus)) w here ( w , - w0) /w 0 is the percentage difference in wages between area 1 and area o and (w j -w us)lw us is the percentage deviation in w a g e s in area i from the n ation's (the differential dis played in tables 2 and 3). Although the results for 1974 show a rough correspondence between skill-adjusted and observed (actual) wage differentials, substan tial differences are also clearly evident. Detroit, Anaheim, Birmingham, San Diego, Cleveland, Houston, and Boston, for example, all have observed wage differentials that exceed the skilladjusted differential by at least 8 percentage points, which is the approximate differential required for statistical significance at the 5 per cent level. Only Akron exhibits the opposite phenomenon— a skill-adjusted differential that is at least 8 percentage points higher than the observed differential. The five SMSAs with the highest skill-adjusted wages are New York, Pater son, San Francisco, Detroit, and Chicago. The five lowest SMSAs are Tampa, Ft. Worth, Greensboro, Norfolk, and Buffalo. The results for 1983 show a stronger correspondence between skill-adjusted and observed wage differentials. By this year, no SMSA except San Jose has an observed wage dif ferential that differs from the skill-adjusted differ ential by at least 8 percentage points. Only one of the five highest-wage SMSAs in 1974 (San Francis co) remains in the top five in 1983- The remain ing four in 1983 are San Jose, Anaheim, Seattle, and Minneapolis-St. Paul. Two of the lowest-wage SMSAs in 1974 (Tampa and Norfolk) remain among the five lowest SMSAs in 1983. The remaining three in 1983 are Albany, Birmingham, and Miami. The changes in the differentials between 1974 and 1983 are presented in SMSAs with the largest increases are Dallas, Ft. Worth, Houston, Minneapolis-St. Paul, and Greensboro. SMSAs with the largest decreases are Albany, New York, Paterson, Rochester, and Akron. Most of the cities associated with rapid growth during the last decade exhibit increases in both skill-adjusted and observed wage differen tials. In some instances, the skill-adjusted and actual changes in wage differentials differ sub stantially. SMSAs that show increases in the skilladjusted differential, but a decline in the actual wage differentials, are Houston, Anaheim, and Sacramento. For these SMSAs, the skill-adjusted increase is presumably offset by a decline in average skill level. Cleveland’s skill-adjusted and ob served wage differentials fell between 1974 and 1983; the actual wage declined more rapidly than the skill-adjusted wage. Since the relative decline in the actual wage differential, with respect to the skill-adjusted wage differential, has to be offset by a decline in average skill level of the area’s work force, this indicates that Cleveland suffered a de cline in the average skill of the area’s labor force. New Orleans, Philadelphia (trivial ly), Atlanta, and Akron show decreases in the skill-adjusted wage differential, but an increase in table 4. the actual wage differential. For these SMSAs, the skill-adjusted decrease is presumably offset by an increase in the average skill level. Large diver gences between the skill-adjusted and actual Change in Wage Differentials from 1974 to 1983 (percentage point change) Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 SMSA Dallas Ft. Worth Houston Minneapolis-St. Paul Greensboro San Bernardino Seattle Tampa San Jose Buffalo Kansas City Newark Anaheim St. Louis Los Angeles Milwaukee Pittsburgh Sacramento Denver NonSMSAs and other SMSAs Cincinnati San Francisco Norfolk New Orleans Philadelphia Boston Birmingham Cleveland Atlanta Nassau-Suffolk Baltimore Gary Chicago Portland Miami Columbus San Diego Detroit Indianapolis Akron Rochester Paterson New York Albany Skilladjusted 10.9 7.8 7.0 6.9 6.5 6.1 5.8 5.2 4.7 4.3 3.9 3-6 3.3 2.5 2.3 2.2 2.1 2.1 1.7 1.0 0.8 0.6 -0.2 -1.1 -1.6 -1.9 -2.2 -2.4 -3.8 -4.2 -4.2 -4.5 -5.3 -5.4 -6.3 -6.5 -7.2 -7.8 -8.0 -9.2 -9.4 -10.1 -11.5 -14.3 Actual 13-1 3.5 4.1 6.9 5.1 2.1 -2.3 6.2 6.2 0.4 0.9 -9.1 -4.3 2.3 2.2 -0.1 7.0 -2.0 0.7 1.6 1.7 5.6 0.3 9.6 -1.6 -3.7 -5.2 -7.4 3.4 -8.8 -0.6 -6.9 -1.6 -7.1 -11.6 -6.0 -16.7 -14.5 -11.5 -2.4 -8.9 -5.1 -10.2 -24.1 NOTE: Wage differentials are derived from Current Population Survey files, using the technique described in the text. changes (even if the changes are in the same direction) have similar interpretation. Other SMSAs with large differences between the two measures are San Jose, Birmingham, Gary, San Diego, Detroit, and Albany. Based on the estimates above, have skill-adjusted metropolitan wage differen tials converged since 1974? This question can be answered by calculating the change in the coeffi cient of variation from 1974 to 1983. The coeffi cient of variation is the standard deviation (com puted from the sample of SMSA-level wage differentials) divided by the mean; thus, it is an index of the degree of dispersion in the sample. This measure indicates substantial convergence for both sets of differentials, declining by 22 per cent for the skill-adjusted differentials and by 46 percent for the actual wage differentials. Because the observed wage differential is composed of the skill-adjusted wage differential and a differen tial related to differences in actual skills, the fact that observed wages converged more than twice as much as skill-adjusted wages suggests that variations across metropolitan areas in actual skill levels also declined during the period.3 Why do we observe relatively strong wage convergence during the 1974-1983 period? Following our demand-side approach, one could attribute convergence to the expand ing scope of most product markets (both domes tically and internationally), increased competition faced by geographically concentrated firms that may have had some power to influence price, the relative decline of industries that make products using relatively large amounts of local natural resources, and the emergence of manufacturing industries that require smaller-scale plants. III. Conclusion The objective of this paper was to provide esti mates of variations across metropolitan areas in the wage employers pay a worker of given skills and training, and then to compare how these dif ferentials have changed over the past decade. Based upon 1974 and 1983 data from the we find substantial variations in skill-adjusted wages in both 1974 and 1983, as well as significant deviations between skilladjusted and observed wage levels. We also find that the wage differentials and skill differentials converged significantly during this same period. Cleveland’s skill-adjusted and actual wage levels Population Survey, 3 Current The change in average skill level could be the result of changes in actual skills or of changes in the market compensation of the skills between 1974 and 1983. The general similarity of the 1974 and 1983 wage regressions, however, suggests that most of ihe change in skill level reflects actual changes in skills. also converged toward the national averages. Over the last decade, Cleveland closed the gap by 2.4 percentage points for skill-adjusted wages and by nearly 9 percentage points for actual wages. The reduction in Cleveland’s wage differentials and the general convergence in wages and skills could influence Cleveland’s economic future in at least two ways. First, the incentive for firms to move out of Cleveland might diminish, since convergence in wages reduces the potential cost savings of a move. Second, the wage differential might not be as crit ical a factor in economic growth as it once was. In fact, labor supply-side factors, such as labor climate and local amenities and public services, might become more important influences on economic development. References Bellante, Don. “The North-South Differential and the Migration of Heterogeneous Labor,” vol. 69, no. 1 (March 1979), pp. 166-75. Amer ican Economic Review, Borts, George H. “The Equalization of Returns and Regional Economic Growth, vol. 50, no. 3 (June I960), pp. 319-47. "American Economic Review, Economic Growth ------- , and Jerome L Stein. New York: Columbia Univer sity Press, 1964. in a Free Market. Coelho, Philip R, and Moheb A Ghali. “The End of the North-South Wage Differential,” vol. 6, no. 5 (December 1971), pp. 932-37. can Economic Review, Ameri Eberts, Randall W., and Joe A Stone. “The Deter minants of Firm Location,” Department of Economics, University of Oregon, 1985. Working Paper, Hanoch, Giora. “An Economic Analysis of Earn ings and Schooling,” vol. 2, no. 3 (Summer 1967), pp. 310-29. Resources, Journal of Human Johnson, George E. “Intermetropolitan Wage Dif ferentials in the United States,” in Jack E. Tri plett, ed., Chicago: University of Chicago Press, 1983, pp. 309-30. The Measurement oj Labor Cost. Schooling Experience, and Learn Mincer, Jacob. New York: National Bureau of Economic Research, Distributed by Columbia University Press, 1974. ing. Muth, Richard F. “Differential Growth Among Large U.S. Cities,” in James P. Quirk and Arvid M. Zarley, eds., Lawrence, KS: University of Kansas Press, 1968. nomics. Papers in Quantitative Eco _________ “Intermetropolitan Wage Differentials in the United States. Comment,” in Jack E. Tri plett, ed., Chicago: University of Chicago Press, 1983, pp. 330-32. The Measurement oj Labor Cost. Sahling, Leonard G., and Sharon P. Smith. “Regional Wage Differentials: Has the South Risen Again?” vol. 65, no. 1 (February 1983), pp. 131-35. tics, Review oj Economics and Statis Scully, Gerald W. “Interstate Wage Differentials: A Cross Sectional Analysis,” vol. 59, no. 5 (December 1969), pp. 757-73. Review, American Economic The Effects of Supplemental Income and Labor Productivity on Metropolitan Labor Cost Differentials by Thomas F. Luce Thomas F . Luce is associate direc tor, Wharton Philadelphia Economic Monitoring Project, University of Pennsylvania. The author would like to thank Anita A . Summers and Michael F. Fogarty for their helpful comments on preliminary drafts. Introduction Regional variations in labor costs attract a great deal of attention because of their potential to affect the regional distribution of economic activ ity. Because of the major role that labor costs play in total production costs, regional differences in labor costs may translate into dramatic regional differences in profitability. Profitability, in turn, is a major determinant of whether or not existing firms will expand and of whether or not new firms are likely to locate in a given region. Most studies of regional variations in labor costs are based solely on data for payroll per employee or wages. However, the measure of labor costs that is most relevant to the profitability of a firm is the total cost of the labor needed to produce its output. There are at least two steps involved in getting from simple wage data to estimates of labor costs per unit of output. First, nonwage income to workers ( “supplemental income”) must be added to wages to get total labor costs per hour. In 1977, the value of measurable supplemental income paid to manufacturing workers was, on average, about 20 percent of the value of wages paid to those workers. This percentage showed consider able variation, however, among regions and states. (See Garofalo and Fogarty [1984].) By 1982, supplemental income had increased to nearly 25 percent of wages. Second, labor costs per hour must be scaled by the amount of output per hour that is attributable to labor inputs, as opposed to other inputs, if the objective is to measure labor costs per unit of output. The amount of output generated by a unit of labor input varies for dif ferent workers and for different production pro cesses. Labor productivity will be different among workers possessing different skills or other per sonal characteristics. Similarly, productivity will vary for a single worker according to the amount of other factors of production (machinery, energy, or raw materials) used in a particular production process. The research described here has been directed toward incorporating these two corrections to the raw wage rates to obtain a more accurate measure of labor costs per unit of output. The supplemental income estimates are a direct extension of work done at the state level by Garofalo and Fogarty. The strategy employed to control for labor productivity differs significantly from that used in most other studies. Researchers in terested in analyzing regional labor productivity patterns face hard decisions regarding the tech niques and data available to them. One option is to use indirect indicators of productivity that can be measured reliably, but which may or may not be reliable proxies for labor productivity. This op tion uses the personal characteristics of workers to measure productivity. The strength of the tech nique is that it uses data that are relatively access ible and reliable. It has two major weaknesses: (1) Strong assumptions are required about the rela tionship between the indirect measures (personal characteristics) and labor productivity. (2) The method does not control for differences across industries or regions in capital intensity. The other choice—the one used in this analysis— is to attempt to estimate labor pro ductivity directly from data that measure output and input levels. The strength of this approach is that the effects of differences in the quality of labor of differences in the mix of inputs (capital intensity) are both captured by the mea sure. This approach also has two major weakness es: (1) Reliable data are not easily available at the regional level for some of the measures, especial ly capital inputs. Some variables must, therefore, be estimated (with some error) from the data that are available. (2) The option requires one to make fairly strong assumptions about the nature of pro duction processes across industries and regions. Neither approach is entirely satis factory7since each requires strong assumptions. It could be argued that the first approach (based on characteristics of the labor force) provides the more relevant measure for new firms seeking a location, because these kinds of firms are not tied to an existing technology or physical plant. On the other hand, the second approach, because it controls for the effects on labor productivity of existing capital intensities, may be the better measure for capturing the potential of existing firms to expand in their current location. Indexes for wage rates, supplement al income, and labor productivity have been gen erated for each of the 20 commonly reported Man ufacturing subsectors (two-digit SIC industries) in the 20 largest (based on 1980 employment) Stan dard Metropolitan Statistical Areas (SMSAs). and I. Wages The wage data collected for this research support the finding by other researchers that wage rates vary significantly among metropolitan areas and regions in the United States. The first column of shows 1982 Manufacturing wage indexes for the 20 largest SMSAs. The indexes represent production worker hourly wages in the SMSA as a percent of the national average. In order to con trol for the fact that different SMSAs have different industrial structures, each SMSA’s index compares the cost of that SMSA’s employment mix com puted from the SMSA’s wage structure to the cost of the same mix at national average wages. This means that an artificially high index number will not be produced simply because an SMSA has greater-than-average concentrations of employ ment in industries that have wages that are higher than the average for Manufacturing as a whole. The first thing that is clear from the wage indexes is that, in 1982, there was a great deal of variation in Manufacturing wages among these large SMSAs. Wages in the lowest-wage SMSA (Nassau) were only 75 percent of those in the highest-wage SMSA (Pittsburgh). Eight of the table 1 SMSAs showed wages less than the national aver age, but the weighted average wage for the 20 was 2 percent greater than the national average. The most striking feature of the regional averages is that all of the SMSAs in the North Central region showed Manufacturing wages in 1982 that exceeded the national average. Wages in Cleveland approximately matched the regional average at about 8 percent above the national average and were greater than in all but six of the 20 largest SMSAs. Wages in the North east, South, and West were close to the national average, but the indexes are far from being uni formly distributed within these regions. The Northeast, for instance, shows the lowest regional average despite the fact that it contains the highest-wage SMSA in the sample (Pittsburgh). Similarly, wages in the South and West range from a low of 91 percent of the average (Dallas) to a maximum of 115 percent (San Francisco). These regional patterns, particu larly the finding that the Northeastern SMSAs showed lower wages on average than those in the South and West, are somewhat surprising. In light of the often-cited difference between wages in the “Sunbelt” and the “Frostbelt,” one might have expected greater regional differences than those revealed by the data. One possible explana tion for the patterns is that wages have converged over time as the result of equilibrating forces at work in the national economy. In regions where Manufacturing employment is in decline, one would expect downward pressure on relative wages. Examination of column (7) of provides some support for this view. Wages in the slower-growing Northeast and North Central regions have indeed declined relative to those in the South and West. However, the decline was significantly greater in the Northeast than the North Central and there are some clear excep tions at the SMSA level— for example, relative wages in Pittsburgh increased between 1977 and 1982 despite significant declines in Manufactur ing employment during the period. table 1 n. Supplemental Income Estimates of supplemental income for the 20 largest SMSAs showed even more variation than the wage indexes, ranging from about 75 percent of the national average to more than 130 percent of the average. The supplemental income data include available in the both mandatory supplements to wages like social security and worker’s compensation, and voluntary supplements like health and life insurance. Other, less easily measured fringe benefits, such as free parking or subsidized cafeterias, are not included. Unfortunately, the regional data are reported at the state level only, with no detail Census of Manufactures across industries. Industry detail is available only in the national data. Estimates of supplemental income for each of the 20 Manufacturing sectors in the SMSAs have been generated by combining wage data from the SMSAs with the information about regional variations in fringe benefits rates contained in the state level data and with the information about variations among industrial sectors contained in the more detailed national data. The procedure assumes supplementary in come in a given industry and SMSA to be the product of (1) the level of wages in that industry' and SMSA (2) the average supplemental income rate (supplemental income divided by wages) for the industry in the nation, and (3) the average supplemental income rate in total Manufacturing in the SMSA’s home state (controlling for the Manufacturing employment mix in the state). The supplemental income estimates for each industry in an SMSA are then combined in the same way as the wage estimates to get the mix-controlled index for total Manufacturing in the SMSA The supplemental income indexes in column (2), compare the supplemen tal income cost of the SMSA’s employment mix to the cost of the same mix at the national average supplemental income rates. The estimates for 1982 show the differences among the SMSAs’ fringe benefits rates to be much more substantial than for wages. The supplemental income rate in table 1, Manufacturing Wage, Supplemental Income and Labor Productivity Indexes: 20 Largest SMSAs, 1982 United States =100 (1 ) Simple wage index Twenty largest SMSAs3 102.2 (2 ) (3 ) (4 ) (5 ) Supplemental income index Total labor cost index Labor productivity index Corrected labor cost index (6 ) (7 ) Column ( 5 ) Column ( 1 ) minus change column ( 1 ) from 1977 (8 ) Column ( 5 ) change from 1977 98.1 101.5 99.1 102.4 0.2 -0.8 -0.7 0.4 -6.0 1.6 3-4 -6.4 3.8 23.5 -2.6 -0.8 -2.2 -2.0 -3.6 -2.8 5.5 -3.3 -5.5 -1.8 -5.9 -4.9 20.6 Northeast3 Boston Nassau Newark New York Philadelphia Pittsburgh 98.6 96.5 89.1 96.5 95.5 101.6 118.9 93.3 85.4 74.1 85.4 83.8 110.2 131.4 97.7 94.5 86.4 94.4 93-5 103.3 121.6 98.7 104.3 95.2 94.6 105.0 98.0 85.3 99.0 90.5 90.7 99.9 89.1 105.4 142.4 North Central3 Chicago Cleveland Detroit Minneapolis St. Louis 108.0 103.2 107.7 118.4 109.0 104.8 109.9 106.5 114.9 129.0 93.2 97.2 108.5 103.9 109.2 120.6 106.1 103.2 95.9 94.8 94.1 92.0 103.4 104.8 113-0 109.5 116.0 131.1 102.6 98.5 5.0 6.3 8.3 12.7 -6.4 -6.3 -0.9 -0.1 -1.7 -1.3 3.9 0.1 5.3 1.8 -0.4 South3 Atlanta Baltimore Dallas Houston Washington, DC*3 100.4 93.8 111.3 91.3 109.1 101.4 87.8 77.9 119.2 74.3 88.8 137.5 98.0 90.7 112.9 88.0 105.1 107.6 101.6 102.6 99.3 103.9 100.1 99.2 96.4 88.4 113.7 84.7 105.0 108.5 -4.0 -5.4 2.4 -6.6 -4.1 7.1 1.7 -2.4 4.5 1.0 4.9 -4.5 -0.1 0.2 0.7 -3.2 6.6 -4.0 West3 Anaheim Denver Los Angeles San Francisco 101.1 99.1 105.7 98.3 114.8 96.7 94.8 92.0 95.0 111.0 100.3 98.3 103.1 97.7 114.1 102.5 98.4 101.0 97.9 99.9 102.1 94.0 112.2 -3.2 0.8 -3.6 -4.3 -2.6 2.5 4.0 4.8 2.4 1.5 0.2 4.2 2.2 -1.2 2.5 103.9 101.7 -3.2 3-4 7.4 2.2 SOURCE: Com puted from Census o f M anufactures, 1977 and 1982, Bureau o f the Census, U.S. Department o f Com m erce. a. Aggregate indexes are weighted averages o f the SMSA estimates with weights based on manufacturing em ploym ent in the 20 largest SMSAs. b. Supplemental incom e index for Washington, DC based on 1977 supplemental incom e data. 11 Nassau (the SMSA showing the lowest index number) was only 54 percent of that in Washing ton, DC (the SMSA with the highest rate). The overall pattern of supplemen tal labor costs reflects the wage pattern. Higherthan-average wage rates tend to be associated with higher-than-average supplemental income rates. Eight of the 12 SMSAs with higher-thanaverage wages had higher-than-average supple mental income rates. Similarly, all eight of the lower-than-average wage SMSAs also showed lower-than-average supplemental income. Cleve land’s standing in supplemental income rates mirrored its position in wages— only three of the 20 SMSAs showed supplemental income rates higher than Cleveland’s. Total labor costs in a specific industry and SMSA are calculated as the sum of wages and supplemental income. The industryspecific estimates are then combined in the same way as the wage and supplemental income mea sures to get a mix-corrected estimate of the laborcost index for total Manufacturing. Because of the rough correspondence between the wage in dexes and the supplemental income indexes, and because supplemental income represents a much lower percentage of total compensation than wages, the total labor-cost indexes (colum n [3]) do not differ dramatically from the wage indexes. In only four of the 20 SMSAs is the difference greater than three percentage points (Atlanta, Dal las, Houston, and Washington, DC). In general, adding supplemental income to the labor-cost indexes increases the spread among the SMSAs, but not by a substantial amount. III. Labor Productivity Factoring labor productivity into the labor-cost estimates is also very important. If higher-thanaverage wages in an SMSA reflect higher-thanaverage labor productivity, then the index for the SMSA from column (3) of will overstate any relative disadvantage that the SMSA might have in competing with other regions for jobs. The potential for labor productivity to significantly alter the competitive position of an SMSA is greater than it is for supplemental income. Supplemental income represented only about 24 percent of total compensation in 1982. Any adjustment made for labor productivity, however, affects 100 percent of total compensa tion. Consequently, equivalent percentage differ ences in the two factors in an SMSA will have dif ferent effects on the overall measure of labor costs in the SMSA, with the productivity adjust ment being the greater of the two. Differences among SMSAs in labor productivity can arise from two different sources. First, differences in labor quality due to skill lev table 1 els, education, or experience are likely to be reflected in differences in productivity. Many ana lysts, therefore, use various labor-force character istics as proxy measures for productivity. How ever, there is another important source of productivity differences. The amounts of other factors of production that are used in combina tion with labor will affect labor productivity inde pendently of the quality of labor. The productivity of otherwise identical workers will be different depending on the amount of capital (such as machinery and equipment) used in combination with them in the production process. Ideally, controlling for labor pro ductivity differences arising from both sources would require that the researcher have industryand SMSA-specific data for output, capital inputs, labor inputs, and all other inputs. In addition, knowledge about the production process, itself— how capital and labor are combined at different levels of output— is required. With this kind of data, it would be possible to separate the portion of output directly attributable to labor from the part attributable to the other inputs. These kinds of data are not readily available, particularly at the SMSA level. However, the reports value-added data by industry and SMSA From these data, it is possible to estimate the amount of output that is directly attributable to a unit (one hour) of labor input, after controlling for the amount of capital used in the production process. The labor cost per unit of output can then be estimated by div iding the cost per unit of labor input by output per unit of labor input. The procedure used in this research to make this calculation involved two steps. First, capital inputs were estimated by sub tracting total labor costs (including the costs for nonproduction workers) from value added and dividing this difference by an estimate of the rate of return to capital. Second, the value of output attributable to a unit of labor input was estimated by assuming that capital and labor are combined in a specific way in the production process.1The Census of Manufactures I ln theory, value added represents only the contribution to the total value of output that is made at the stage of production in question. A n y contributions to value that are made at prior stages in the total production process (such as by the refining of raw materials, or pre assembly of components) are not included in the “value added" at the stage of production under analysis. In reality, the value-added data reported by the Bureau of the Census includes some value that cannot be directly attributed to the labor or capital brought to bear at the stage of production in question. The capital estimates used here thus overstate capital stock. H ow this overstatement affects the relative measures used in this analysis depends on the extent to which the magnitude of noise in the data varies from S M S A to S M S A , a piece of information which is not available. The production process assumed for the analysis is a constant retums-to-scale Cobb-Douglas production function with a capital exponent of 0.282 (taken from Hulten and Schwab [1984]). result is a measure of how effectively the SMSA’s labor force is combined with the existing capital plant. By estimating productivity directly from output data (albeit with some strong assump tions), it is not necessary to make any assump tions about how labor force characteristics, such as education or age, affect productivity. If an SMSA’s labor force possesses productivityenhancing characteristics, the impact should be captured in the estimate of output that is directly attributable to labor inputs. Labor productivity estimates derived by using this procedure show much less variation across the 20 largest SMSAs than either the wage or supplemental income indexes. reports the labor produc Column (4) of tivity indexes for the 20 SMSAs. The index repre sents labor productivity in Manufacturing in the SMSA as a percentage of national average labor productivity in Manufacturing. Productivity in the lowest-productivity SMSA (Pittsburgh) is about 85 percent of the national average and about 81 per cent of the value for the highest-productivity SMSA in the group (New York).2 A primary reason for investigating labor productivity is to test whether higher-thanaverage labor costs in an SMSA reflect higherthan-average labor productivity. Comparisons of the third and fourth columns of suggest that this is not the case in the 20 largest SMSAs. Indeed, the simple correlation coefficient— a mea sure of how closely two variables move together— between the labor productivity indexes and the wage indexes is negative, indicating that, in these SMSAs, higher-than-average wage indexes are asso ciated with lower-than-average labor productivity.3 The result of this negative relation ship is that, when labor productivity is factored into the labor-cost indexes, the spread among the SMSAs increases. Column (5) shows the labor cost per unit of output indexes. The lowest-cost SMSA (Dallas) showed labor costs in 1982 that were just under 60 percent of those in the highest-cost SMSA (Pittsburgh). table 1 table 1 2 The very low index for Pittsburgh is largely due to the index for the S M S A 's dominant sector — Primary Metals. Reported value added in this sector for 1982 w as less than total labor costs for the sector, a relationship which is conceptually troublesome and which is inconsistent with the labor productivity calculation. The difference between reported value of shipments and cost of materials was there fore substituted for reported value added in the productivity estimation procedure. Consequently, the productivity measure for Pittsburgh should be viewed with caution, since it is likely that the problems resulting from the use of available value-added data (see fn. 1) are particularly acute in Pittsburgh's case. V j The correlation coefficient is 0.52. IV. Combined Effects of Supplemental Income and Labor Productivity The supplemental income and labor productivity adjustments to the simple wage index tend to op erate in the same direction. This was true for 17 of the 20 largest SMSAs. In each of the seven SMSAs where the supplemental income adjustment increased the labor-cost index, the productivity adjustment also increased it. Similarly, in 10 of the 13 SMSAs where the supplemental income cor rection decreased the labor-cost index, the pro ductivity correction also resulted in a decrease. The net change in the labor-cost measure resulting from the two adjustments is shown in column (6) of In 11 of the 20 SMSAs, the net effect of the two adjustments was to decrease the labor-cost index. In these SMSAs, the simple wage index overstates relative labor costs. In the other nine SMSAs (including Cleve land), the simple index understates costs relative to the national average. The magnitude of the under- or overstatement varied substantially from SMSA to SMSA, with the understatement being the greatest for Pittsburgh, and the overstatement being the greatest for Dallas. Overall, these results suggest that simple wage measurements will tend to distort regional labor-cost differentials. O n average, the wage indexes understate relative labor costs in the higher-cost, North Central SMSAs, and over state them in the lower-cost SMSAs in the South and West. In addition, the productivity correc tion has a very significant effect on the measured change in labor costs between 1977 and 1982. The increases in costs in the South and West re flected in the simple wage indexes are largely off set by improving relative labor productivity during the period (colum n [8], O n the other hand, the decline in relative wages in the North Central region is overwhelmed by the decline in the relative productivity measure. Only in the Northeast does the productivity correction have little effect on the measured change in labor costs. The net effect is that the competitive posi tion (as measured by the productivity-corrected labor-cost indexes) of the Northeastern SMSAs im proved on average between 1977 and 1982, while the North Central’s position deteriorated, and those of the South and West remained unchanged. What are the implications of laborcost differentials of the magnitude found in 7? Statistical analysis, relating employment growth between 1977 and 1982 to relative labor costs in 1977 in the 20 largest SMSAs, suggests that they have been significant in the past. (See Summers and Luce [1985].) The finding was that, after control ling for the effects of national employment trends, unionization rates, right-to-work legisla table 1. table 1). table tion, energy7costs, vulnerability to international competition, state and local taxes, cost of living, and local amenities, a labor-cost differential of 50 percent in 1977, like the one that existed between Dallas and Detroit, was associated with a subsequent employment growth differential of almost 3 percent per year. The actual differential for these two SMSAs for the period from 1977 to 1982 was about 10 percent per year, implying that the labor-cost differential explained almost 30 percent of the total difference in growth rates. case. Manufacturing employment declined much more quickly in these six SMSAs between 1977 and 1982 than in the other 14, or in the nation as a whole. In the six, total Manufacturing employ ment declined by more than 5 percent per year over this time period, compared to a decline of less than 1 percent per year in the other 14. V. Relative Labor Costs in Cleveland In Manufacturing as a whole, Cleveland fell into Manufacturing Wage, Supplemental Income and Labor Productivity Indexes: Cleveland SMSA, 1982 United States =100 (1 ) Simple wage index 1 4 (2) (3) (4) (5 ) Supplemental income index Total labor cost index Labor productivity index Corrected labor cost index (6 ) (7 ) Column ( 5 ) Column ( 5 ) minus change column ( 1 ) from 1977 Total manufacturing3 107.7 114.9 109.2 94.1 116.0 8.3 2.2 Durables3 Lumber products Furniture and fixtures Stone, clay and glass Primary metals Fabricated metals Non-elec. machinery Elec. machinery Trans, equipm ent Instruments Other durables 109.0 108.2 110.8 100.2 116.5 115.5 118.3 106.9 118.7 119.2 94.6 102.3 107.8 116.9 107.0 103.9 104.3 121.7 118.9 123.8 7.9 -1.2 113.3 124.9 112.7 100.8 106.0 110.6 109.4 112.0 101.5 113-2 113.2 107.5 118.5 107.2 95.5 100.5 2.4 -2.7 -20.1 -0.5 23.5 2.8 11.9 -2.1 Nondurables3 Food and kindred Textiles Apparel Paper and allied Printing and publishing Chemicals Petroleum products Rubber and plastics 102.7 102.0 85.1 161.6 89.1 110.3 95.7 74.2 91.6 107.5 108.9 90.8 172.4 95.1 117.7 102.1 79.2 97.7 103.5 103-3 86.0 163.1 90.2 111.4 97.0 75.2 92.8 92.7 93.7 112.0 125.8 81.1 87.1 92.5 94.0 91.9 111.3 111.7 106.2 117.0 105.6 94.5 99.4 97.3 93.0 95.2 86.9 114.9 98.4 90.5 87.0 103.1 109.0 105.6 115.5 111.7 110.3 76.8 129.7 111.3 128.0 104.9 80.0 101.0 -6.9 4.1 10.4 7.2 17.6 -13.9 3.4 11.1 16.1 9.0 8.3 -8.3 -31.9 22.2 17.7 9.2 5.8 9.4 -29.3 -2.8 -4.1 1.8 -22.1 -6.0 13.5 14.9 11.9 2.1 -18.7 -3-8 SOURCE: Com puted from Census o f M anufactures, 1977 and 1982, Bureau o f the Census, U.S. Department o f Com m erce, a. Aggregate indexes control for industrial structure. TABLE 2 For the SMSAs showing higherthan-average labor costs and lower-than-average productivity in 1982 (Philadelphia, Pittsburgh, Chi cago, Cleveland, Detroit, and Baltimore) the im plications of this finding are particularly sobering. The statistical analysis implies that those SMSAs would have had to possess very significant cost advantages from other sources, such as greaterthan-average access to input or output markets, to have been competitive with other areas in the Unit ed States. This does not appear to have been the the group of SMSAs in 1982 (composed primarily of the older SMSAs in the North and East) with higher-than-average wages, higher-than-average supplemental income, and lower-than-average labor productivity. It is of interest to examine whether this pattern carries over into the specific industrial sectors that are of greatest importance to the region. shows the 1982 labor-cost measures, described above, broken out by the 18 sectors for which data are available for Cleveland. Table 2 Some caution should be exercised in evaluating the results presented in The primary reason for this is the level of industrial disaggregation used in the analysis. In the same way that total Manufacturing measures that do not control for different industrial structures across SMSAs may over- or understate labor-cost differ ences, the two-digit SIC breakdowns in may reflect differences between Cleveland and the nation in industrial structure at a finer level of disaggregation. This problem, in fact, appears to be a factor in at least two of the sectors shown in It is likely that the very low wage index for Petroleum Products and the very high index for Apparel are largely the result of this issue. However, since these two sectors, together, accounted for less than 5 percent of Manufactur ing employment in the region, they have very lit tle impact on the overall indexes. Nearly 70 percent of 1982 produc tion worker employment in Manufacturing in the Cleveland SMSA was contained in the five sectors beginning with Primary Metals in Each of these sectors showed higher wages and supple mental income in Cleveland than in the nation as a whole. In addition, only one of the five ( Elec tric Machinery) showed labor productivity signifi cantly above the national average. Two others (Fabricated Metals and Transportation Equip ment) showed labor productivity within five per cent of the average. However, productivity advan tages in none of these sectors were large enough to offset the significantly higher-than-average wage and supplemental income rates. Overall, productivity-conected labor costs exceeded the national average in all but two of the reported 18 sectors. In addition, the re gion’s competitive position deteriorated between 1977 and 1982 in eight of the 18 sectors and in three of the region’s five largest sectors (Primary and Fabricated Metals, and Nonelectric Machin ery). Labor costs clearly cannot be viewed as a factor enhancing the region’s desirability to firms competing in national and international markets. What impact are differences of the magnitude found in Cleveland likely to have on future employment growth or decline in the region? The research cited in previous sections suggests that the impact was very significant between 1977 and 1982. The findings implied that a labor-cost differential like the one found for Cleveland in 1977 (14 percent) was associated with subsequent employment growth in Manufac turing, which was about 0.8 percent per year less than it would have been if labor costs had been equal to the national average. This represents more than one-fifth of the total difference between the growth rate in the Cleveland SMSA and that in the nation between 1977 and 1982 (when the average difference was about 3 6 per table 2. table 2 table 2. table 2. cent per year). Although other factors working to Cleveland’s disadvantage explain the majority of the region’s slower-than-average employment growth in the period, the effect of higher-thanaverage labor costs cannot be ignored. A 0.8 per cent per year shortfall in growth represents about 7,000 Manufacturing jobs in the SMSA over the five-year period from 1977 to 1982. VI. Conclusions Manufacturing labor costs varied significantly among large SMSAs in 1982. Most of the variation was attributable to differences in wage rates. When supplemental income was added to wages to get total labor costs per hour, the spread among SMSAs increased, but not by a substantial amount. Correcting for differences among SMSAs in labor productivity tended to increase the dif ferentials by more than the supplemental income adjustment but by a magnitude that was less than the original wage differentials. The data for the 20 largest SMSAs do not support the proposition that higher-than-average wage rates are associated with greater-than-average labor productivity. Labor costs in 1982 for the Cleveland SMSA were significantly greater than the na tional average. O f the overall 16 percentage point differential, about 50 percent (or eight percentage points) was due to greater-than-average wage rates. Another 40 percent of the total difference was attributable to lower-than-average labor productiv ity, with the remaining 10 percent being due to greater-than-average supplemental income rates. The higher-than-average labor costs in Cleveland are likely to have had a dam pening effect on employment growth in Manufac turing in the region. In the group of the 20 largest SMSAs, labor-cost differentials of the magnitude evident in Cleveland in 1977 were associated with employment growth about 0.8 percent per year less than if labor costs had equaled the national average. This represents about one-fifth of the total difference in Manufacturing employ ment growth rates between Cleveland and the nation between 1977 and 1982. The overall implication of this research for the Cleveland area is that, in order to compete effectively with other areas of the coun try for Manufacturing jobs, other characteristics of the region must be sufficiently advantageous to overcome the region’s relatively high labor costs. Many of the same market forces that operated in the past to create the higher-than-average wages in the region are likely to lead in the future to some moderation, but this is a slow and painful process. Wages in Cleveland as a percent of national average wages declined by only 2 per cent between 1977 and 1982 — a period when Manufacturing employment in the region 15 decreased by 25 percent. In addition, the margi nal improvement in the region’s competitive position embodied in the relative wage decline was more than offset by a decrease in relative labor productivity in the region. Perhaps the most important mes sage from the analysis is that there is room for improvement in the SMSA in one component of labor costs— labor productivity—that can be en hanced over a shorter time horizon by actors within the region. Any improvements in this direc tion will require both a commitment by labor to productivity-enhancing changes in work rules and incentive structures, and by management to invest in the region to maintain and improve the physical plant. Neither group, working alone, can significantly improve the region’s ability to com pete in national and international markets. References Bureau of the Census, Census of Manufactures, Geographic Area Statistics U.S. Department of Commerce, 1977 and 1982. Garofalo, Gasper A., and Michael S. Fogarty, “Census Facts,” The Regional Economic Issues Program, Cleveland, May, 1984, pp. 26-30. REI Review, Hulten, Charles R., and Robert M. Schwab, “Regional Productivity Growth in U.S. Manu facturing: 1951-1978, 1984, pp. 152-162. Summers, Anita A., and Thomas F. Luce, nomic Review”Watch. “The American Eco Eco nomic Report on the Philadelphia Metropoli tan Area, 1985, University of Pennsylvania Press, Philadelphia, 1985. Reducing Risk in Wire Transfer Systems by E. J. Stevens E .J . Stevens is an assistant vice president and economist at the Federal Reserve Bank of Cleveland. The author would like to acknowl edge useful conversations with David B. Humphrey. Introduction Hundreds of billions of dollars in payments are made each day in the United States. The system that enables this enormous sum to change hands includes several different mechanisms. Probably the largest number of payments, but with the smallest total dollar value, are made by using coins and paper money. Another very large num ber of payments, with a daily total value in the neighborhood of $75 billion, are made by using checks, credit cards, and direct transfers through automated clearinghouses. The smallest number of payments, but representing by far the largest total dollar value— frequently $500 billion a day— are made using so-called wire transfers of funds. Wire transfers move balances elec tronically at Federal Reserve Banks from one bank’s deposit account to another’s on the same day. Transfers can be carried out over any of sev eral wire networks (large-dollar transfer systems) connecting banks to one another and to the Fed eral Reserve Banks.1 In this way, banks make pay ments that handle their own short-term financing transactions as well as payments on behalf of themselves and their customers. These payments, in turn, reflect much of the dollar-denominated securities and foreign exchange market trading of the world. March 27, 1986 was the effective date of a Federal Reserve Board of Governors’ policy to control risks in large-dollar transfer sys tems. Adjustment to that policy has been smooth, as expected, for two reasons.2 First, consultation and public comment on the need for and nature of the program have been ongoing for a number of years. The actual policy was announced in May of last year. Since then, both the Federal Reserve Banks and private consultants have been conduct ing informational meetings for banks across the nation. Second, the risk-control mechanism that became effective on March 27 embodies only a modest initial effort at risk reduction. With the mechanism in place, however, future steps to re duce risk become more feasible. How smoothly fu ture risk reduction can be assimilated will de pend on the ease with which financing practices of banks and institutional arrangements for mak ing certain kinds of payments can adapt to the ris ing cost of risk im plied by the risk-control policy. This article briefly describes sources of risk in large-dollar transfer systems and dis cusses major features of the new mechanism for risk control.3 Then, examples of potential changes 2 This expectation was supported by a survey done just before March. See “ Findings: Survey on Implementation Status of Reduction of Paym ent System Risk," Bank Administration Institute, Ja n uary 23, 1986. I 3 A full description of the policy m ay be found in “ Policy Statement Regarding Risks on Large-Dollar Wire Transfer System s” (Docket The word "bank” will be used here in a generic sense, and in No. R-0515), Board of Governors of the Federal Reserve System . Dis cludes commercial banks, thrift institutions, Edge A c t and Agree cussion of the risk problem is in: E . J . Stevens, "Risk in Large Dollar ment Corporations, U . S . Branches and Agencies of Foreign Banks, and Transfer S yste m s,” F R B of Cleveland New York Article XII Investment Companies. 2-16. Economic Review, Fall 1984, pp. 17 in financing and payments practices that might facilitate future risk reduction are examined. I. Risk Exposure The risk being controlled is the threat that pay ments made over one of the large-dollar transfer systems can’t be settled. None of these systems operates on a real time, cash-in-advance basis that would continuously settle by deducting each pay ment, minute by minute, as it occurs, from the bal ance in an account. Instead, they are “batched” set tlement systems that update accounts only at the end of the day by the net of payments and receipts during the day. It is possible, therefore, for a de pository institution to transfer large sums during the day before it has received all the funds needed to settle its account at a Federal Reserve Bank. If the needed funds can’t be acquired, a settlement failure occurs. A settlement failure is a rare event in the United States. Many banks have failed to open in the morning, but few in modem history have failed to settle their accounts at a Federal Reserve Bank the previous evening. Ultimately, who stands to lose in the event of a settlement failure depends in part on whether the largedollar transfer system involved is a net settlement system, or Fedwire. The leading example of a net settlement system is CHIPS (Clearing House Interbank Payments System), a private telecom munications clearinghouse payments network operated by the New York Clearing House. Partic ipants exchange provisional payments messages during the day, but payments become final only at the end of each day when the net position (receipts, minus payments) of each participant is settled through accounts at Federal Reserve Banks.4 Inability of a participant to settle in this type of system suggests that one or more other participants or their customers are at risk because the Federal Reserve will not effect a net settle ment order at day’s end if one or more partici pants have insufficient balances. O n the other hand, Fedwire, a wire transfer system operated by the Federal Reserve, makes payments by transfer ring funds directly from one depository’s account at a Federal Reserve Bank to that of another.5 Inability of a Fedwire user to cover its payments at the end of a day means that a Reserve Bank takes the loss, because funds received by a bank over Fedwire during the day are irrevocable once notification of a payment is received. In both cases, risk arises because a bank can send more funds before the end of a day than are covered by its initial balance, plus its receipts, to that point during the day. Such a prac tice results in a “daylight overdraft.” For example, consider a bank continuously borrowing overnight in the federal funds market: each morning it returns the previous day’s borrowing over Fedwire, but can’t actually cover that return of funds until later in the day when new borrowing has been arranged and received. The risk is that a bank might be unable to arrange sufficient new borrowing and therefore fail to repay its daylight overdraft. Daylight overdrafts reflect daylight credit provided to the overdrafting bank either by the Federal Reserve on Fedwire, or by other banks on a net settlement system. The practice of relying on daylight credit creates credit risk for banks vis-a-vis their customers, for Federal Reserve Banks vis-a-vis Fedwire users, and for par ticipants in net settlement systems vis-a-vis one another. Systemic risk is also created in the last case because the unexpected failure of one bank to settle might have a ripple effect as that failure makes it impossible for other banks to settle. In such a case, there is the potential for causing a classic banking crisis that could disrupt financial markets worldwide.6 Rapid growth of large-dollar transfers relative to reserve deposit balances sug gests that banks commonly resort to daylight credit to finance payments during the day.7 The Federal Reserve does not con done daylight overdrafts and, until relatively re cently, they were probably rare. It was not until 1979 that the first measurement of daylight over drafts was taken. Therefore, aggregate values of transfers relative to banks’ deposit balances at Federal Reserve Banks is only suggestive of the likely growth of daylight overdrafts. Transfers were only about 20 percent of balances in 1950, 150 percent in 1970, but approaching 3000 percent in the past few years. Now, with use of powerful computerized accounting systems, it is possible for a bank to maintain an on-line monitor of its own and customers’ daylight overdrafts. The Fed eral Reserve is able to monitor the daylight over drafts of a bank across all large-dollar networks, at least after the fact. In the future, large-dollar trans- This brief description simplifies a more complex settlement pro cess. Only 22 banks' accounts actually receive a debit or credit at Federal Reserve Banks. Ten of these banks settle for the remaining 112 participants. A failure might reflect the inability of one of the 22 to settle 6 A thorough analysis of systemic risk is in David B. Humphrey, "Paym ents Finality and Risk of Settlement Failure: Implications for Financial M arkets." Paper prepared for Conference on Technology its own position, or of one of the associate banks to meet its settlement and the Regulation of Financial M arkets, N ew York University, M ay obligation with a settling participant. 1985. 5 It is immaterial that the depositories m ay hold accounts at differ ent Reserve Banks because the Federal Reserve Banks "settle- up” among themselves at the end of each day. 7 Marcia L . Stigum cites the example of a large m oney center bank with daily payments 2V& times its Market, total assets. The Money Hom ew ood, Illinois, D ow Jo nes, Irwin, 1983. p. 585-6. fer systems conceivably could operate on-line real time monitors that would prevent the use of daylight credit completely, thus requiring that cash be available in advance of each payment. Daylight credit exposure is not a unique indicator of risk. Risk depends on the probability that institutions will not cover their daylight overdrafts by the end of a day, as well as, in the event of an actual failure to settle, the probability that claimants won’t recover some or all of their loss in the liquidation of a failed insti tution. Payment system risk then depends jointly on the amount of daylight credit, on the sound ness of institutions in daylight overdraft positions, and on the ability of depository institutions to control the amount of payments-related credit extended to other depository institutions during a day. Systemic risk—the risk that otherwise sound institutions w ill be swept up in a cascade of set tlement failures— depends as well on the interre latedness of institutions in the payments system. This is influenced heavily by the ability of the central bank, in its role as lender-of-last-resort, to prevent or isolate a settlement failure by provid ing overnight credit at the end of a day. Reliance on daylight credit is not troubling in itself. Rather, it is the uncontrolled and unrationed provision of daylight credit that is troubling. As long as daylight credit is unrationed, risk creation is subsidized and daylight credit becomes overused. Fedwire has no explicit price for providing daylight credit and, because there is no well-developed private market for daylight credit, has little basis for setting such a price. Until the current risk-control policy began to be developed, Fedwire also did not have an effective limit on daylight overdrafts for any but visible problem banks. It can be argued that there is im plicit pricing of daylight credit in net settlement systems.8 Receivers of funds transfers (suppliers of daylight credit) face a cost in the form of some pro bability of loss. They therefore have an incentive to lim it the amount of daylight credit they extend to each other participant. However, this argument is weak, unless the computerized net settlement system provides a feature that both allows partic ipants to set such limits, and enforces them by preventing transfers that would breach a limit. Moreover, the whole argument breaks down when, as appears to have been the case, there is a widespread presumption among banks that the Federal Reserve, as the lender-of-last-resort, would lend to a participant that is otherwise unable to settle rather than let a settlement failure take place and risk a systemic wave of failures. 8 For discussion of risk and pricing, see David L . Mengle, “ Daylight Overdrafts and Payment System Risks," Economic Review, eral Reserve Bank of Richmond, M ay/June 1985 pp. 14-27. Fed II. The Mechanism for Risk Control The risk-control policy establishes three require ments for every net settlement system: 1) each participant should be able to set a bilateral limit on the net amount of daylight overdraft credit it is willing to extend to each other participant; 2) each participant should be subject to a lim it on the amount of daylight overdraft credit it uses; 3) the net settlement system should include an on line monitor to reject or hold payments that would breach either limit.9 In the case of Fedwire, banks will be subject to a daylight overdraft lim it in the form of a dual “cap.” One part of the cap limits a bank’s average daylight overdraft position during a two-week required reserve maintenance period. The other part limits a bank’s overdraft during any single day of that two-week period. A potential problem with inde pendent daylight overdraft caps for each largedollar system is that they would not distinguish institutions using only one system from those using two or more systems. Consequently, each net settlement system must provide data to the Federal Reserve so that it can monitor the risk exposure each bank creates simultaneously over all systems relative to that bank’s daylight over draft cap on Fedwire. If a bank’s overdrafts across all systems exceed this limit, the Federal Reserve Bank could counsel the bank and/or advise the appropriate examiner about the situation, or the Federal Reserve could reject a bank’s Fedwire transfers that exceed its overdraft limit. A bank seeking permission to run daylight overdrafts must undertake a selfevaluation of its creditworthiness, credit policies, and operational controls and procedures. This selfevaluation must include a review by its own board of directors, and the bank must maintain records as a basis for examiner inspection and comment to the directors. The bank thereby will establish its own overdraft limitations, but these must lie within Federal Reserve guidelines. The guidelines are expressed in terms of a multiple of the insti tution’s capital. (See box.) Should this volunatry process not be taken seriously, “...the Board (of Governors) will reconsider its options, including the adoption of regulations designed to impose explicit limits on daylight credit exposure.”10 In summary, each depository insti tution, including each Federal Reserve Bank, can now manage the net amount of daylight credit it extends to each other institution; each institution must undergo self-evaluation necessary to obtain 9 This was a feature of the Board's interim risk-reduction policy adopted in 1982. Policy Statement Regarding Risks on Large-Dollar Wire Transfer System s, p. 10. The Cross-System Net Debit ‘Cap’ At the heart of the new risk-control policy is a cross-system sender net debit cap. The sender is a bank, making pay ments over Fedwire. A net debit cap is a dollar lim it on the amount of daylight credit a bank may draw by sending pay ments in excess of the sum of its opening balance and pay ments received up to any point during the day on Fedwire. The lim it is “cross-system” in that, for banks that participate in net settlement systems such as CHIPS, the amount of day light credit allowed under the lim it set on Fedwire will be reduced by the net amount of daylight credit the bank has drawn on those net settlement systems.3 Clearly, a bank’s cross-system daylight credit use, or net debit position, must vary over a day, beginning and ending at zero, but rising above zero whenever the opening balance, plus payments received, fall short of pay ments made. The cross-system net debit cap has two forms. One is a limit on the two-week average of a bank’s maximum daily net debit position, with the average taken over each two-week required reserve maintenance period. Averaging provides flexibility for banks to operate within the unpredic table ebb and flow of payments traffic, while abiding by the intent of the risk-control policy. The other form of the cap is a lim it on a bank’s maximum net debit during each day of the two-week period. This cap is higher than the two-week average cap, but effectively puts a limit on the flexibility built into the averaging process. If a bank is at the one-day limit for one or more days of the period, then it must be below the two-week average for one or more days in order to stay within the average. Dual Cap Multiple of Adjusted Primary Capital Cap Class High Above Average Average No Cap Two-W eek Average 2.0 1.5 1.0 0 Plus Single Pay 3.0 2.5 1.5 0 The Board’s Policy Statement includes a discussion of the cap-setting procedure banks should employ and how self judgements of creditworthiness, credit policies, and opera tional controls and procedures might be combined into the single summary self-classification required to obtain a cap higher than zero.b a. The Fedwire limit will not be raised if a bank has been a net supplier o f credit on a net settlement system. b. Other details o f the procedure also are included in the Statement, includ ing a definition o f adjusted primary capital; treatment o f Edge Act and Agreement Corporations, U. S. Branches and Agencies o f Foreign Banks, and New York Article XII Investment Companies; and implications for Book- a nonzero lim it on the aggregate net amount of daylight credit it draws from all systems during an interval; the Federal Reserve Banks will monitor the daylight overdraft positions of institutions on Fedwire relative to their self-imposed caps, nor mally after the fact, but net of any daylight credit obtained on other funds transfer networks. III. Institutional Adjustments for Risk Reduction Incredulity was a common reaction to early dis cussions of reducing risk on large-dollar transfer systems.11How could half-a-trillion dollars or more of daily payments possibly be resequenced so that, with only $20-30 billion of cash deposits, those payments could still be made, but with less reliance on daylight overdrafts? Each sender might wait until enough payments were received before payments were sent, but every delayed send would, of course, mean a delayed receipt for someone else. Given the small cash base and limited time during which transfer networks are open (the working day sometimes extended into the even ing), the result seemed more likely to be “grid lock” than smoothly functioning transfers of funds. The emphasis on creating a risk-control mechanism first, with high overdraft limits based on selfevaluation, seems to have submerged this kind of reaction. But when future steps are taken to use the mechanism to reduce risk, how will smoothly operating payments be maintained consistent with reduced daylight overdrafts and reduced risk? Two kinds of changes, induced by market incentives, should take place that could achieve the desired result. One kind would pur chase reduced risk directly, as individual banks reallocate their operating and portfolio resources to live within overdraft limits. The other kind would result from innovations in standard arrange ments for interbank payments and financing. Banks may reduce the amount of daylight credit they extend as well as reduce their own use of daylight credit simply because nationwide attention has focused on the problem. Heightened awareness and bet ter information may bring more prudent behav ior. W hile many banks have monitored and man aged their own and their customers’ daylight overdraft positions for many years, others appar ently have not. As a result of the educational pro gram and preparation accompanying im plem en tation of the Board of Governors’ risk policy, banks now may be less generous in accommodat ing other banks’ and customers’ use of daylight credit, thereby reducing their own need for day light credit. Setting more prudent limits, or col- Direct Risk Reduction: entry Securities Transfers, Automated Clearinghouses, Net Settlement Ser vices, and additional matters. BOX 1 11 See, for example, Stigum. lecting fees for scheduled extensions of daylight credit to customers, would have this effect. Sim ilarly, with the ability to specify binding bilateral net credit limits in net settlement networks, banks may be less generous in accommodating other banks’ use of daylight credit. Risk reduction will then result both from reduced daylight over drafts and from improved credit quality resulting from continuous, explicit risk management. Banks also might delay making some payments until later in the day in order to reduce their reliance on daylight credit. O f course, the resulting delayed receipts might in crease reliance on daylight credit at other banks. However, many depositories and customers never use daylight credit and, in fact, maintain positive balances throughout the day. Thus, some overall re duction in daylight credit is possible through more careful management of the timing of payments. Banks could elect to hold larger overnight balances at Federal Reserve Banks from which to make payments during the day. This might seem to be an expensive adjustment cost ing a bank the foregone earnings on those extra re serves. However, a bank can elect to hold addition al sums as a clearing balance on which earnings credits can be used to pay for priced services. In either case, banks might make this a part of a leastcostly method of reducing daylight overdrafts. Risk declines as bank capital grows, providing more room for institutions to operate within caps set on a “times capital” basis. Maintaining a higher capital position might also seem to be an expensive adjustment, but may be worth the price. Moreover, many banks are already adding, or planning to add, to capital as they adjust to potential loan quality problems and comply with regulatory guidelines for safety and soundness. Even without any change in daylight overdraft practices, more highly capitalized insti tutions might present lower risk. Another fertile field for reducing daylight overdrafts lies in the liability manage ment of depository institutions. About two-thirds of Fedwire transfers reflect federal funds transac tions, as borrowing banks repay the previous day’s borrowing and then, typically, replace that with fresh borrowing for the current day. Extend ing the maturity of bank financing could yield substantial dividends in reduced Fedwire traffic and reduced daylight overdrafts of Federal Reserve Bank accounts. Risk exposure of the Federal Reserve Banks certainly would decline, but risk exposure of others might grow. Longerterm financing would add to lenders’ risk of illiq uidity (that is, of using costly methods to meet unexpected needs for cash) and, all else unchanged, add to lenders’ and borrowers’ inter est rate risk (that is, of unexpected changes in maturity rate spreads). Uninsured lenders, replac ing overnight with longer maturity loans, would also face a slightly different credit risk. No longer could they rely on Federal Reserve Banks to assume credit risk each morning, as they had when overnight loans had been returned. The “musical chairs” of repayment thus would be spaced further apart. Moving the bearer of risk from Federal Reserve Banks to private market lenders does not represent evasion of risk-reduction pol icy. W idening the scope of market scrutiny and the opportunity for risk pricing should be expected to encourage more conservative behav ior by borrowing banks. Substantial reductions in daylight overdrafts at individual banks could emerge from innovations in some long standing market practices. Some of these innovations might only evade the risk-control mechanism by shifting risks outside the monitor, and will not be acceptable.12 Others would, in fact, reduce risk and are to be encouraged. Distinguishing between the innovations will require careful investigation. The three examples of suggested changes discussed here might be acceptable if carefully structured and are offered to indicate the range of ideas being developed in the market in response to the risk-control policy. An alternative to replacing over night financing with longer-term borrowing would be to develop a “rollover” practice in overnight credit markets. Borrower and lender might agree that, unless either wished to termi nate the entire credit, all or part would be rolled over at the relevant daily rate each day. A single daily transfer could cover interest, plus any agreed change in the outstanding amount of the loan. This would eliminate the need to transfer the full amount of borrowing both back and forth each day. Credit risk from overnight lending would remain, but would not become a daily daylight payment risk either for the Federal Reserve or for participants in net settlement systems. Access to a rollover loan, as well as its price, presumably w ould depend on the credit worthiness of the borrower as viewed with more intense lender scrutiny than for a typical over night loan today. In this way, the transfer of risk from Federal Reserve Banks and participants in net settlement systems should generate incentives for more conservative behavior by borrowers. Another substantial portion of the traffic on large-dollar transfer systems flows among banks that, for themselves or for dealer customers, are settling securities or foreign Innovations: -1 The Policy Statement (pp. 30-31) specifically, “reaffirms its _L L d (earlier) policy that institutions m ay not use Fedwire or other payments networks as a method of avoiding risk-reduction measures." 21 22 exchange transactions. Current practice typically involves gross next-day settlement of securities transactions, meaning that banks send one another payments for each transaction. Each day, two banks active in handling security market operations typically will send each other m ulti m illion dollar payments that are more or less offsetting. These payments are initiated and received in automated systems on the basis of trades known in advance because they were done on the previous day. The alternative would be for two banks to offset the payments due to one another, replacing those two payments with a single trans fer of the net difference due to one or the other institution. Daylight credit risk would be reduced if the banks adopted new legal agreements defin ing obligations to be for this net position rather than for gross positions.13 Heretofore, the incen tive for this kind of economizing on payments traffic was primarily the cost of a funds transfer— at most a few dollars per transfer. The additional incentive of avoiding more costly means of day light overdraft reduction might provide the impe tus for devising offset arrangements. As in the case of federal funds rollover, offset payments would not eliminate all risk. Banks would be ex posed to risk of a failure to settle the net amount due, but the amount at risk would be much smaller than the gross amounts now exposed. Development of a day-loan market is another institutional change frequently cited as promising daylight overdraft relief. The Federal Funds market is the source of one-day maturity loans of cash in the form of deposit balances at a Federal Reserve Bank. Similarly, a day-loan market would be the source of loans of cash, but with same-day maturity. Just as banks may charge a fee to customers who daylight overdraft their accounts, so too, for a fee, banks might be able to borrow and lend cash for repayment later in the day. Such a procedure seems technologically feas ible, especially if it were encouraged by provision for priority-funds transfer messages that would bypass a queue of payment orders on large-dollar transfer systems. Some banks will always have positive balances that might be loaned to others who want to make payments but who are at their daylight overdraft limits. A day-loan market is not an institu tional development that would directly reduce risk. Rather, it would transfer risk from the Fed eral Reserve Banks and the whole set of partici pants in net settlement networks to the institutions -1 O J. Such an arrangement, said to be the first of its kind, is expected to start operating soon in London, involving settle ment of foreign exchange transactions among major international banks. See “ International Financing Review ," Issue 622, M ay 1 7 ,1 9 8 6 , pp. 1496-7. making day loans. However, it may indirectly reduce risk by making exposures more visible so that market discipline would ration credit to risky institutions with increased certainty. These three examples of institu tional changes— rollover, offset, and day loans— have not happened yet, but they, and others like them, suggest promising ways in which market practices might be expected to adjust to future efforts to use the new risk-control mechanism to reduce risk in large-dollar transfer systems. IV. Concluding Remarks An important result of the risk-control policy now in place is that each depository institution’s cross system use of daylight credit can be monitored relative to caps that are themselves related to the institution’s self-evaluated creditworthiness. Initial caps are not expected to result in any significant disruption in large-dollar funds transfer service. Nonetheless, some depository institutions are having to adjust their operations to meet the pol icy limitations. This, plus the adjustments of other institutions recently sensitized to the risks, should at least dampen the growth of daylight overdraft risk exposures. However, conclusions must await experience under the new limitations because payments patterns may change in response to these initial adjustments, perhaps creating day light overdraft problems for institutions that had not previously experienced them. Once the situation settles down, the Federal Reserve Board of Governors fully expects to move further toward reducing risk, perhaps, for example, by ratcheting down “timescapital” cross-system daylight overdraft limits. In the meantime, banks can develop operational and institutional changes that will reduce and redirect risk without disrupting the payments system. In return, Federal Reserve Banks’ risk exposure on Fedwire should diminish and market discipline should play a larger role in controlling risk. Economic Commentary The Difficulty in Explaining Postwar Stability KJ. Kowalewski and Eric Kades December 1, 1985 How Desirable Is Dollar Depreciation? Gerald H. Anderson December 15, 1985 A Correct Value for the Dollar? Owen F. Humpage and Nicholas V. Karamouzis January 1, 1986 Bank Holding Company Voluntary Nonbanking Asset Divestitures Gary Whalen January 15, 1986 Junk Bonds and Public Policy Jerome S. Fons February71, 1986 The Dynamics of Federal Debt Can We Count on Private Pensions? John B. Carlson and E. J. Stevens July 1, 1985 James F. Siekmeier February 15, 1986 Is Manufacturing Disappearing? American Automobile Manufacturing: It’s Turn ing Japanese Michael F. Bryan July 15, 1985 Solutions to the International Debt Problem Gerald H. Anderson August 1, 1985 Medicaid: Federalism and the Reagan Budget Proposals Michael F. Bryan and Michael W. Dvorak March 1, 1986 Should We Be Concerned About the Speed of the Depreciation? Owen F. Humpage March 15, 1986 Paul Gary Wyckoff August 15, 1985 The Government Securities Market and Pro posed Regulation The Dollar in the Eighties James J. Balazsyjr. April 1, 1986 Owen F. Humpage and Nicholas V. Karamouzis September 1, 1985 Interstate Banking: Its Impact on Ohio Banks Thomas M. Buynak September 15, 1985 The M1 Target and Disinflation Policy W illiam T. Gavin October 1, 1985 International Trade and the Fourth District’s Recovery Robert H. Schnorbus October 15, 1985 Reserve Borrowings and the Money Market Richard L. Mugel November 1, 1985 Bank Earnings: Comparing the Extremes Paul R. Watro November 15, 1985 A Revised Picture: Has Our View of the Econ omy Changed? Theodore G. Bernard April 15, 1986 Monetary Policy Debates Reflect Theoretical Issues James G. Hoehn May 1, 1986 How Good Are Corporate Earnings? Paul R. Watro May 15, 1986 The Thrift Industry: Reconstruction in Progress Thomas M. Buynak June 1, 1986 The Emerging Service Economy Patricia E. Beeson and Michael F. Bryan June 15, 1986 Economic Review Quarter I 1985 Beauty and the Bulls: The Investment Character istics of Paintings by Michael F. Bryan The Reserve Market and the Information Con tent of Ml Announcements by W illiam T. Gavin and Nicholas V. Karamouzis Quarter IV 1985 Stochastic Interest Rates in the Aggregate LifeCycle/Permanent Income Cum Rational Expec tations Model by Kim J. Kowalewski New Classical and New Keynesian Models of Business Cycles by Eric Kades Quarter II 1985 Vector Autoregressive Forecasts of Recession and Recovery: Is Less More? by Gordon Schlegel Revenue Sharing and Local Public Expenditure: Old Questions, New Answers by Paul Gary Wyckoff Quarter I 1986 The Impact of Regional Difference in Unionism on Employment by Edward Montgomery The Changing Nature of Regional Wage Differ entials from 1975 to 1983 by Lorie Jackson Quarter III 1985 The Impact of Bank Holding Company Consoli dation: Evidence from Shareholder Returns by Gary Whalen The National Debt: A Secular Perspective by John B. Carlson and E. J. Stevens The Ohio Economy: A Time Series Analysis by James G. Hoehn and James J. Balazsy, Jr. Labor Market Conditions in Ohio Versus the Rest of the United States: 1973-1984 by James L. Medoff