MATH GURU ONLY

THE IMPACT OF MARKET STRUCTURE ON WAGES, FRINGE BENEFITS, AND TURNOVER

JAMES E. LONG and ALBERT N. LINK*

This paper examines the relationship between labor compensation and the structure of the product market, which is measured by the industry concentra- tion ratio and by dummy variables for the existence and type of government regulation. Unlike previous studies that have estimated the impact of concen- tration and regulation on wages or earnings, this study extends the analysis to include the effect of market structure on employer-provided pensions and in- surance and on voluntary labor turnover. The hypothesis that product market power raises labor compensation is supported by empirical results indicating that concentration increases wages and fringes but lowers volun- tary labor turnover. Regulations that set minimum prices and restrict entry raise labor compensation, since wage premiums due to regulation are not off- set by lower pensions and insurance or higher turnover. Other forms of regulation, such as profit regulation in public utilities, are found to reduce labor compensation, as evidenced by higher turnover or lower wages and fringes, or both.

A LARGE number of studies have exam- ined the impact of product market

structure on labor earnings. It has been sug- gested that firms witb market power pay relatively bigber wages because (1) work- ers in concentrated industries capture part of tbe monopoly profits, (2) labor costs in concentrated industries can be more easily passed on to consumers, and (3) tbe small number of firms in concentrated industries makes it easier for unions to organize and raise wages. Empirical evidence on tbe relationsbip between concentration and

•James Long is an associate professor of economics at Auburn University and Albert Link is a professor of economics at the University of North Carolina at Greensboro,

wages is mixed, bowever. Weiss, Masters, Hawortb and Rasmussen, Asbenfelter and Jobnson, and Hendricks find tbat con- centration bas no statistically significant effect on wages once industrial character- istics and employee quality are held con- stant.' In contrast, Dalton and Ford esti-

'Leonard Weiss, "Concentration and Labor Earn- ings," American Economic Review, Vol, 56, No, 1 (March 1966), pp, 96- 117; Stanley H, Masters, "Wages and Plant Size: An Interindustry Analysis," Review of Economic? and Statistics, Vol, 51, No, 3 (August 1969), pp, 341-45; Charles T, Haworth and David W, Rasmussen, "Human Capital and Inter-Industry Wages in Manufacturing," Review of Economics and Statistics, Vol, 53, No, 4 (November 1971), pp,376-80; Orley Ashenfelter and George E, Johnson, "Unionism, Relative Wages and Labor Quality in U,S, Manu- facturing Industries," International Economic Re-

Industrial and Labor Relations Review, Vol, 36, No, 2 (January 1983), < 0019-7939/83/3602-0239$01,00

1983 by Cornell University,

239

240 INDUSTRIAL AND LABOR RELATIONS REVIEW

mate that firms in concentrated industries pay higher wages than competitive firms, a relationship Haworth and Reuther find occurring only during cyclical periods of slack demand and stable prices.^ In sum- marizing the effects of market structure on wages, Hendricks points out that the esti- mated effect of concentration depends on the control variables and samples used and on the occupations and time periods exam- ined.'

To the extent that government regu- lation can affect the pricing policies and structure of industries, product market regulation is another potential influence on labor earnings. For example, by limiting the entry of firms into the market, regula- tion may reduce competition and enable unions to raise wages above the level that would have existed in the absence of regu- lation. The resistance of firms to demands for higher wages may be increased, how- ever, by regulations that set maximum prices to insure minimum profits. Hend- ricks investigates the effect of regulation on earnings and finds that annual earnings are lower in regulated industries than in unregulated manufacturing, a result that is consistent with Weiss's findings.^ Never- theless, labor earnings are relatively higher in industries in which regulatory authori- ties set minimum prices and restrict entry, as in trucking and airlines.^ Also, Ehrenberg

view, Vol, 13, No, 3 (October 1972), pp, 488-507; and Wallace Hendricks, "Regulation and Labor Earnings," Belt Journal of Economics, Vol, 8, No, 2 (Autumn 1977), pp, 483-96,

2James A, Dalton and E, J, Ford, "Concentration and Labor Earnings in Manufacturing and Utilities," Industrial and Labor Relations Review, Vol, 31, No, 1 (October 1977), pp, 45- 60, and James A, Dalton and E, J, Ford, "Concentration and Professional Earnings in Manufacturing," Industrial and Labor Relations Review, Vol, 31, No, 3 (April 1978), pp, 379-84; Charles T. Haworth and Carol Jean Reuther, "Indus- trial Concentration and Interindustry Wage Deter- mination," Review of Economics and Statistics, Vol, 60, No, 1 (February 1978), pp, 85-95,

'Wallace Hendricks, "Unionism, Oligopoly, and Rigid Wages," Review of Economics and Statistics, Vol, 63, No, 2 (May 1981), pp, 198-205,

••Hendricks, "Regulation and Labor Earnings," and Weiss, "Concentration and Labor Earnings,"

'See, for example, Hendricks, "Regulation and Labor Earnings," and Thomas C, Moore, "The Bene- ficiaries of Trucking Regulation," Journal of Law and

estimates that telephone industry employ- ees in New York have substantially higher earnings than other comparable employ- ees in the state,^

These studies differ in the data bases, variable definitions, and time periods used, which complicates any direct comparison of their results. Yet, one factor is common to all these studies: they have focused almost exclusively on wages or nominal earnings. None, that is, directly considers the influ- ence of market structure on nonwage com- pensation such as fringe benefits. Employer contributions to employee pension and in- surance plans have grown rapidly in recent years and now average close to 10 percent of wages and salaries; in large manufacturing firms, the percentage is much higher.' Ideally, for estimating the social costs of market imperfections, one would want to know the impact of concentration and regulation on the total (wage plus non- wage) compensation of workers.

Three empirical issues are therefore con- sidered in this paper. First, we examine the impact of market structure on wage levels, using a data base that has some im- portant advantages over those used in previ- ous studies. Second, we analyze the relation- ship among concentration, regulation, and employer expenditures on private pension and insurance funds. And finally, we estimate the effect of market structure on labor turnover, holding constant the influence of concentration and regulation on wage rates.

Market Structure and Wages

Inadequate control for worker "quality" can bias the regression estimates of the net

Economics, Vol, 21, No, 2(October 1978), pp, 327-44, 'Ronald G, Ehrenberg, The Regulatory Process

and Labor Earnings (New York: Academic Press, 1979), pp, 69-90,

'For evidence, see James E, Long and Frank A, Scott, "The Income Tax and Nonwage Compensation," Review of Economics and Statistics, Vol, 64, No, 2 (May 1982), pp, 211-19, The 10 percent figure applies to voluntary rather than legally required employer contributions, such as OASDHL For a sur- vey of employee fringe benefits by industry, see Cham- ber of Commerce of the United States, Employee Benefits 1980 (Washington, D,C,: Chamber of Com- merce, 1981),

MARKET STRUCTURE AND LABOR COMPENSATION 241

impact of market structure on wages. Eor this reason, National Longitudinal Survey (NLS) data have some desirable properties for wage analyses. Eirst, the NLS contains data on actual firm-specific work experi- ence and prior employment history, along with other important productivity controls such as educational attainment, duration of job training, and health status. Second, the NLS reports individuals' hourly wage rates, which can be used as dependent vari- ables in wage equations. Other studies have resorted to industry averages or have prox- ied individual wage rates with annual earn- ings standardized for time worked.* Of course, the NLS data are not without draw- backs, namely, the small sample sizes and narrow age ranges.

We have taken the NLS sample of mature males (aged forty-five to fifty-nine in 1966) and estimated wage equations with the natural logarithm of the hourly wage in 1966 as the dependent variable. The inde- pendent variables are years of work experi- ence with the current employer, £XP; ex- perience squared, EXP^; years of formal schooling, EDUC; months of job-related training, TRAIN; a dummy variable equal to one if health limits working and equal to zero if otherwise, HLTH; a dummy variable equal to one if married with spouse present and zero if otherwise, MSP; a dummy vari- able equal to one if white and zero if other- wise, WHITE; dummy variables for region, SOUTH and WEST; the percent of work- ers in the industry covered by collective bar- gaining, [/;' the percent of the industry that

'Researchers using the tapes from the 1970 Census 1-1000 Public Use Sample measure annual hours worked as the product of current weekly hours of work and weeks worked for the prior year (with both hours and weeks measured as intervals), a practice tbat may bias estimates of tbe hourly wage rate.

'The unionization figures refer to the percent of all production workers in the industry covered by col- lective bargaining during the period 1968-72, as estimated from the Expenditures for Employee Com- pensation Survey administered by the U.S. Bureau of Labor Statistics. These data can be found in Richard B. Freeman and James L. Medoff, "New Estimates of Private Sector Unionism in the United States," Indus- trial and Labor Relations Review, Vol. 32, No. 2 (Jan- uary 1979), pp. 143-74.

Ideally, we would like to have information on the union status of each individual in the sample, but

is employed in plants having 500 or more employees, LGFIRM; the industry con- centration ratio, C;'" the interaction be- tween unionization and concentration, U'C; and dummy variables for regulated industries, REG, MAX, MIN, and F^R." Our sample is restricted to 1,514 wage and salary employees in private, regulated in- dustries and unregulated manufacturing. The regression results appear in Table 1.

The basic earnings model. Equation 1, explains over 40 percent of the variation in 1966 wages; and all of the control vari- ables for employee and industry character- istics are statistically significant with the expected signs. The coefficients of C and U'C are highly significant and their signs indicate that concentration raises wages although its impact diminishes with union- ization.'^ A unit increase in concentration raises wages by .345, .245, .145, and .045 percent at unionization levels of 20, 40, 60, and 80 percent, respectively." Thus, at a

sucb infonnation is not provided in the census data used by Weiss and Hendricks or in the NLS data for the year 1966. Such data are available in the 1971 NLS, bowever, and we have compared the wage equation parameters when the industry level of unionization is replaced by a dummy variable for union membersbip (more correctly, coverage of tbe individual by a collec- tive bargaining contract). These results indicate that the coefficients of the concentration and regulation variables do notvary witb the measure of unionization.

'"Concentration estimates for the industry categories in the NLS sample were kindly provided by Wallace Hendricks.

"M/IX denotes industries in which maximum prices are regulated (telephone, electricity, gas and steam, water); MIN denotes industries in which minimum prices are regulated and also entry is restricted (air transportation, trucking); and VAR denotes all otber regulated industries (railroad, bus, taxicab, ware- housing, water transportation, services incidental to transportation, radio and television, and sanitary). REG includes all of these regulated industries. See Hendricks, "Regulation and Labor Earnings," pp. 485-87, for a discussion of the primary regulatory authority and its powers in various industries.

'2A negative relationship between wages and [/• C was also found by Weiss, "Concentration and Labor Earnings"; Frederic M. Sherer, Industrial Market Structure and Economic Performance, 2d ed. (Chicago: Rand McNally, 1980), p. 359; and Hendricks, "Union- ism, Oligopoly, and Rigid Wages."

I'These estimates are obtained by evaluating the expression

3 WAGE/ dC = .00445 - .00005 U.

242 INDUSTRIAL AND LABOR RELATIONS REVIEW

Table L Regression Analyses of the Effect of Concentration and Regulation on Wage

Rates, 1966. (^-values in parentheses)

Variables

C

u.c

REG

MAX

MIN

VAR

U

LGFIRM

EXP

EXP^

MSP

EDUC

TRAIN

SOUTH

WEST

WHITE

Constant

R̂

F-level

N

Equation 1

,00445** (3,06)

- ,00005* (-2,41)

- ,01804 (-,60)

—

,00436** (3,39)

,33277* (2,36)

,01216** (4,72)

- ,00012 (-1,78)

,09119** (2,92)

,04275** (14,72)

,00204** (3,50)

-,14971** (-6,89)

,14187** (5,25)

,15275** (6,90)

4,57168** (50,50)

,4250

85,52

1514

Equation 2

,00492** (3,42)

-,00003 (-1,72)

—

-,00211 (-,04)

,14863** (3,24)

-,16692** (-4,27)

,00349** (2,62)

,2393 (1,71)

,01197** (4,69)

-,00011 (-1,57)

,09882** (3,19)

,04156** (14,45)

,00208** (3,62)

-,14113** (-6,55)

,14673** (5,49)

,14882** (6,77)

4,57487** (50,86)

,4395

78,29

1514

unionization level of 70 percent (the sample mean), wages in high-concentration indus- tries (C equals 70 percent) are 4.75 percent higher than wages in low-concentration in- dustries (C equals 20 percent). In contrast, the wage differential between high-union- ization {U equals 70 percent) and low- unionization industries {U equals 20 percent) is 8.8 percent at the mean con- centration level of 52 percent.

In Equation 2, the single dummy variable for industry regulation {REG) is replaced by separate variables for the specific type of regulation {MAX, MIN, and VAR). The estimated coefficients of the concentration terms now imply a much larger impact of market power on wages—about a 14 percent wage differential between high- and low- concentration industries. Controlling for the specific type of industry regulation is thus important in estimating the net impact of concentration on wages, as well as in estimating the union wage effect (which increases to 9.6 percent in Equation 2).

The finding that concentration raises wages is very robust and does not depend on cyclical factors such as high unemploy- ment and stable prices, as suggested by Haworth and Reuther, or on concentra- tion's being less than some critical value, as estimated by Dalton and Ford.'* This conclusion (as well as those of previous studies) should he tempered, however, hy the recognized weaknesses of concentration ratios, including the fact that the industry definitions used in their measurement are arbitrary and do not necessarily conform

*Significant at the ,05 level in a two-tailed test, **Significanl at the ,01 level in a two-tailed test.

''Haworth and Reuther, in "Industrial Concentra- tion and Interindustry Wage Determination," find the concentration impact on wages to be positive and significant only during a recessionary period. The year 1966 was one of low unemployment (2,9 percent). When our wage equations were estimated with NLS data for 1971, a year of high unemployment (5,0 per- cent), the impacts of high concentration and MIN- regulation on wages were similar to those reported in Table 1 and equally significant,

Dalton and Ford, in "Concentration and Labor Earnings," estimate that increases in concentration above the 50 percent level do not affect manufac- turing wages. We find that estimating the wage equa- tions with concentration-squared terms leaves the estimated wage differentials between high- and low- concentration industries virtually unchanged.

MARKET STRUCTURE AND LABOR COMPENSATION 243

to economic markets and also the fact that no information is provided about the mar- ket shares of individual firms.

The estimated coefficient of REG in Equation 1 is not statistically different from zero, wbicb indicates that wages do not differ between regulated industries and un- regulated manufacturing, once human capital, location, and industry character- istics are held constant. The estimates in Equation 2 substantiate Hendricks's hy- pothesis that a single dummy variable mis- represents the wage impact of industry regulation.'5 Regulation has no effect, for example, on wages in utility industries in which maximum prices {MAX) are set and entry is restricted, such as the telephone, electricity, gas and steam, and water indus- tries. This finding suggests that the impact on labor costs of "cost price-through" utility pricing schemes is offset by the in- flexibility of rate changes due to regulatory lags.'^

Setting minimum prices (MIN) and con- trolling entry into tbe market, on the other hand, raises wages by about 15 percent, suggesting tbat union power in air trans- portation and trucking is increased by regu- lations that reduce competition from non- union labor. Finally, wages are nearly 17 percent lower in railroads, local passenger transit, warehousing, radio and television, and other industries in wbich regulatory practices are more varied {VAR) and have not altered market structure and pricing policies in ways tbat might increase union power.

Assigning the degree and type of regula- tion to an industry also involves some arbitrary judgments that qualify the infer- ences drawn from the coefficients of the regulation variables. In addition, a problem in interpreting the coefficient of MIN is that air transportation and trucking are in- dustries characterized by spatial production limitations that act to curb the entry of new firms outside the union's jurisdiction. In long-line trucking, for example, a union firmly established in a few key cities within

a region will be protected from nonunion competition because new entrants must almost always operate in the same key cities. Levinson argues that these spatial limitations of production, rather than the control of entry by regulatory authorities, are responsible for the Teamsters' success in raising wages.^'

In this paper, we are unable to determine which entry barrier—spatial production limitations or regulation—is more im- portant in raising wages in air transporta- tion and trucking.'^ Nonetheless, we be- lieve it is important to include the regula- tion variables in the wage equation, if for no other reason than to serve as controls for estimating the wage impact of concentra- tion.

Since the earnings model includes a num- ber of control variables for worker quality and industry characteristics, the findings in Table 1 suggest tbat employees in high- concentration and M/N-regulation indus- tries receive economic rent in the form of wage premiums. Tbe correct measure for determining whether workers receive eco- nomic rent, however, is total labor compen- sation, wbicb equals wages and salaries plus sucb nonwage compensation as em- ployer-provided fringe benefits (pensions and health insurance), job satisfaction, and employment stability. The impact of mar- ket structure on nonwage compensation is

'^Hendricks, "Regulation and Labor Earnings." "For additional discussion, ̂ see Ehrenberg, The

Regulatory Process and Labor Earnings, pp. 11-12.

"Harold M. Levinson, "Unionism, Concentration, and Wage Changes: Toward A tJnified Theory," Industrial and Labor Relations Review, Vol. 20, No. 2 (January 1967), pp. 198- 205.

'*Both theory and empirical evidence suggest that the latter effect may dominate. Moore, "The Bene- ficiaries of Trucking Regulation," argues that the operating-ratio regulation in trucking tends to in- crease labor costs, aside from any additional union strength due to reduced competition from nonunion drivers. Empirically, the annual growth in hourly earnings in trucking has slowed considerably since 1977, relative to earnings growth during earlier years and relative to manufacturing wage increases since 1977. Entry into trucking has been less restricted since 1977, particularly for short-haul carriers, and de- regulation policies have been gaining support. Since there is little reason to expect that spatial production limitations have changed radically since 1977, the relatively slower wage growth in trucking may reflect the influence of greater competition.

244 INDUSTRIAL AND LABOR RELATIONS REVIEW

examined in the following sections of the paper.

Market Structure and Fringe Benefits

Data on one form of nonwage compensa- tion—voluntary employer contributions for pension and profit-sharing plans and for group health and life insurance—can be derived from U.S. Department of Commerce estimates of employee compensation by industry.'8 In estimating the relationship between market structure and employer expenditures on these fringe benefits, it is important to control for those determinants of fringe benefits that correlate with con- centration and regulation. The variables expected to affect fringes have already been discussed in detail elsewhere, so only a brief description of the model is presented here.^"

Fringe benefits should vary positively with total employee compensation because of tbe positive income elasticity of demand for fringes and the positive tax-rate elas- ticity that results from the preferential tax treatment of fringe benefits.^' Unionization is thought to have a positive effect on fringe benefits, since union leaders can inform their members of the tax advantages of fringe benefits and of the cost savings of group insurance and since employee prefer-

"U.S. Department of Commerce, Survey of Current Business, Vol. 59, No. 7 (July 1979), pp. 54-55. Vol- untary employer contributions for pensions and insur- ance were estimated by first subtracting wages and salaries from total employee compensation and then subtracting tbe legally required employer con- tributions to OASDHI. OASDHI payments by industry were calculated as (.mOb)(WAGE)(FTE) for WAGE > 517,700 or ($1,071) (FTE) for WAGE <' $17,700, where WA GE is average annual wage and salary earnings per full-time employee and FTE is total industry full-time employment.

2»See Robert G. Rice, "Skill, Earnings, and the Growth of Wage Supplements," American Economic Review, Vol. 56, No. 2 (May 1966), pp. 583-93; Bevars Mabry, "The Economics of Fringe Benefits," Industrial Relations, Vol. 12, No. 1 (February 1973), pp. 95- 106; Richard B. Freeman, "Tbe Effect of Unionism on Fringe Benefits," Industrial and Labor Relations Review, Vol. 34, No. 4 (July 1981), pp. 489-509; and Long and Scott, "The Income Tax and Nonwage Compensation."

2'See John S. Nolan, "Taxation of Fringe Benefits," National Tax Journal, Vol. 33, No. 3 (September 1977), pp. 359-68.

ences for fringe benefits can be directly transmitted to management by collective bargaining. Because of rate differentials be- tween group and individual health or life insurance policies, fringe benefits should vary directly with firm size, assuming the demand for group insurance is price-elastic. Employers may have a preference for certain types of fringe benefits that reduce labor turnover, sucb as nonvesting pension plans or paid vacations whose length increases with tenure. These kinds of fringe benefits are most likely to be offered when turnover costs are high, which is the case if much of employees' human capital is firm-specific.

The impact of concentration and regula- tion on fringe benefits is obtained by esti- mating the equation: (1) FBi =a +bC, +cRi + dXi , where FB, is annual voluntary employer expenditures in 1978 on pensions, profit sharing, and insurance, measured as dollars per full-time employee; C; is the industry concentration ratio; /?, consists of dummy variables indicating the existence and type of regulation, as defined earlier; and the vector X, includes the control variables TOTCOMP (average annual total com- pensation per employee), FIRMSZ (average firm size), KLRA TIO (the ratio of capital to labor), U (the percent of industry employees covered by collective bargaining), and U'C (an interaction between unionization and concentration). Data for these variables are available for twenty-eight two- and three- digit industries, which include manufac- turing, transportation, communication, and public utilities.^^ The ordinary least

'^Concentration ratios for two-digit manufactur- ing industries came from William G. Shepherd, The Economics of Industrial Organization (Englewood Cliffs, N.J.: Prentice-Hall, 1979), p. 202. Concentra- tion data for regulated industries were provided by Wallace Hendricks. For two industries (local and inter-urban passenger transit and trucking and ware- housing), it was necessary to average his data for cen- sus industries (using industry employment as weights) to match the Department of Commerce industry defi- nitions. Average firm size was measured as the num- ber of employees per establisbment in 1977, as re- ported in the U.S. Bureau of Census, County Business Pattems 1977, United States Summary, CBP-79-1

MARKET STRUCTURE AND LABOR COMPENSATION 245

square estimates of tbe model are reported in Table 2.2'

Table 2. Regression Analyses of tbe Effect of Concentration and Regulation on Employer Contributions for Pensions and

Group Insurance, 1978. (^-values in parentheses)

Variables

C

u«c

REG

MAX

MIN

VAR

U

TOTCOMP

FIRMSZ

KLRATIO

Constanl

R̂

F-level

N

Equation 1

-31.189* (-2.55)

.854** (3.90)

- 626.523 (-1.67)

—

—

—

-11.516 (-.81)

.085* (2.44)

4.205 (1.92)

129.576** (5.72)

317.782 (.51)

.936

41.51

28

Equation 2

- 19.804 (-1.48)

.653* (2.64)

—

-230.861 (-.37)

-401.641 (-1.03)

-986.703* (-2.09)

-.086 (-.01)

.072 (1.99)

3.904 (1.75)

124.217** (5.24)

21.618 (.03)

.946

35.26

28

*Significant at the .05 level in a two-tailed test. **Significatit at the .01 level in a two-tailed test.

(Washington, D.C: GPO, 1979), pp. 1-2. TOT- COMP and KLRATIO (which is proxied by the ratio of corporate capital consumption allowances to in- dustry employment) came from the LI.S. Department of Gommerce. Survey of Current Business, Vol. 59, No. 7 (July 1979), pp. 54, 55, and 60. tjnionizationdata are from Freeman and Medoff, "New Estimates of Private Sector Llnionism in the LInited States."

"Freeman, in "The Effect of Unionism on Fringe Benefits," p. 498, has pointed out that Equation

Tbe concentration terms are bigbly sig- nificant in Equation 1, wbich includes only tbe single regulation variable. Tbe co- efficients of C and U'C indicate tbat product market power raises fringe benefits, once unionization exceeds 37 percent.̂ ^ Tbe marginal effects of concentration on fringe benefits are - 14, +3, +20, and +37 dollars at unionization levels of 20, 40, 60, and 80 percent, respectively. At a unionization level of 44 percent (tbe sample mean), tberefore, pension and insurance contributions are |320 bigber in bigb-concentration (C equals 70 percent) tban in low-concentra- tion (C equals 20 percent) industries. Wben tbe tbree regulation variables are included in place of REG, as in Equation 2, tbe abso- lute values of tbe concentration coefficients are diminisbed; but tbe U'C term,bow- ever, remains positive and bigbly signifi- cant. Tbe estimates indicate tbat market power raises fringe benefits at unionization levels above 31 percent; at tbe mean level, employer contributions for pensions and in- surance are $425 bigber in bigb- tban in low-concentration industries. Tbe impact of concentration on fringe benefits is tbus mucb smaller tban tbe $1,575 effect due to unionization.^*

1 is subject to simultaneity bias since TOTCOMP includes voluntary employer contributions for fringe benefits along with wages and salaries plus legally required supplements. Gorrecting for this problem, as he suggests, does not appreciably change our estimates of the impact of concentration and regulation on fringe benefits. Gonsequently, the corrected estimates are not reported in the text.

2*Note that the positive and significant coefficient of U'C in Table 2 contrasts with that observed in Table I and in other wage studies. Together, these re- sults indicate that in highly concentrated industries, increased unionization raises the sbare of total com- pensation allocated to fringe benefits—a finding consistent with the theory and evidence presented by Freeman, "Tbe Effect of Unionism on Fringe Bene- fits." The tax savings from fringe benefits increase witb concentration; and unions can inform members of the advantages of fringes and make worker prefer- ences concerning the compensation mix known to management.

2*Tbis estimate assumes a concentration level of 49 percent (the sample mean) and a unionization differ- ential of 50 percent (for example, U equals 70 percent relative to U equals 20 percent).

246 INDUSTRIAL AND LABOR RELATIONS REVIEW

When measured by a single variable, industry regulation reduces fringe benefits by $627, although the statistical significance of REG is marginal. Equation 2 suggests that the negative effect of regulation on fringes should not be attributed to all types of regulation. Pension and insurance con- tributions are relatively lower (by |987) only under VAR-type regulation, in which authorities do not consistently restrict entry or set minimum and maximum prices.

We know of no previous studies that have related fringe benefit expenditures to con- centration and regulation; consequently, there are no benchmarks against which our estimates can be compared. Moreover, since we use aggregate industry data, our con- clusions should be regarded as tentative. In the next section we propose an alternative test to determine the influence of concen- tration and regulation on nonwage com- pensation.

Market Structure and Labor Turnover

An alternative method for estimating the effect of market structure on nonwage com- pensation is derived from the assumption that, other things equal, individuals will quit their present jobs if they perceive that an alternative job offers a net advantage. If concentration or regulation generates positive (negative) economic rent in terms of higher (lower) wage or nonwage returns, workers in concentrated or regulated indus- tries will therefore be less (more) likely to quit and move to other jobs than workers in competitive industries. Interindustry turn- over differentials not resulting from wage differences among concentrated, regulated, and competitive industries can thus be as- sumed to result from differences in nonwage compensation.26

2'A number of recent studies have used turnover data to analyze nonwage differentials by race and union status. For example, see Duane E, Leigh, "Unions and Nonwage Racial Discrimination," Industrial and Labor Relations Review, Vol, 32, No, 4 (July 1979), pp, 439-50; and Richard B, Freeman, "The Effects of Unionism on Worker Attachment to Firms," Journal of Labor Research, Vol, 1, No, 1 (Spring 1980), pp, 29-62, James E, Long has used quit behavior to estimate whether government workers

The impact of market structure on labor turnover has not been extensively re- searched, Hendricks indicates that regulated industries offer more stable employment than manufacturing, and Ehrenberg reports that quit rates are much lower for New York Telephone employees than for workers in manufacturing industries. Nevertheless, neither conclusion is supported by multi- variate analysis that controls for nonregula- tion variables that influence turnover.^'

Other empirical studies have found a significant, negative relationship between concentration and the industry's average annual quit rate. Burton and Parker suggest that quits are inversely related to concentra- tion because (1) the relatively high profits of concentrated industries may allow those firms to use labor in an inefficient (in other words, X-efficient) manner, which may be a characteristic of employment that is attrac- tive to workers; (2) the fewness of firms raises the probability of collusive no-raiding agreements as well as reduces alternatives for intra-industry mobility; and (3) fringe benefits may be higher in concentrated industries.28 Parsons argues that entrance into high-concentration, high-wage indus- tries generally requires investments by the worker, which may diminish the tendency to quit.29

These concentration-turnover findings came from regression models of the form

(2) QR= )So+ fi,W+ I3,A + ^ , 2 ,

where QR is the industry quit rate, W is the current wage, A represents alternative job possibilities, and Z is a vector of other deter-

are overpaid relative to private sector employees, in "Are Government Workers Overpaid? Alternative Evidence," Journal of Human Resources, Vol, 17, No, 1 (Winter 1982), pp, 123-31,

"Hendricks, "Regulation and Labor Earnings," p, 492; Ehrenberg, The Regulatory Process and Labor Earnings, pp, 108- 110,

28John F, Burton and John E, Parker, "Interindustry Variations in Voluntary Labor Mobility," Industrial and Labor Relations Review, Vol, 22, No, 2 (January 1969), pp, 199-216,

"Donald O, Parsons, "Specific Human Capital: An Application to Quit Rates and Layoff Rates," Journal of Political Economy, Vol, 80, No, 5 (November/ December 1972), pp, 1120-43,

MARKET STRUCTURE AND LABOR COMPENSATION 247

minants of the quit rate (such as pensions, race, gender, and concentration). This specification can be criticized on two grounds. First, using industry aggregates as the units of observation introduces a simul- taneous-equation bias because it is ambigu- ous whether Equation 2 is a supply or de- mand relationship,'" With individual work- ers as the unit of observation, however. Equation 2 can be treated as a supply rela- tionship, because the compensation levels facing individual workers are exogenous to them, A second statistical problem is the inclusion of wages as an explanatory vari- able, Flanagan argues that this introduces multicoUinearity (since wages may be re- lated to race, gender, concentration, and other explanatory variables) and a simul- taneity problem (employers may pay lower wages to workers more likely to quit, for example)."

With these problems in mind, we esti- mate the relationship between market structure and voluntary turnover using the following model specified by Flanagan:'^

(3) a = /8o+ J8,(M .̂- - M )̂+ /̂ 2 2 , .

where Q, is the quit behavior of the indi- vidual, Wi is the individual's current wage, W is the mean market wage for workers with similar human capital and wage-related characteristics, and Z, contains other deter- minants of quit behavior. The probability of quitting the current employer is assumed, other things equal, to increase as the differ- ence between the actual current wage and the "potential" market wage (hereafter, RESID) decreases. For our analysis, the vector Z includes dummy variables indicat- ing whether the current employer is in a regulated industry {REG, MAX, MIN, or VAR); the concentration ratio (C) in the individual's current (three-digit census) industry; unionization {UNION) in the

'"For further discussion, see Schiller and Weiss, "The Impact of Private Pensions on Firm Attach- ment," p, 379,

"Robert J, Flanagan, "Discrimination Theory, Labor Turnover, and Racial Unemployment Differ- entials," Journal of Human Resources, Vol, 13, No, 2 (Spring 1978), pp, 187-207,

"Ibid,

current industry; years of work experience with the current employer {TENURE); and dummy variables for married with spouse present {MSP) and race {WHITE). If high concentration and regulation gen- erate positive (negative) economic rents, the coefficients of C and the regulation vari- ables are expected to be negative (positive) and statistically significant. Since the effects of concentration and regulation on the cur- rent wage will be held constant by the RESID term, negative coefficients on the market-structure variables will indicate that concentration and regulation raise non- wage compensation. From previous studies, we expect the quit probability to vary in- versely with TENURE, MSP, and UNION and directly with WHITE.

Turnover data for individual workers and for the explanatory variables required to estimate Equation 3 were obtained from the National Longitudinal Surveys (NLS) of middle-aged men (aged forty-five to fifty- nine)," In the NLS, individuals are asked to compare their employers at two consecu- tive survey dates and to indicate whether any change in employers is due to a volun- tary quit or to layoff or dismissal. In the sample of middle-aged males, comparison data on employers are reported for the years 1966-67, 1967-69, 1969-71, and 1971- 73. Labor turnover can thus be measured as a binary variable taking a value of one if a worker quits his or her job in one year and has a different employer in a later year, and taking value of zero if that worker does not change or involuntarily changes employers. Combining data for the four intervals pro- vides a sample of 4,408 observations, in- cluding 177 instances of job quitting.

The variable RESID is calculated by subtracting individual workers' "potential" (predicted) wage from their actual wage at the beginning of a time interval. The pre- dicted wage is based on the coefficients from a market wage equation, using the natural logarithm of the hourly wage as the dependent variable and including the following explanatory variables: years of

"Data sources for the concentration and unioniza- tion variables may be found in footnotes 10 and 11 above.

248 INDUSTRIAL AND LABOR RELATIONS REVIEW

schooling; tenure and its square; unioniza- tion; months of job-related training; and dummy variables for health, marital status, race, and geographic region. Since the turnover observations are pooled by time intervals, four separate wage equations are estimated. The wage equation samples are restricted to private wage and salary em- ployees in manufacturing and regulated industries.

Our estimates of the impact of market structure on voluntary turnover are reported in Table 3." Since the dependent variable is dichotomous, the quit equations are esti- mated with probit analysis. The coefficient of concentration is negative and highly significant in both equations, indicating that the probability of quitting decreases as concentration rises. In contrast, the co- efficients of regulation (REG) are positive, which indicates that individuals employed in regulated industries are more likely to voluntarily change employers than workers in unregulated manufacturing. Equation 2 reveals, however, that the positive impact of regulation on turnover is statistically sig- nificant only for MAX- and VAR-iype regulation.

The coefficients of the remaining vari- ables carry the predicted signs, and RESID, TENURE, and UNION are highly signifi- cant. Since including RESID in the quit equation holds constant any "wage differ- ential" incentive for changing employers, the negative coefficient of C in Table 3 sug- gests that quitting decreases as concentra- tion rises because nonwage compensation is greater in highly concentrated industries than in competitive industries. This find- ing is consistent with the estimates in Table 2, which indicate that employer expendi- tures on pensions and insurance increase with concentration. The results in Tables 2 and 3 suggest that workers in VAR-regula-

. Probit Analyses of the Effect of Con- centration and Regulation on Voluntary

Labor Turnover, 1966-73. (asymptotic lvalues in parentheses)

'*To determine whether our estimates are biased by pooling different time intervals, we included the national rate of unemployment during the first year of each interval as a crude measure of labor market conditions. This variable was never statistically sig- nificant and its inclusion had no effect on the market structure coefficients; it was therefore dropped from the equation.

Variables

C

REG

MAX

MIN

VAR

RESID

TENURE

UNION

MSP

WHITE

Constant

- 2 X log likelihood

ratio

N

Equation I

- .0068** (-3.13)

.3337** (3.11)

-.2014* (-2.10)

- .0489** ( - 10.64)

- .0059** (-2.67)

-.1484 (-1.24)

.0973 (1.13)

-.4761** (-2.74)

231.36

4408

Equolion 2

-.0111** (-3.64)

—

.7398** (2.57)

.0721 (.42)

.5687** (3.37)

-.1785 (-1.83)

- .0493** (-10.64)

- .0048* (-2.06)

-.1545 (-1.24)

.1250 (1.43)

- ,3767* ( - 2.08)

235.95

4408

*Significant at the .05 level in a two-tailed test. **Significant at the .01 level in a two-tailed test.

tion industries are relatively more likely to quit because they receive lower fringe bene- fits than comparable workers in other regu- lated industries and manufacturing. The insignificance of the MIN terms in Tables 2 and 3 implies that setting minimum prices and restricting market entry raises wages but does not affect other kinds of labor com- pensation. The coefficients oiMAX suggest that regulation in the utility industries re- duces certain kinds of nonwage compensa- tion (other than pension and insurance

MARKET STRUCTURE AND LABOR COMPENSATION 249

payments) and, other things equal, results in higher labor turnover.'^

Summary and Implications Previous studies of the effect of con-

centration and industry regulation on labor compensation have focused predominantly on wages or earnings while ignoring fringe benefits and other kinds of nonwage com- pensation. Consequently, the implications of those studies regarding the impact of market structure on total labor earnings must be considered tentative. In this study, we have empirically examined the effect of concentration and regulation on labor compensation by estimating the deter- minants of hourly wages, fringe benefits in the form of employer contributions to pen- sion and insurance plans, and voluntary labor turnover. The regression results reveal a positive and highly significant relation- ship between concentration and both wages and fringes, and a significant negative rela- tionship between concentration and job quitting. These findings are highly con- sistent with the hypothesis that, other things equal, labor compensation is relatively higher in concentrated industries than in more competitive ones. Thus, it is rather ironic that certain members of Congress with close ties to organized labor, notably Senator Edward M. Kennedy, have sup- ported legislation intended to reduce in- dustrial concentration in the United States.56

'^These kinds of nonwage compensation could include vacation and sick-leave pay, bonuses, over- time premiums, contributions to savings plans, holiday funds, employee education funds, job security, and job satisfaction. For information on the various kinds of fringe benefits, see Chamber of Commerce of the United States, Employee Benefits 1980; and Freeman, "The Effect of Unionism on Fringe Bene- fits." Freeman's data (p. 496) indicate that employer contributions for pensions and insurance make up only about 36 percent of measurable voluntary fringe benefits.

"Opening statement of Edward M. Kennedy in U.S. Congress, Senate, Committeeon the Judiciary, Mergers and Economic Concentration, Part 1, Hearings before the Subcommittee on Antitrust, Monopoly, and Business Rights (Washington, D.C: GPO, 1979). For additional discussion, see A. F. Ehrbar, "Bigness

Industry regulations that set minimum prices and restrict market entry, as in air- lines and trucking, are estimated to raise labor compensation by means of increasing hourly wages. This result implies that the opposition to deregulation on the part of the Air Line Pilots Association and the Teamsters has been well founded. The rela- tively higher incidence of job quitting among workers in the utility industries (telephone, electricity, gas, and water) suggests that profit regulation through the imposition of maximum prices reduces labor compensation. Stronger evidence that certain kinds of regulation may decrease labor earnings is contained in the finding that in the remaining regulated industries, both wages and fringe benefits are relatively lower, wbereas voluntary labor turnover is relatively higher.

The implications of this study for public policies designed to deconcentrate or de- regulate industries depend on two issues. First, has labor quality adequately been beld constant in the empirical analyses, so that compensation differentials reflect eco- nomic rent rather than payment for some productivity factor? Second, do the empiri- cal findings for mature men apply to other groups of workers as well? If we assume that the relatively higher labor compensa- tion in concentrated and M/N-regulated industries constitutes economic rent en- joyed by the average worker in those indus- tries, then our findings imply that decon- centration and deregulation would reduce unit costs and lead to more efficient re- source allocation throughout the economy. The income losses (reduced wages or fringes) incurred by the workers now em- ployed in the affected industries would be more than offset by gains to consumers and other workers.

The assumption that relatively higher labor compensation (or profits) represents economic rent seems more valid in the case of industries in which entry has been arti- ficially restricted by regulation tban in

Becomes the Target of the Trustbusters," Fortune, Vol. 99, No. 6 (March 26, 1979), pp. 34- 39.

250 INDUSTRIAL AND LABOR RELATIONS REVIEW

highly concentrated, unregulated manu- facturing. Studies by Brozen, Demsetz, and others suggest that the relatively high profit rates in concentrated manufacturing indus- tries reflect the greater efficiency and tech- nological advancement of dominant firms." Kendrick and Grossman find that the slowdown in manufacturing produc- tivity growth since the mid-1960s is sig-

"Yale Brozen, "The Significance of Profit Data for Antitrust Policy," in J. Fred Weston and Sam Peltz- man, eds.. Public Policy Toward Mergers (Pacific Palisades, Calif.: Goodyear, 1969), pp. 110-27; Harold Demsetz, The Market Concentration Doctrine (Washington, D.C: American Enterprise Institute, 1973).

nificantly less in more concentrated indus- tries.'^ The relatively higher wages and fringe benefits observed in concentrated industries may thus be compensation for tbe higher skill levels of workers. Finally, to the extent that labor turnover raises costs and retards productivity growth, the finding that concentration reduces job quitting is certainly not supportive of public poli- cies to reduce industrial concentration.

"John W. Kendrick and Elliot S. Grossman, Pro- ductivity in the United States: Trends and Cycles (Baltimore, Md.: Johns Hopkins University Press, 1980), pp. 106-11.