""7142' - .1. . 'WU‘ 11 11-151? 111.111.11.11.1-'*-1.'. LEI U" 1‘3" I'&1.-.., 1‘. E.- %”s 1":‘:;’ 1“th _‘ fl" '.|." “WOW '11". .‘1 I. ‘ 145;}..‘u1' I ”I :I"... '(r l 2*. J;- ’5" ' . ' .I.- . ., Q." - . |.I\'|" \ I ' 4 ‘ ‘I *I":‘ .“ “'3 if“ “0"1‘11' 'll ‘I‘ttz I4” 1 cf: '1" I .' . “I \ " ‘5‘)" ‘:"::I 1‘ ' ’: I .. 11' ,:\~.‘. 'I'\ ~— I. .1. 1,2“ " - 11.111 "'-' I:II1'- 1 {'I'Jl‘i‘." “1: @21- (‘1’. .18.; 1...: .11‘)'1":"1~.~Lt~'.;.1‘""flf "311' 1" {fig-,1; ~41: 'l‘j‘E; $2.: ' ‘ C," 1:»; 6.5;“: f91346" “‘4 (I'M mfiIEJI I.‘ ”Fag. yit-é. 1.I"‘4:§”.:k'y_-':~‘. "5:"; f1 ‘ 1'31“. _‘. 'géfi. 1'51"]:2’1k vr .4 - w. '1. 4..."; :1 . "1r..4’_.'". ’3‘ . -- 1‘3 . fl-‘QV. by!“ '11'._‘\l111\,.1)|:\: 'v'fiifi' . ' ‘ ‘ . . . .- _ "1'11‘ fi'fifir. 45:22“ . " ‘9‘:ng ..‘, .358“; 5365:511in L 0 . 111. "." " A * - ‘ 'n 4'- ' ‘ "u' M‘ . "“3? {5121.911 1 ' _ - . '13:; W 9-13; . . .. 3.335%} 3“ 1514;111:321 : ' i . “ . .. . 432.121 ’~.~ v «11‘ ‘ 1 1: 1 . ~ 11111" :1 w v . 3.1.1. .. "f - "v m1 '«11‘1 ' ‘ ‘I' .-'~‘ i... .-.‘_\o- 4 7’ ~ 1 ‘. :m‘w ‘1' ~4- .‘4 f. ~ 'Slm-Jr ~‘ 'Agv a "i H, ‘ 3“}k‘1 1-1345)" 3 ~, 23.; ‘.- . . .. 7 ~th ‘ '1'1 .5.»« . --(‘4’< 1 ‘l C ‘4 L 1 . -‘; .\._*- ‘n . ... '4'; ~ f.‘ .- -5. -1 ’Pr 1 - n- . R ' r . n.’ ‘ 1' ---- "- ‘J—o ‘ ‘ 153:3 4:; .J-‘n-m ".1 1' no 111“ ._.~;- :1 a .' ._. .< .' 5-1. , 6—;V-4 4 2"3‘ M 1.111 I -? .s.” 'r- . - 05:: "1:143:13? . _ :L‘.‘. _._ - ' ‘ ' . '1. A y _ .3 4. ‘51. ‘ '4 - 1- - _ ‘N-‘.‘.'-‘.:i .. fl ._ a; 1.22:._J.fl . : LC'-‘ ' ;‘ 5.3“:‘V. ,‘i’w .. 311.1, ’4 :11 u I 7“. E”. Lvlm llLllllLflLlL LILLILLILILILllLLlLLl fifi'E-S's LIBRARY Michigan Saba University This is to certify that the thesis entitled UNION EFFECTS ON THE SIZE DISTRIBUTION OF EARNINGS presented by Nguyen Thanh Quan has been accepted towards fulfillment of the requirements for Ph . D. degree in M55 WMfl/W Major professor Dateméflgg 0-7 639 4 A2 M833 WC??? JAN obi-20h? -: .: © 1978 NGUYEN THANH QUAN ALL RI GHTS RESERVED UNION EFFECTS ON THE SIZE DISTRIBUTION OF EARNINGS By Nguyen Thanh Quan A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Economics 1978 ABSTRACT UNION EFFECTS ON THE SIZE DISTRIBUTIONS OF EARNINGS By Nguyen Thanh Quan There has been much work on unions' effects on relative wages, but little is said on their effects on the functional dis- tribution and practically nothing on their effects on the size distribution. This study is an attempt to fill this void, for several sets of microdata, to estimate the incidence of unionism and its relative wage effect by earnings class, make assumptions about the unions' effect on the incidence of periods of unemployment and about what they do to relative factor shares, then proceeds to estimate what they do to the size distribution. The functional equation to be estimated by ordinary least squares is of the form Ln E.. = f (UNIONij, X.. 13 Z..) 13’ 13 Nguyen T. Quan where: Eij is the hourly earnings UNION ij is the membership status X is the vector of demographic, industry and occupation 13 characteristics 2.. is the vector of indicators of human capital for the ith worker in the jth earnings class Since the observed Ej is a weighted average of union and nonunion earnings, the latter value can be determined as n A . = . + N N. - EJ EJ/(l U IO J m) where: m is the estimated differential due to unionism. The mean union earnings is easily derived. For each earnings class the frequency distribution of both types of earnings can be determined and a Lorenz curve generated by a Pareto distribution of the form Y = {l - (l-x)OL}]/B 0 < a, 8 §_l is fitted over the whole range. From the estimated Lorenz curve the Gini index, as a measure of inequality, is derived for both union and nonunion earnings. The data used are the Survey of Working Conditions, and the National Longitudinal Survey--Mature Men. The evidence suggests that the distribution of union earnings is relatively more equal than that of nonunion over a number of years. To My Parents ii ACKNOWLEDGMENTS I wish to thank the members of my dissertation committee, Professors Daniel S. Hamermesh, Daniel B. Suits, Bruce T. Allen, and Anthony Y. C. Koo for their guidance and helpful suggestions during the course of this research. I am especially indebted to my committee chairman, Professor Daniel S. Hamermesh for the generous manner in which he shared his time and intellectual capital with me. Appreciation is also extended to Harriet Dhanak, Director of Political Data Archives at Michigan State University. Her patience and good humor have made my struggle with data tapes most bearable. Last, but not least, I would like to thank Pat Trommater, Denise Amburgey, and Nancy Heath for typing the many drafts. iii TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES - Chapter I. II. III. IV. INTRODUCTION THEORETICAL FRAMEWORK NNNNN m-DWN—J Introduction . . . Unionism and Relative Wages . Determinants of Competitive Wages . Earnings Shares . Measures of Inequality and Functional Forms for the Lorenz Curve . THE NATIONAL SURVEY OF WORKING CONDITIONS . 3.1 3.2 3. 3 3.4 3. 5 3. 6 3. 7 THE WOR 4 4 4 l 2 3 Introduction . . Explanatory Variables . . . 3. 2. 1 Population Characteristics . 3.2.2 Regional Characteristics 3.2.3 Occupational Characteristics 3. 2. 4 Human Capital . . 3.2.5 Working Conditions Estimation . . Empirical Results . . Determination of Union and Nonunion Earnings Distributions . Lorenz Curves and Measures of Inequality Conclusion NATIONAL LONGITUDINAL SURVEY: OLDER MALE KERS . . . . . . . . Introduction . Characteristics of Older Workers Empirical Results iv Page vii Chapter Page 4.4 Determination of Union and Nonunion Earnings Distributions . . . . 74 4.5 Lorenz Curves and Measures of Inequality. . . 79 4.6 Conclusion . . . . . . . . . . . . . 88 V. THE NATIONAL LONGITUDINAL SURVEY: DISAGGREGATES ESTIMATES . . . . . . . . . . . . . . . 91 5.1 Introduction . . . . . . 91 5.2 Union- Nonunion Earnings Differential . . . . 92 5.3 Union and Nonunion Relative Shares . . . . . 97 5.4 Lorenz Curves and Inequality . . . . . . . lO6 5. 5 Conclusion . . . . . . . . . . . . . lZl VI. SYNTHESIS AND RECONCILIATION OF TWO SAMPLES . . . 125 6. l Introduction . . . . . . l25 6. 2 The Survey of Working Conditions: Older Workers . . . . . . . . . l27 6. 3 Conclusion . . . . . . . . . . . . . l33 APPENDIX . . . . . . . . . . . . . . . . . l38 REFERENCES LIST OF TABLES Hourly Earnings Regressions, Working Conditions Data, 1969 . . . . Incidence of Unionism by Deciles Among Private Workers, Survey of Working Conditions, 1969 Relative Shares of Union and Nonunion Earnings by Deciles, Survey of Working Conditions, 1969 Occupational Incidence by Deciles, Survey of Working Conditions, 1969 . . . . . . . Estimates of Lorenz Curves and Gini Coefficients, Survey of Working Conditions, 1969 Hourly Earnings Regressions, National Longitudinal Survey, Men . . . . Incidence of Unionism by Deciles Among Private and Public Workers, National Longitudinal Survey, 1969 and 1971 . . . . Relative Shares of Union and Nonunion Earnings by Deciles, National Longitudinal Survey, 1969 and 1971 . . . Difference in Relative Shares of Union and Nonunion Earnings by Deciles, National Longitudinal Survey, 1969 and 1971 . . . Occupational Incidence by Deciles, National Longi- tudinal Survey, 1969 Occupational Incidence by Deciles, National Longi- tudinal Survey, 1971 . . . Estimates of Lorenz Curves and Gini Coefficients, National Longitudinal Survey, 1969 and 1971 Union-Nonunion Earnings Differentials in Manufactur- ing and Nonmanufacturing, National Longtidinal Survey, 1969 and 1971 . . vi Page 34 42 43 46 48 61 75 78 83 84 85 87 93 Table 5.2 A.2 A.3 Page Incidence of Unionism by Deciles in Manufacturing and Nonmanufacturing, National Longitudinal Survey, 1969 . 99 Incidence of Unionism by Deciles in Manufacturing and Nonmanufacturing, National Longitudinal Survey, 1971 . 100 Relative Shares of Union and Nonunion Earnings by Deciles in Manufacturing and Nonmanufacturing, National Longitudinal Survey, 1969 . . . . . . . 103 Relative Shares of Union and Nonunion Earnings by Deciles in Manufacturing and Nonmanufacturing, National Longitudinal Survey, 1971 . . . . . . . 104 GINI Coefficients for Union and Nonunion Earnings in Manufacturing and Nonmanufacturing, National Longi- tudinal Survey, 1969 and 1971 . . . . . . . . 119 Union-Nonunion Earnings Differentials for Three Age Groups in the Survey of Working Conditions, 1969 . . 127 Incidence of Unionism by Deciles Aming Workers with Age 35 Years and Above, Survey of Working Conditions, 1969 . . . . . . . . . . . . . . . . . 129 Relative Shares of Union and Nonunion Earnings by Deciles for the Sample of Workers with Age 35 Years and Above . . . . . . . . . . . . . . . 131 Estimates of Lorenz Curves and Gini Coefficients for the Sample of Workers with Age 35 Years and Above, Survey of Working Conditions, 1969.. . . . . 134 Hourly Earnings Regressions, National Longitudinal Survey, Disaggregated Data, 1969 . . . . . . . 139 Hourly Earnings Regressions, National Longitudinal Survey, Disaggregated Data, 1971 . . . . . . . 142 Hourly Earnings Regressions, Various Age Groups. Working Conditions Data, 1969 . . . . , . 9 , 145 vii Figure 3.1 LIST OF FIGURES Lorenz Curves for Union and Nonunion Earnings, Survey of Working Conditions, 1969 . Lorenz Curves for Union and Nonunion Earnings, National Longitudinal Survey, 1969 Lorenz Curves for Union and Nonunion Earnings, National LongitudinaT Survey, 1971 Lorenz Curves for Union and Nonunion Earnings, Manufacturing, National Longitudinal Survey, 1969 Lorenz Curves for Union and Nonunion Earnings, Manufacturing, National Longitudinal Survey, 1971 Lorenz Curves for Union and Nonunion Earnings, Blue Collar Workers in Manufacturing, National Longitudi- nal Survey, 1969 . . . Lorenz Curves for Union and Nonunion Earnings, Blue Collar Workers in Manufacturing, National Longitudi- nal Survey, 1971 . . Lorenz Curves for Union and Nonunion Earnings, Nonmanufacturing, National Longitudinal Survey, 1969 . . . Lorenz Curves for Union and Nonunion Earnings, Nonmanufacturing, National Longitudinal Survey, 1971 . . . Lorenz Curves for Union and Nonunion Earnings, Blue Collar Workers in Nonmanufacturing, National Longitu- dinal Survey, 1969 . . . . . Lorenz Curves for Union and Nonunion Earnings, Blue Collar Workers in Nonmanufacturing, National Longitudinal Survey, 1971 viii Page 45 8O 81 108 109 110 111 112 113 114 115 Figure Page 5.9 Lorenz Curves for Union and Nonunion Earnings, White Collar Workers in Nonmanufacturing, National Longi- tudinal Survey, 1969 . . . . . . . . . . . 116 5.10 Lorenz Curves for Union and Nonunion Earnings, White Collar Workers in Nonmanufacturing, National Longi- tudinal Survey, 1971 . . . . . . . . . . . 117 6.1 Lorenz Curves for Union and Nonunion Earnings, 35 Years Old and Above, Survey of Working Conditions, 1969 . . . . . . . . . . . . . . . . 132 ix CHAPTER I INTRODUCTION Is a union a latter-day version of Robin-Hood's merry men striking out from local headquarters to rescue financially deprived workers? In other words, can trade unions increase the workers' share in the distribution of income at the expense of the receivers of rent, interest, and profits and then redistribute this share equitably to people at different earnings levels? One of the basic objectives of unions is to raise relative earnings of their members. In doing so they may have gained at the expense of nonunion workers, whose money wages are depressed as labor 1 is reallocated, rather than capital, and they may have gained at the expense of weak unions. The size of this gain is difficult to ascertain, as Albert Rees states succinctly We tend to overemphasize the role of unions, both in . . . their own industries and . . . the economy as a whole. . . . The other two thirds may have their wages and salaries influ- enced by what the unions do, but I feel there are very strong independent forces on the demand side that govern their rates of pay. . . . Even in the . . . unionized [one third] there are some very weak. . . unions that have had very little to do with the wages of their members. . In a series of rough guesses, I would say perhaps a third of the trade unions have raised the wages of their members by 15 percent to 20 percent above what they might be in a non- union situation; another third by perhaps 5 percent to 10 percent, and the remaining third not at all. . . . The high figures tend to be found, not in periods of inflation, but in periods of prOSperity combined with stable prices. . . . In [an inflationary] period like 1946-1948, for example, the union pe0p1e may even lag behind simply because of the rigidities involved in the collective bargaining process.2 Why is it that the magnitude of the effects of a successful union seems to be in the neighborhood of 10 to 20 percent rather than say, 1 or 2 percent or even 100 or 200 percent? Since the basic goal of unions is to raise wage rates, what kinds of forces limit the suc- cess of unions that temporarily achieve very large gains and drive them back toward a more usual impact on relative wages? There are economic restraints which mitigate the impact of unionism on relative wages. The magnitude of the union effect is based on the elasticity of the demand for union labor, more Spe- cifically on the elasticity of a Marshallian derived demand. The demand is more inelastic the more essential is union labor to the production of the final product, the more inelastic the demand for the final product, the smaller the ratio of the cost of union labor to the total cost of the product, and the more inelastic the supply of the other factors of production. In a smaller framework, the impact of the union on relative wages is the extent to which a union raises the wages of its members and the other workers for whom it bargains above the wages of com- parable but unorganized workers. However, there are problems in comparing union and nonunion wages. If we compare wages in union and nonunion plants in the same industry and labor market, we may observe only small differences even where the union is effective, for nonunion employers will often be forced to raise wages if they want to prevent the unionization of their workers. Moreover, a union wage increase will not necessarily affect the wages of nonunion workers in other industries; the effect can be in either direction. In one case the higher union wages in one industry may stimulate an increase in union wages in other industries which in turn spill over onto nonunion wages in those industries. In other cases, the effect operates through the labor market. The higher wages in the union sector will tend to check the growth of employment in that sector. This will increase the supply of labor to the nonunion sector and tend to check increases in nonunion wages.3 0n the supply side, there are empirical limits which prevent the clear distinction between the effect of the union from the effect of other forces that contribute to wage differentials in the absence of unions.4 For one thing, the wages of two comparable but unorganized workers having the same occupations will not usually be the same if both live in two different cities. There will almost always be some differential whose size is determined by such factors as the age and skill of the workers, working conditions, the size of the cities, the local market conditions and the area in which the cities are located. In view of the above problems, estimates of union impact on wages cannot be regarded as exact but should give us some rough order of magnitude of the union effect. Some studies have investi- gated a large number of industries simultaneously, classifying them according to the degree of unionization and industrial concentra- tion.5 Other studies have examined intensively the effects of unions on earnings in a single industry.6 More recently, with the avail- ability of microeconomic data, some studies have incorporated union status as a personal characteristic into the wage equation.7 Furthermore, there are specialized studies on the impact of public 8 and public school teachers.9 unions on the wages of firefighters In summary, all the studies mentioned concentrate on improv- ing the estimates of union-nonunion relative wages differential for various sectors of the economy without examining the effects of unionization on labor's share. Johnson and Mieszkowski10 and Diewertn have attempted to analyze the impact of unions on the distribution of income and have shown, via a general equilibrium approach, "that most, if not all, of the gains of union labor are made at the expense of nonunionized workers, and not at the expense I012 of earnings on capital. What is then the union effect on the size distribution? Freeman]3 in an unpublished paper, measures inequality in the general distribution of wages by using both the standard devia- tion in the logarithms and the coefficients of variation. This technique is applied to the wages of unionized and nonunionized workers in both the manufacturing and nonmanufacturing sectors, while at the same time controlling for individual characteristics. The crux of Freeman's paper is a test of the standardization hypothe- sis whereby collective bargaining tends to equalize wage rates across establishments, replace personal rates with formal job rates within plants and reduce white collar-blue collar differentials in enterprises. Results show that unionism reduces dispersion of wages within the organized sector; and by more than offsetting the increase in dispersion of earnings across industries, unionism reduces inequality on net. The present study is an attempt to put forth another method of measuring relative inquality due to unionism and, for two sets of microdata, to estimate the incidence of unionism by wage levels and its relative effect, to make assumptions about the unions' effect on the incidence of periods of unemployment and about what they do to relative earnings shares, then proceed to estimate what they do to the size distribution. The functional equation to be estimated by ordinary least squares is of the form LnEi = f (Unioni, X1: 21) (1) where: Ei is the average hourly earnings Unioni is membership status Xi is the vector of demographic, industry, and occupation characteristics Z1 is the vector of indicators of human capital for the ith worker. Since the observed average hourly earnings Ei is a weighted average of union and nonunion earnings, the latter value can be determined as follows: m ll . u . n Un1oni - Ei + (l - Un1oni) Ei (2) then E? Ei/(l + Unioni - m) (3) where: m is the estimated percentage diffential due to unionism The average union earnings is easily derived from equation (2). For each earnings class the frequency distribution of both types of earnings can be determined on a Lorenz curve generated by a Pareto distribution of the form v=[1-(1-x)0‘1‘/B Oa1 ’8 (20) where: o and B are the parameters to be estimated, and O < a, 8_: 1. The first characteristic of this functional form is that it includes the Lorenz curve specification corresponding to the Pareto distribution of income as the special case of B = l, a < 1.2] Second, the equality line q = p is generated by the case a = B = 1. Furthermore, the function is continuous and is twice differentiable and possesses the proper corvexity and slope constraints proposed by Kakwani and Podder.22 The Gini coefficient.isderived as follows: G = 1 - 2 f3 [1 - (1 - p)0‘]‘/B dp (21) Using the transformation of variables, u = (1 - p)“ G = l - g-f (1 - u)”B 111/0"1 du (22) 1 O 19 2 11 OY‘ G=]--0—LB(E’-B_+]) (23) 1 + 1) is the Beta distribution. where: B (4;, 8 Given estimates of a and B the Gini coefficient can be easily obtained from either standard tables or generally available computer subroutines. Let a and B be the consistent estimates of a and B respectively, then the asymptotic variance of an estimate of Gini coefficient based on equation (23) will be Var (e) = (3%)2 Var (&)+ (3 -—§)2 Var (s) + 24—543) cov (a, £1 (24) where: 3% = 32.5. [e + 11 (3;) - 11(1 + — L+L)] 3‘31" IE.» 1) (25) 8G _ 2 1 1 1 1 1 y‘gg-z-I (§+1)- ‘i’(1+a+§113(5a 8+ 1) where: V(—) is the Euler's psi function which can be numerically computed by making use of the following relationship: QI-J W()-‘i’(l+ +‘§)=§(-——3—-—- ‘) (26) l 0‘ k=01+ pol-J Ql-J 4. 7&- 91-” Thus, if the variances and co-variances of estimates of a and B are known, the standard error of the Gini coefficient based on equation (23) can be computed. The existence of the asymptotic standard errors provides us some basis for statistical test of significance of the difference between Gini coefficients. CHAPTER II: FOOTNOTES 1H. Gregg Lewis, Unionism and Relative Wages in the United States (Chicago: University of Chicago Press, 1963). 2Sherwin Rosen, "Trade Union, Power, Threat Effects and the Extent of Organization," Review of Economic Studies 36 (April 1969): 185-196. 3Also,wages of union members will be rendered higher by unionization of others in the same industry or closely related industries. See Albert R. Rees, The Economics of Work and Pay (New York: Harper and Row, 1973); H. Gregg Lewis, Unionism and Relative Wages in the United States (Chicago: University of Chicago Press, 1963); Sherwin Rosen, "Trade Union, Power, Threat Effects and the Extent of Organization," Review of Economic Studies 36 (April 1969): 185-196; Orley Ashenfelter, "The Effect of Unionization on Wages in the Public Sector: The Case of Fire Fighters," Industrial and Labor Relations Review 24 (January 1971): 191-202; for dis- cussions on this problem. 4This approach is found also in studies for sex discrimina- tion in wages. See Ronald Oaxaca, "Male-Female Wage Differentials in Urban Labor Markets,“ International Economic Review 14 (October 1973): 693-709; also "Sex Discrimination in Wages,“’in Discrimination in Labor Markets, ed.: Orley Ashenfelter and Albert Rees (Princeton: Princeton University Press, 1973) for a detailed analysis. 5Gary S. Becker, Human Capital: A Theoretical and Empirical Analysis with Special Reference to Education—(New York: National Bureau of Economfic Research,'1964); 3150 Human Capital and the Personal Distribution of Income: An Analytical Approach, Woytinsky Lecture No. 1 (Ann Arbor: University of Michigan, 1967). 6Jacob Mincer, "The Distribution of Labor Incomes: A Survey with Special Reference to the Human Capital Approach," Journal of Economic Literature 8 (March 1970): 1-26; also Schooling, Expe- rience and Earnings (New York: National Bureau of Economic Research,‘l974). 7Gary 5. Becker and Barry R. Chiswick, "Education and the Distribution of Earnings," American Economic Review 56 (May 1966): 358-369. 20 21 8See Yoram Ben-Porath, "The Production of Human Capital and the Life Cycle of Earnings," Journal of Political Economy_75 (August 1967): 352-365 for a detailed analysis. 9Jacob Mincer, "The Distribution of Labor Incomes: A Survey with Special Reference to the Human Capital Approach," Journal of Economic Literature 8 (March 1970): 1-26; also Schooling, Experience and EarningsTTNew York: National Bureau of Economic Research, 1974)} 10Orley Ashenfelter and George Johnson, "Unionism, Relative Wages, and Labor Quality in U. S. Manufacturing Industries," International Economic Review 13 (October 1972): 488-508 relax this assumption and posit, instead, that Ui and Schoolingi are endo- genous. HThe truncation of the sample by deciles does not allow a least squares estimation technique to provide an unbiased and con- sistent estimate of mp for each pth decile. The problem can be alleviated by using nonlinear regression, maximum likelihood esti- mation and an instrumental variable technique. Another way is to use values of X for which the expected value of Y given X is well below the truncation point. See David L. Crawford, "Estimating Earnings Functions from Truncated Samples," Discussion paper, Institute for Research on Poverty, University of Wisconsin, Madison, July 1975 and Jerry A. Hausman and David A. Wise, "Social Experi- mentation Truncated Distributions, and Efficient Estimation," Econometrica 45 (May 1977): 919-938 for details. 12By using m instead of hp, the pth decile union nonunion differential, in the determination of earnings components, we mitigate the problem of workers being pushed into a higher bracket while others are pulled down as unionism may have different degrees of influence along the earnings range. For example, regression equations encompassing binary explanatory variables for the differ- ent deciles show union-nonunion earnings differentials of -4.4 percent to 0.0 percent in the lower deciles and about 2.3 percent to 3 percent in the higher deciles, with a drastic decline of -13.0 percent in the highest earnings bracket. The negative sign implies that nonunion wages are higher than union wages. 13A. B. Atkinson, "0n the Measurement of Inequality," Jgurnal of Economic Theory 2 (September 1970): 244-263. 14Hugh Dalton, "The Measurement of the Inequality of Incomes," Economic Journal 30 (September 1920): 348-361. 15Atkinson, op. cit. 22 16Amartya Sen, On Economic Inquglity (Oxford: Clarendon Press, 1973). 17Ibid., p. 3. 18A rigorous mathematical treatment can be found in Atkinson, op. cit., and Sen, op. cit. 19Robert Rasche, J. Gaffney, A. K00, and N. Obst, “Functional Forms for Estimating the Lorenz Curve," Working Paper No. 7706, Department of Economics, Michigan State University, February 1978. 20113101. 2lNanak C. Kakwani and N. Podder, "On the Estimation of Lorenz Curves from Grouped Observations," International Economic Review 14 (June 1973): 278-291. 2211ml. CHAPTER III THE NATIONAL SURVEY OF WORKING CONDITIONS 3.1 Introduction In this chapter we investigate data provided by the National Survey of Working Conditions. The survey was conducted in November and December, 1969, and January 1970 by the Survey Research Center of the Institute for Social Research, the University of Michigan. It covers 1533 self-employed and salaried men and women, each with 660 variables of information regarding their actual job situations and areas affected by the job. This study selects a sample of 1238 individuals 16 to 65 years old who are fully employed and whose weekly working hours are less then 70 hours. The natural logarithm of average hourly earnings is regressed on 24 relevant variables gleaned from the 660 variables provided by the survey. Most of these variables are of qualitative nature. There are four broad categories of variables summarizing population and regional characteristics, occupational characteristics, human capital and working conditions. The variables comprising the latter category are unique to this sample. 23 24 3.2 Explanatory Variables 3.2.1 Population Characteristics It is an observed fact that women in general earn less than men, and blacks and some other minority groups earn less than whites. The causes of these differentials are complex. In part they arise from current discrimination in the labor market and in part are themselves the result of past discrimination in the labor market and in education or training opportunities.1 Since it is a violation of federal laws to pay lower wage rates to women and blacks for the same job in the same work place than to other workers, the most common form of discrimination is to deny them employment in jobs for which they are qualified or demand higher qualifications for the same wage as others. Hence, minorities will be overrepre- sented in the work forces of the employer who pays the lowest wage within an occupation. Moreover, employers frequently prefer minor- ities for certain kinds of jobs because the limited options avail- able to blacks and women and their concentration in certain occupa- tions make them dependable sources of labor. Stereotypes are thus developed.2 On the other hand, the clustering of minorities within certain occupations has the effect of further depressing relative wages and reinforcing employers' profit motives to hire blacks, for example, where all-black work forces can be employed for less than white work forces. In consideration of the above, for each of the sex and race characteristics, dummy variables are used, one for being female or 25 black, zero otherwise. A negative relationship between earnings and these variables is expected. 3.2.2 Regional Characteristics Economic theory has asserted that income differentials between areas would induce migration flows, and all of the evidence suggests that this indeed has occurred. But the theory also predicts that in the long run, the migration of labor, accompanied by a reverse migration of capital, will bring the incomes of different areas into equality. Nothing of that nature has occurred. Classi- cal theory deals only with a system in equilibrium that is disturbed by a once-and-for all shock, whereas most disequilibria are not instigated by single events. The growth of industrial societies has stimulated urbanization, tended to require a broad range of skills and professional competence, to create increasing levels of general education, to restricture the work forces, to meet new labor and management requirements. These and other characteristics create more disparity in wages between urban and rural areas, northern and southern regions. To capture these discrepancies, two binary variables, repre- senting large urban areas and southern regions, are incorporated into the earnings function. From the sample under study, 34 percent of individuals report themselves to be living in large size SMSAs with a population of 750,000 and above, while 24 percent of respon- dents are from the South. Coefficients for the two variables should 26 show higher earnings in urban areas and lower earnings in the South relative to rural and other geographical areas. 3.2.3 Occupational Characteristics One's earnings depend (n1 one's occupation. Occupational differentials reflect differences among workers in levels and types 3 of skills and in conditions of work. Rees points out that in the very short run, supply is inelastic and largely determines the number of people in the occupation, while demand determines their wage. With time and better information on the perspective of training costs, with firmer establishment of tastes and expected earnings, the number of qualified people will increase. Thus, in the long run, the supply of labor to an occupation could be highly elastic. However, even assuming a perfectly competitive market, occupational differentials still remain. The advantages of an occupation include not only the salary and any amenities attached with it, but also the prestige, status and satisfaction which are derived from it. Hence, there should be a job-related premium in order to compensate for an occupation which is unpleasant and which is held in low esteem.4 The subject of nonpecuniary compensation will be taken up in a later section. One component of occupational wage differentials is due to rents from scarce natural talents. Clearly, the earnings of a Nobel laureate include rents which make up all of the differences in earnings between him and the average member of the profession. With the advent of industrialization and the stronger and wider emphasis 27 on acquiring a skill through training and education, a major com- ponent of earnings differentials is the return on investment in acquiring the skills. The private costs of the training needed to obtain skills must be recouped over the worker's working life with a rate of return equal to that of other equally risky investments or equal to the subjective rate of discount used by each individual in making his initial decision. In order to make sure of receiving the "just and fair" pecuniary compensation,members of certain pro- fessions have erected barriers to entry for the express reason of controlling the number of entrants and thus decreasing the elastic- ity of supply. This is the raison d'Etre of almost all license boards, guilds, craft unions and the like. In view of the above facts, this study has included in the earnings equation dummy variables which describe workers falling in the occupational categories of professionals, clerical, craftsmen, Operatives, and service occupation. Some interactive variables are also included to take into consideration the barriers to entry created by unions. 3.2.4 Human Capital It would be redundant to talk about the effect of training and education on earnings differentials in this section, but it should be relevant to point out that up to this point the term "training" and "educational level" are too general and do not entail any measure of the length of time involved in those processes. In contrast, human capital models point out individual investment 28 behavior as a basic factor in earnings differentials. Further, and most importantly, the model takes the length of training as the basic source of heterogeneity. J. Mincer5 points out that training increases productivity and thus raises the real wage, but the time spent in training necessitates postponement of earnings to a later age. The various amounts of training are undertaken in the hope that future earnings are sufficiently large to compensate for the cost of training, mainly the foregone earnings. From these simple 7 Becker,8 and Ben-Porath9 have assumptions, Mincer,6 Chiswick, expanded the human capital model to account for some qualitative features of observed distributions of earnings such as the life- cycle hypothesis and the optimization of earnings over time. Closer to this paper's concern is the correlation between earnings and investment in human capital. Empirical results show that the goodness-of-fit, measured by the coefficient of determina- tion, is highest for individual earnings when schooling and expe- rience variables are incorporated in a parabolic earnings profile in logarithm.10 The schooling variable is the number of years of school the individual has completed, while the experience variable is the person's age at his last birthday minus the years of school- ing minus the first six years of his life. To take into account diminishing returns on investment, the experience variable is employed as a quadratic, reflecting the nonlinearity of the earn- ings function. The sign of the coefficient of the squared term is expected to be significantly negative. 29 3.2.5 Working Conditions Adam Smith11 argued that the wage should reflect compensa- tion for unpleasant jobs or for jobs held in low esteem. It is not only compensation for opportunities foregone, but also it should cover the psychic, and in some instance, the physical cost of hold— ing a job. A pictorial example of occupational hazards is the tribulation of Charlie Chaplin in Modern Times, where an assembly line worker, despite the simplicity of such task as tightening bolts, could be severely affected mentally and physically on and off a job that is so repetitious and fast-paced. Hence, besides the North-South difference, rural versus urban zones, the lack and length of training, earnings differentials can be explained by the nonpecuniary aspects of the jobs. Assuming that the marginal rate of substitution between pecuniary and nonpecuniary returns diminishes with income, the question is how much an individual worker is willing to pay for more and more pecuniary goods in order to keep him on the same indiffer- ence curve. To answer this question, Rosen12 states that workers operating within the confines of a budget constraint will tend to pay for fewer pecuniary goods if kept at a lower preference level. Workers' indifference maps reflect tastes. On the other hand, firms do supply the nonpecuniary benefits. Operating under increas- ing long run marginal cost, firms require a higher proportion of pecuniary returns, namely market price, in order to produce more nonpecuniary benefits. Thus the firm will have to be at a higher 3O isoprofit curve. For a given market price, the higher profit firm will provide more of the nonpecuniary goods. The price which clears the market is the locus of equilibrium points between the isoprofit and indifference maps. R. E. 8. Lucas13 attempts to test this hypothesis by including in the wage equation proxies for nonsedentary, hazardous, repetitive and supervisory aspect of jobs. His results were mixed, except for the "repetitive" variable which significantly explained a small percentage of the increase in the wage. Expanding on the hypothesis that differential wages are due to nonpecuniary aspects of occupation, this thesis attempts to view work pace and work scheduling as substitutes for individual skills. The heterogeneity of the labor pool forces an employer to use certain work requirements, especially work pace, as a screening device for identifying productive workers.14 Technological innova- tions and mass production have transformed the idyllic Smithian pin shop into a cobweb of assembly lines along which workers produce more output at faster speed. Those workers who are able to race at the prescribed speed are judged to have superior ability and are com- pensated accordingly. Higher wages then induce people to work under higher "speed" or harder working conditions, and, at the same time increase their desire to move toward a less hectic, more leisurely occupation which in essence requires a higher educational attainment. 0n the other hand, excessively harsh working conditions, in the eyes of the involved worker, create an atmosphere which is conducive toward union organization. The latter represents a mechanism for 31 mitigating the problems of the work place. Giving support to this hypothesis, Duncan and Stafford15 show that higher earnings of union members reflect in part nonpecuniary benefits rather than rents. In the sample under study, proxies for working conditions came from respondents' description of their jobs. Solicited descrip- tions of jobs have the following format. "A job that requires that you work very fast. Would you say this is a lot like your main job, somewhat like your main, a little like your main job, or not at all like your main job?" The subjective answers are incorporated into a binary variable. For those respondents who claim that their jobs are a lot, somewhat, or a little like the description, the variable is assigned a value one and zero otherwise. Similarly, binary explanatory variables are also used for jobs which require hard work and physical effort, and for jobs which allow some degree of freedom on how the task is performed. Moreover, to take into account the previous proposition that, besides the intellectual skill pro- vided by formal schooling, physical skill may cause, at least in the short run, large wage differences, another binary variable is included in the earnings equation for jobs which are described as requiring some "high degree of skill." 3.3 Estimation The basic equation is a regression of the natural logarithm of average hourly earnings on a set of dummy variables representing population characteristics, human capital indicators, industrial and occupational characteristics. Separate regressions were run to 32 allow for interactions between union membership and industry, as well as between union membership and occupation. The general form of the equation is as follows: 1n E1 = f (Unioni, Experiencei, Experienceg, Educationi, Femalei, Blacki, Urbani, Southi, Working Conditionsi, Industryi, Occupationi) (27) where: E1 is the average hourly earnings received by the ith worker. Other continuous variables are education, i.e., the years of school- ing completed, and experience (which is Age-Education-6). Measured as dummy variables are union status, female, black, urban, south, working conditions, industry and occupation. Variables which encom- pass working conditions are: FAST WORK, FREEDOM, SKILL, HARDWORK, PHYSICAL WORK. The Variable FREEDOM is used to describe ”A job that allows you a lot of freedom as to how you do your work." In other words, FREEDOM is not meant to describe idleness or free time on the job, but it is a proxy for a job which provides responsibility to the holder. The other working conditions variables are self- explanatory. Industry variables are construction, manufacturing, transportation, wholesale and retail trade, finance and service. Occupation variables are professionals and managers, clerical, craftsman, operatives and service workers. The reference group sub- sumed in the constant is a nonunionized white worker not residing in the south and holding a farming job. 33 3.4 Empirical Results Table 3.1 gives the results of the least-squares estimation of equation (27). In column 1 estimated coefficients for the rele- vant variables are presented, while in column 2 estimates of the interactions between unionism on the one hand and industry and occupation on the other hand are introduced. Columns 3 and 4 tabu- late regression results for the blue-collar and white-collar workers samples. Estimates are classified by subgroups such as working conditions, industries, occupations and union interactions. The value of the adjusted R2 is relatively high given the nature of the data set. For 1237 observations, the adjusted R2 is about .53. As expected, the estimated coefficients for human capital indicators and individual characteristics have the theoretically correct sign and are highly significant. In the overall sample (column 1) the experience variable contributes, with significant diminishing returns, about 5.1 percent to the explanation of the 16 This contribu- variations in earnings given all other regressors. tion drops to 4.7 percent in the blue-collar workers sample, but increases drastically up to 88.4 percent in the white-collar workers sample (column 3 and 4 respectively). At the same time the education variable contributes about 8.7 percent to the explanation in earnings for the overall sample, while it is 5.8 percent for the blue collar sample and 10.7 percent for the white collar sample. This result reinforces the notion that formal schooling and post-schooling 34 TABLE 3.l.--Hour1y Earnings Regressions, Working Conditions Data, 1969 (absolute value of standard errors in parentheses) Blue White 5;:1;g:§g*y (1) (2) Collar Collar (3) (4) Constant .1052 .1169 .1656 .3560 .0877) (.0827) (.1158) (.1611) UNION .1239 .1183 .2129 -.0174 .0249) (.0614) (.0450) (.0400) Exper .0170 .0166 .0149 .0196 .0021) (.0021) (.0027) (.0003) Experz .00027 -.00026 -.00022 -.00032 .00004) (.00004) (.00005) (.00007) Education .0593 .0598 .0487 .0672 .0055) (.0055) (.0079) (.0082) Female -.4399 -.4322 -.5020 -.4277 .0267) (.0259) (.0379) (.0356) Black .0191 .0244 .0285 .0571 .0361) (.0360) (.0446) (.0606) SMSA .1610 .1558 .1065 .1847 .0251) (.0251) (.0344) (.0353) SOUTH -.1393 -.1332 -.1584 -.1051 .0290) (.0289) (.0385) (.0436) Working Conditions Fast Work .0590 .0144 (.0358) (.0378) Freedom .0601 .0579 .0328 .0804 (.0228) (.0226) (.0307) (.0334) Skill .0894 .0912 .1175 .0431 (.0252) (.0249) (.0325) (.0385) Hard work .0479 .0613 -.00297 .0911 (.0274 (.0254) (.0390) (.0376) TABLE 3.l.--Continued 35 Blue White Explanatory Variables (1) (2) CgLLar CoLLar Physical -.0820 -.O8l8 -.0525 -.0921 .0282) (.0278) (.0345) (.0477) Industries Construction .1032 -.0086 .1474 .1103 .0600) (.0770) (.0585) (.1644) Manufacture .0407 .0567 .1060 .0073 .0429) (.0438) (.0424) (.0662) Transport .0489 .0186 .1190 -.0647 .0555) (.0758) (.0595) (.0881) Wholesale -.l388 -.1821 -.2086 Retail .0432) (.0430) (.0615) Finance .0543 .0484 .0199 .0641) (.0644) (.7538) Service .1054 -.1024 -.1640 .0430) (.0431) (.0598) Occupations Professionals/ .2488 .2627 -.O67O Managers .0469) (.0376) (.0971) Clerical .1498 .1589 -.l73l .0441) (.0331) (.0983) Craftsmen .0885 .0828 .0652 .0427) (.0345) (.0490) Operatives .0475 -.0053 .0382 .0529) (.0679) (.0669) Service -.0266 -.0319 -.2700 .0477) (.0477) (.216) TABLE 3.1.--Continued 36 Blue White 1212:1122” (0 <2) 61‘1" “(‘14. 3 4 Interactions UCONSTR .1949 .1083) UMANUF -.Ol42 .0732) UTRANSP .040] .1067) UWHRT .1505 .0902) USERVI .0959 .0784) UCRAFTS .0470 .0929) UOPER .2860 .2306) UCRAFCO .2892 .1225) N 1237 1237 636 582 82 .5280 .5289 .5393 .5555 37 experience are strong elements in explaining variations in earnings within the white-collar occupation. Female workers earn on the average 36 percent less than their male counterparts with a greater discrepancy in the blue-collar sample. Earnings differential between black and white workers is small and not at all significant. As far as residential location and regional observations are con- cerned, workers living in large urban areas have a 17.5 percent edge and southerners earn about 15 percent less than workers in other 17 White collar workers in urban areas have parts of the country. higher earnings than any group; at the same time, they do not suffer financially as much for residing in the south. Variables depicting working conditions attempt to explain the effects of nonpecuniary aspects of the jobs on earnings. These binary variables encompass mainly the physical side of a job, while subsuming in the constant its creative aspect. The estimated coef- ficients for variables in the overall sample such as freedom on the job (FREEDOM), job requiring manual skill (SKILL), job requiring hard work (HARD WORK) are all statistically significant at the 10 percent level and have the expected sign. Freedom on the job pro- vides the worker with some degree of responsibility which implies either intellectual or manual ability. One unexpected result is the highly significant, but negative, estimated coefficient for the variable depicting physical effort on the job (PHYSICAL). One reason is that jobs on which physical efforts are needed are gener- ally menial jobs and for which machine utilization is minimal. 38 The regressions for blue-collar and white-collar sub-groups provides some interesting contrasts, as seen in column 3 and 4. For the blue-collar workers, coefficients for FASTWORK become highly significant, as are those for SKILL; while the estimate for PHYSICAL is statistically significant at the 13 percent level only, FREEDOM and HARD WORK are not at all significant. In the white-collar sub- group, FAST WORK has very little effect on earnings, as expected from that category of workers, but coefficients for FREEDOM, HARD WORK, and PHYSICAL do show a significant explained variation in earnings. The first two variables have positive estimated coeffi- cients while the latter is negative. In other words, work pace is best measured by the proxy variable FASTWORK for the blue-collar worker to reflect the intensive utilization of capital, while it is best accounted by the proxy FREEDOM in the white-collar group to reflect the more flexible work schedule and the higher degree of extensive responsibility. Moreover, hard work is a disutility for a white-collar worker, so that he must be compensated proportionately, and having to provide physical effort is considered menial. In contrast, for a blue-collar worker, hard and physical jobs are con- sidered the norms, so that these proxies do not provide any sta- tistical effect on earnings, while acquiring a skill should reflect opportunity costs and obstacles to entry. The estimated relative effect of unionism on earnings in the overall sample is .1239 or a union-nonunion earnings differential of 39 ‘8 The difference in the effects of unionism on earn- 13.19 percent. ings is clearly significant when the sample is disaggregated into blue-collar and white-collar. The proportionate differential is about 23.7 percent in the blue-collar category while it is almost nil in the white-collar. The industry and occupation variables subsume, as reference groups, agriculture and mining, farmers and laborers. In the over- all sample, the estimated coefficients for construction, wholesale and retail trade, and service industries are significantly large. Among the different classifications of occupations, professionals and managers, clerical and craftsmen make significant contributions in explaining variations in earnings. One interesting result con- cerns the effect of unionism on certain sectors, especially in construction and wholesale retail trade19 (column 2). The estimated coefficients for construction and wholesale retail trade are -.0086 and -.1821 in the nonunion sector. The former value is not statis- tically significant. However, the coefficient for the unionized construction industry variable (UCONSTR) is .1949 and that of whole- sale retail trade (UWHRT) is .1606. Both are statistically signifi- cant. Thus, the union-nonunion earnings differential within the two industries are respectively 37 percent and 32 percent.20 The effects of unionism in manufacturing, transportation and service industry (portrayed by UMANUF, UTRANSP, and USERVI reSpectively) are not at all significant, similarly for craftsmen and operatives 40 (UCRAFTS and UOPER). As expected, the blue-collar sample shows the stronger relative effects of construction, manufacturing and trans- portation on earnings, while the estimated coefficients for the occupation variables such as craftsmen and operatives are small in magnitude and are not statistically significant. This is due to a simultaneity problem between the dependent variable and the explana- tory variables CRAFTSMEN and OPERATIVES. On the other hand, typical white-collar industries and occupations, such as wholesale-retail trade, service and clerical work show a significant relative effect on earnings in the white-collar sample.21 In summary, the empirical results show that: (l) besides human capital indicators, nonpecuniary aspects of the work place contribute significantly in explaining variations in earnings. Moreover, these effects are occupation specific; (2) the proportion- ate union nonunion earnings differential in 1969 is about 13.2 per- cent; this value increases to 23.7 percent for the blue collar sample, but becomes negligible in the white-collar sample, and (3) in some industries, such as construction and wholesale-retail trade, unionism provides an edge greater than 30 percent relative to non- union earnings. The relative effect of unionism is found to be small and not significant in manufacturing, transportation, and service industries. 41 3.5 Determination of Union and Nonunion Earniggs Distributions The observed average hourly earnings Ep for the pth decile is a weighted average of the mean hourly union earnings E: and the mean hourly nonunion earnings E2. The latter value can be deter- mined as E: = Ep/(l + UNIONi - m), where m is the estimated differ- ential due to unionism and UNIONi is the membership status of the ith individual. The mean union earnings is derived as E: = (Ep - E:)/UNION1 + E2. The union-nonunion earnings differential have been estimated from the overall sample as m = 13.19 percent. Essential for the computation of the components of earnings is the incidence of unionism over the ten deciles. Table 3.2 shows the percentage of unionized workers for the different deciles. The skewed bell shape distribution of union membership over the spectrum of earnings, with a large mode encompassing the seventh, eighth, and ninth deciles,is clearly implied. As expected, union membership is small in the lower deciles, less than 15 percent, but increases progressively up to the ninth decile. In the highest decile member- ship plummets down to 28 percent. The low percentage of unionized workers in the highest decile is due to the high concentration of professionals and other white-collar workers within that earnings bracket. Table 3.3 shows earnings shares received by deciles of unionized and nonunionized recipients. In order to compare the relative shares, column 3 of Table 3.3 provides the resulting dif- ference between nonunion and union earnings shares. The results do 42 TABLE 3.2.--Incidence of Unionism by Deciles Among Private Workers, Survey of Working Conditions, 1969 (in percent) Earnings Rank Unionization Rate Lowest Decile 8.87 2nd Decile 14.52 3rd Decile 28.23 4th Decile 33.06 5th Decile 34.68 6th Decile 41.13 7th Decile 53.66 8th Decile 51.61 9th Decile 53.66 Highest Decile 27.64 43 TABLE 3.3.--Relative Shares of Union and Nonunion Earnings by Deciles, Survey of Working Conditions, 1969 (in percent) Earnings Rank U???" No2ggion (Egang?]) Lowest Decile 3.57 3.64 .07 2nd Decile 5.21 5.30 .09 3rd Decile 6.31 6.37 .06 4th Decile 7.34 7.41 .07 5th Decile 8.44 8.44 .00 6th Decile 9.92 9.91 -.01 7th Decile 11.13 11.06 -.07 8th Decile 12.23 12.01 -.22 9th Decile 13.96 13.08 -.16 Highest Decile 21.89 22.06 .17 Percent 100 100 44 show some symmetry in the relative effect of union shares about the fifth deci1e.22 In deciles lower than the fifth, the shares of nonunion workers are about .07 percentage points greater than those of union workers. By contrast, beginning with the sixth decile union workers' shares predominate, culminating at .22 percentage points above nonunion shares. This is due to the higher incidence of unionism within the upper middle class. In the highest decile, however, with the decline in union membership, nonunion earnings shares are .17 percentage points greater than union shares. Varia- tions in earnings shares could be further explained by examining the various sources of earnings for a given decile; in other words, what types of occupational jobs fall in what earnings bracket. One can then infer their impact on the earnings shares. This problem will be discussed in the following section. 3.6 Lorenz Curves and Measures of Inequality With the symmetrical variations in the relative shares of both types of earnings, the question arises whether or not these variations are in effect interdecile transfers of earnings toward an improved redistribution among receivers. The problem is basically that of comparing two frequency distributions. Comparison is done by means of a Lorenz curve which is the graphical relationship between the cumulative distribution of earnings and the cumulative distribution of earners. Figure 3.1 shows the Lorenz curves fitted for both union and nonunion cumulative distributions. The Lorenz curve for the nonunion earnings distribution lies above that of 45 Earnings l .. --..—. Union .. ______ Nonunion Receivers Figure 3.1.--Lorenz Curves for Union and Nonunion Earnings, Survey of Working Conditions, 1969. 46 m.» ggmn— . mLmLonmg . mm Pmm.‘ cop om.m ¢¢.N mm.— mm.op me.mp p¢.mo mpwomo umwcmP: oop mm.¢p ww.v mm.“ -.mm Fo.mp mw.om mpwowo gum oop Pn.mp om.m me.o em.m~ om.NP cm.m~ meoma sum 00— mm.PP mm.¢ m_.m m~.¢m mm.om eP.FN mpmomo sun cop mm.mp op.m— mv.o nn.pm mm.o~ mm.mm mpwumo sum cop mm._P mo.m nw.m Fm.mm oo.m~ no.o~ opwomo gum oop Pu.mp mo.w mm.¢P mp.m~ m¢.Fm m~.mp mpwumo gu¢ oo— om.o~ mm.mp oa.m— w¢.o_ mm.o¢ w¢.op opwomo ugm cop op.m— mm.mm mm.pp w¢.op m¢.Fm m~.p~ mpwuwo new cop om.m— op.nm mm.m mo.¢ mm.om m¢.o upwooo awoke; FMWWH tcwwwo mommwmm mo>_Mwwmao cosmmwocu pmwmwopu mpmcomwwmmogq xcmm mmcwccem Apcmocma :Pv mmmp .mcowuwucou mcwxcoz mo >m>czm .mmpmooo an mucmu_o:H chowpoaaoooiu.¢.m m4m a _ea_ aea_ seoeeea_ xu Ammmmcucocon cw mcoccm econcoum mo oa—t> mu:_0moev cmz .xo>c:m pmcwvaamm:04 _e:o_poz .mcowmmocmom mmcwccmm xpcao:11.F.v udmmo op uwumanuc mucocowmwmout cccc. mccc. cmcc. c_mc. mm cccm cccm mccm mccm z Amccc.c Ammmc.c mmmc. mm_. cmc_. mm_. cmmcc are meaoz 4c sec mead: «a _mc_ ccc. vmzcwucounu.F.e m4mcmm -mmew -muwuwm -wcwww mmpcm -cmmwwmfi unawmmmmwwm mmcwcccm momp .>m>c:m pmcwuaawm204 chomumz .mwpwoon an mocwuwocH chowucc=OOO1t.m.v mqmcmm -wwwww -mummww uwcwmw mc.cm -momwwmh -cocmmmmwhm mmcwcccu .mm. .xm>c:m .mcwuaummcoc .mcowpmz .mmpwowo An mocwuwocH .aco.caccooc--.c.e mcc<. 86 of information and search, which have no screening barrier, and for which there is some prevalent or "natural" wage accepted by both the employers and the unorganized workers. This implicit acceptance of some prevalent wage is equivalent to the union standard rate.13 Hence, as with the union rate, there is little variation in differ- ences in the wage structure of the lower earners group. This homo- geneity in wages, coupled with a relatively low level of unionism, tends to push the lower end Of the Lorenz curve toward the line Of total equality. The heterogeneity found in the higher wage earners group derives from prior intensive investment in human capital and acquired skill. Indeed, in the highest decile the mean level of education is about 13.2 years, which reflects the large concentration Of professionals and managers, two occupations wherein the level of organization is low. On the other hand, craftsmen are concentrated strongly in the upper deciles, and union membership is correSpond- ingly at its highest. Hence, union's cumulative shares tend to be greater than nonunion's. Consequently, the Lorenz curve for union earnings lie closer to the line of equality in the upper deciles. The Gini coefficients for the union and nonunion earnings distributions are Obtained from the estimated functional form Of the Lorenz curve discussed in Section 2.5. Table 4.7 shows the results of the estimated coefficients with their corresponding standard errors. The existence of asymptotic standard errors allows statisti- cal testing. The Gini value for union earnings was approximately .265 and .264 in 1969 and 1971 respectively, as for nonunion earnings it was .267 and .265. In other words, union earnings provide a more 87 TABLE 4.7.--Estimates of Lorenz Curves and Gini Coefficients, National Longitudinal Survey, 1969 and 1971 Overall Earnings Union Earnings Nonunion Earnings (l) (2) (3) 1969 a .7372 .7301 .7081 (.0057) (.0041) (.0042) 8 .8060 .8139 .8344 (.0062) (.0045) (.0050) Gini .26517 .26496 .26724 (.00014) (.00015) (.00036) 1971 a .7493 .7429 .7246 (.0039) (.0039) (.0031) 8 .7950 .8020 .8196 (.0042) (.0043) (.0035) Gini .26400 .26377 .26507 (.00013) (.00016) (.00017) NOTE: Figures in Parentheses are asymptotic Standard Errors 88 equal distribution among recipients than nonunion earnings. Further- more, Over the two-year span, the magnitudes of the relevant Gini coefficients tend to decrease. Since the standard errors for the relevant estimated Gini coefficients are known, it is possible to test the hypothesis that the difference between UNION and NONUNION Gini coefficients is zero. The values of the t-statistics are 5.84 and 5.57 in 1969 and 1971, respectively.14 This shows that the UNION Gini is statistically different from NONUNION Gini, and that the former is smaller than the latter. 4.6 Conclusion In analyzing the impact of unionism on the distribution Of earnings of the Older male cohort, we have found the following results. The relative earnings differential due to unionism within that age group is about 16 percent to 18 percent on the average. This differential varies by industry and by occupation. It is greater than 50 percent in the construction industry and is negligible in manufacturing, and within occupations the differential among Opera- tives gravitates around 45 percent to 50 percent, while it is not significant among craftsmen. The separate cumulative distributions of union and nonunion earnings, and the subsequent fitting of Lorenz curves show a more equal distribution Of the nonunion earnings in the lower deciles than the upper deciles. The Gini coefficients obtained by estimating a functional form of the Lorenz curve show that the distribution of union earnings is relatively more equal than that Of nonunion earn- ings over both years. FOOTNOTES: CHAPTER IV 1This aspect of the problem of finding jobs does not con- sider discrimination by employers against older job applicants. 2Arvil V. Adams, "Earnings and Employment of Middle Aged Men: A Special Study of Their Investment in Human Capital," in The Pre-Retirement Years Vol. 4, ed.: Parnes et a1., Manpower Research Monograph NO. 15. 3One explanation is that such programs may be the result of the enforcement of fair employment practices legislation. 4Yoram, Ben-Porath, "The Production of Human Capital and the Life Cycle Of EArnings," Journal of Political Economy 75 (August 1967): 352-365. 5Sherwin Rosen, "Trade Union, Power, Threat Effects and the ngent of Organization," Review of Economic Studies 36 (April 1969): -l96. 6George E. Johnson and K. C. Youmans, "Union Relative Wage Effects--by Age and Education," Industrial and Labor Relation Review 25 (January 1971): 171-179. . ‘7An F test has been carried out for the two sets of loca- t1on variables. Results show that both sets have a highly signifi- cant effect on earnings with F values greater than 25.0. . 8This percentage differential is obtained by taking the gant1109 ()f 6) minus one, where 6 is the estimated coefficient on n10n1sm. . 9However, this differential is still 2.5 percentage points hlghef‘ than the one estimated from the more heterogenous sample USed 1n Chapter III. 0Rosen, Op. cit. . .nsee Richard B. Freeman, "Individual Mobility and Union Voice 111 the Labor Market," American Economic Review Papers and Proceedings, 66 (May 1976): 361-368. 89 9O lettempts to estimate union effects by deciles do not pro- vide any significant results. Dummy variables which capture the interaction between unionism and each decile are also incorporated into the regression equation. The resulting coefficients are highly overestimated. 13Workers at this lower earnings scale are also influenced by minimum wage legislation. 14We again assume the covariance of the estimates Of the Gini coefficients to be equal to zero. CHAPTER V THE NATIONAL LONGITUDINAL SURVEY: DISAGGREGATED ESTIMATES 5.1 Introduction Changing employment patterns within the U. S. economy over the past decade have had an impact on the industrial composition Of union membership. Employment in the manufacturing sector, where unions have traditionally been strongest, has remained relatively stable since 1962, while employment in service-producing industries, including government, has increased by 41 percent.1 Thus, it is not entirely coincidental that unions have made their most sizable gains in the government and nonmanufacturing sectors. Since 1956, only in the government sector have unions consistently gained both in absolute numbers and as a percentage of total employment. Union membership in the manufacturing sector, which had stabilized around 46.7 percent of the total organized work force, declined signifi- cantly, beginning 1970, to 42.8 percent in 1972.2 Union membership in nonmanufacturing has increased mainly due to the higher demand for employment in services. Indeed, the strongest relative gains are in service oriented industry groups and in contract construction, while there are some losses in wholesale and retail trade and in transportation over the same time period.3 91 92 Taking into consideration the patterns of union membership in the aggregate, the sample Of Older male workers from Chapter IV is divided into two main industry groups which are further subdivided into blue collar occupations. Section 5.2 provides the estimated union-nonunion earnings differential in the manufacturing and non- manufacturing sectors, section 5.3 looks at the union incidence and relative shares by earnings deciles, and finally, Gini coefficients for both union and nonunion earnings are determined in Section 5.4. 5.2 Union-Nonunion Earniggs Differential Equation (28) is again used to fit the separate samples. Using the same control variables for both industry groups, the union- nonunion earnings differential is estimated and is presented in Table 5.1. The various regression coefficients for the different control variables are relegated to the Appendix. Within the manufacturing sector, union workers held a 6.30 percent advantage over otherwise identical workers in 1969. This gap increased to 9.43 percent in 1971. The latter figure tends to reflect the inherent rigidity of union contracts. Theoretically, excess labor supply during economic downturns will cause nonunion wages to decline faster than union wages, as employers not parties to collective bargaining agreements--frequently of two to three year duration--are able to react more quickly to labor market changes. Hence, unionism shows its intrinsic power in time of relatively high unemployment. Organized production workers--craftsmen, operatives and nonfarm laborers--in manufacturing had a 8.12 percent edge over 93 TABLE 5.l.--Union-N0nunion Earnings Differentials in Manufacturing and Nonmanufacturing, National Longitudinal Survey, 1969 and 1971 Manufacturing Nonmanufacturing Year Overall Blue Collar Overall Blue Collar White Collar 1969 .0629 .0812 .3184 .4787 .0245 1971 .0943 .0863 .2760 .4210 .0563 94 unorganized workers in 1969. This union-nonunion earnings differ- ential remained at the 8 percent level in 1971. The stability in the differential can again be explained by the spillover enjoyed by unorganized workers. Indeed, in 1969, 62 percent of the blue collar workers in manufacturing were unionized, while in 1971 the sample showed that the degree of unionism remained high, at 57 percent. The threat of unionism may induce nonunion employers to raise wages above the competitive wage rate. Moreover, the presence of unionism in one part of an industry may result in increased wages not only in that part of the industry, but also in related manufacturing industries. Within this institutional framework, nonunionized workers do accumulate some strong and uninterrupted externalities from unionism which push the level of wages above that Of a more competitive and less concentrated sector. Also, besides the poten- tial relative wage advantage due to union membership, the typical worker may benefit from the influence Of the union on the non- pecuniary aspects of his work attachment, especially through griev- ance procedures and seniority systems. One of the factors which explains the narrowing gap in manufacturing wages in union and non- union plants in the same industry is the greater improvement of nonwage benefits in union than in nonunion plants. There are a number of studies in the industrial relations literature showing that nonpecuniary factors are a key determinant in the worker's 4,5 decision to join a union. Almost all of the U. 5. steel workers interviewed by the aforementioned authors do not state receiving 95 higher wages as a reason for joining unions, but believe that unions are a conduit for correcting personal grievances and that union procedures could eliminate any bad work experience. By contrast, the nonmanufacturing sector shows a 31.8 per- cent union-nonunion earnings differential in 1969 which declined to 27.6 percent in 1971.6 This is a marked difference from the widen- ing gap in union-nonunion differential found in the manufacturing sector for the latter year. This decline in differential is not due to a change in the average degree Of unionization for it remains constant at 32 percent over the two year span. Rather this may suggest, on the one hand, a gain of the nonunion sector in spite of unfavorable market conditions, and on the other hand, because of the diversity of the nonmanufacturing sector and hence of the various union bargaining units, unionism may be weaker than otherwise. The gain from the nonunion side can be seen from the occupational mix of the nonmanufacturing sector; there is a relatively high number of professionals and managers along with clerical sales, services, and public workers for whom earnings are determined by the level of education, job tenure, and personal characteristics and/or shift in demand due to change in industrial mix rather than based on piece rate and working conditions. As the sample is narrowed down to encompass only blue collar workers, the union-nonunion differential remains above 40 percent over the two years. In this case, there is lesser variation ir1 earnings due to homogeneity in the sample and strong spillovers from unionism in the blue collar 96 occupations. Furthermore, unionized white collar workers in non- manufacturing industries earned about 2.4 percent more than their nonunionized counterparts in 1969, and this differential increased to 5.6 percent in 1971. At the same time, the unionization rate among blue collar workers went from 18 percent to 22 percent over the two-year span. One result which stands out from Table 5.1 is the signifi- cantly large magnitude Of the union-nonunion differential in non- manufacturing relative tO the manufacturing sector.7 It should be emphasized that comparisons between the two industrial groups are misleading at best. The problem arises due to: (a) conceptual differences between union membership and collective bargaining coverage, and (b) inaccuracy in the measurement Of variables. The fact that an individual belongs to a union cannot have much effect on his earnings. What counts is the extent Of union membership within a worker's place of employment and the extent of collective bargaining coverage among the firms with which his employer must compete.8 Presumably most union members are in organized plants so union membership would be a good proxy for the ability Of the union to influence the terms of employment set by the employer. On the other hand, union membership would not be a very good proxy for the extent Of unionism within the industry, since there are many industries in the nonmanufacturing sector with sub- stantial numbers Of both union and nonunion plants. Hence, it is a weak variable for determining the power of a union in an organized 97 plant to raise wages without taking into account the interdependence among combined effects Of union membership and the extent of col- lective bargaining, and the inclusion only of union status as an explanatory variable in the earnings equation understates the true relationship. The other problem arises from the very broad groupings of industries. The introduction of dummies for such categories as "wholesale and retail trade" or "services" cannot possibly capture industry characteristics like plant size or employment growth. Hence, some Of the earnings difference that we attribute to union- ism is actually due to other industry characteristics. This dif- ference is much wider in nonmanufacturing than in manufacturing. While the grouping "manufacturing" is quite broad, it provides more internal homogeneity than such diverse and heterogeneous classifica- tions as mining, construction, transportation, services, and public administration which constitutes the nonmanufacturing sector. In summary, union workers in manufacturing have a 6 to 9 percent edge in earnings over nonunion workers while those in non- manufacturing have a 28 to 47 percent advantage, with organized production workers having the greatest gain. 5.3 Union and Nonunion Relative Shares Using the estimated union-nonunion earnings differentials presented in Table 5.1, union and nonunion earnings for both the manufacturing and nonmanufacturing sectors are determined. From these values their respective relative shares are computed and 98 presented in Table 5.4 to 5.7. Crucial to the determination of earn- ings and shares is union membership at each earnings decile. Table 5.2 and 5.3 show the incidence Of unionism by earnings bracket for both the 1969 and 1971 years. As the sample is dichotomized into the manufacturing and nonmanufacturing sectors and each is further narrowed down for production workers, the distribution of union membership over the spectrum of earnings follows a nearly flat, slightly skewed bell shape curve with a mode centered around the sixth or seventh decile. In general, in both years, union membership in manufacturing shows a very steep decline within the highest decile with approximately 7 percent of the sample; this is expected, as only relatively high- paying white collar jobs are found in the top decile. In 1971, however, the overall level of union membership in manufacturing showed a substantial decline from what it was in 1969 with some aberrations at the sixth and eighth deciles. This trend has been observed in the aggregate and also in the overall sample which is reported in Chapter IV. As for production workers, there is a relatively smaller percentage Of union workers in the two lowest and highest deciles. Overall, the percentage Of organized blue collar workers is relatively high, greater than 50 percent, and is equally distributed through the third to ninth deciles. Again, the declining trend in union incidence is also observed for the latter year. This high incidence of unionism in the blue-collar sample explains the relative stability of union-nonunion earnings differ- entials which derives from the large spillover effects. 99 m.c. m.mc m.cm m.cc c.c o..ooc caoec.z c... m.cc m.mc m.cc c.cm o..ooc ccc e.c. m.cc c.mc c.cc c.mc o..ooc ecc c..m m... ..cc m... m..c o..oac cc. m.mm m.cc c.cm m... m.cc o..oec coc m.mm m.cc c.cc c... m... o..ooc cam m.cm ..cc c..m c.m. c... a..ooc ecc c..m m... ..c. ...c m.cc e..ooc ccm m.c. c.. m.c m.ce ..ce o..ooc cem c.m m.m ..m c.mm c.cm o..ooc chosen ca..cc cc.c: ca..cc cc.c ..aco>c ca..cc cc.c ..eco>c xcmm mmcwccmm mc.c:pocwcccacoz mcwccpucmccmz choococ cwv mom. .xo>c=m .ccwccpwmcoc .mcowucz .mcwccuocmcccacoz use mc.c:uocwaccz cw mm.wuoo an Emmcowca mo oucocmocmun.m.m m4mc ca..cc mc.c ..aco>c xccm mmcwcccm mc.c:uocmaccEcoz mc.c:uoc$:cmz chwocma cwv pump .am>c:m .mcwuaummcog .mcomuwz .m:.c:uom»=cc§coz new meccauommzcmz cw mopcooo >n Emwcowca22cmocwuwo:~11.m.m m4mo cc..oo w:.m .pmco>o xccm mmcccccu m:.c:uocw==c2coz mc.c:uommzcmz Apcmocmc cvv mom. .ao>c:m .mcwuzuwmcoA .mcowucz .mcwcauumwzcmscoz can mcwccpocwcccz cw mm.mooo An mmcwccmm comczcoz can come: yo moccsm o>wpc.mmtu.¢.m m4mo cc..ou o:.m Ppccm>o m:.c=pocm:cc2coz mc.cauocmcccz Apcmocmc :Fv .nm. .am>c:m Pccwccuwmcog .ccowucz .mcwczuommzccscoz vcc mc.c:uocwzccz c. mc.como an mmcccccm coccccoz can come: we mmccgm m>wucpomtt.m.m m4mc .mmccc.. .ccccc.c .cmccc.. .cc.cc.c .cc.cc.c ccccm. ccmcm. cc.cm. mccc.. mccmm. ec.ececz .ccccc.. .mcccc.c .ccccc.c .c.ccc.. .cc.cc.. cc.cc. mcccm. ccccm. cccc.. .ccmm. cc.c: ..c. ..c.cc.. ..c.cc.. .mcccc.v .cm.cc.c .cm.cc.v cmmcm. mmmmm. cmccm. m.cc.. .mccm. ..cco>c .mc.cc.c .cc.cc.. .ccccc.v ..cccc.v .mcccc.. cmmcm. mc.cm. mc.cm. mmcc.. .cmcm. cc.c:ecz .m.ccc.c .mc.cc.v .cmccc.v .cc.cc.v .mcmcc.v ccmcm. cmccm. ccccm. m.cc.. ccccm. cc.c: ccc. cc..cc cc.c: ca..cc cc.c ..cco>c cc..cc oc.c ..eco>c meccauocmcccscoz mc.c:uucwcccz .Nmp use mom. .xo>c:m .ccwucuwmcom .mcowucz .mcwcapommaccscoz use mc.c:pomeaccz :. mmcwccmm cowcacoz can cows: com mucm_ow»$mou Hszuu.o.m u4m mccmz mc.c:pocm memo: mc.c:pocc memo: mmwuwwwmm memo: mecca“ mccmz towcwncp agoumcmpaxm uzccscoz raccEcoz z nonwage: m a: “mommcucmcca c. mcoccm cccvccgm $0 mc.c> evapomnm>c:m .ccwuzuwmcog .mcowucz .mcowmmwcmmm mmcwccmm xpczozuu...< m4mcom .mcc..c ..cm..c mm.. mcc..- m.c. cmcc.- acacocac ..cc... ..cm... cc.. c.cc.- m.c. cccc. mo>.cccocc ..cc... ..cm... mcm. mcm.. cmm. .cc.. cosmeeecc .c....c .mcc..c .cc. mmcc. ccc. c.c.. .co.co.c cc..ou cm..ou ..< cc..ou ..< cc.c: cc.c cc.c mc.ce.ce> mccoz mc.c:uocm memo: mc.caaoc: memo: mmwumnwm: mccmz mc.czu meow: townwmun zgopmzmpaxm 1::mE:oz 13:05:02 2 num.+::mz umzzwucouuu._.< mgm meow: mcwgauum$ meow: acwgauum: mcmmz mmwflmflmm: meow: mcwgap mammz -uwcwnau xgoumcmpaxm -zcmscoz u::mEcoz z -uowzcmz m m: I‘EI .I‘-.."‘\|l. Ammmmsucmgmn cw mgoggw wgmucmum we mapm> muapomngam Pmcmuaummcod Pacowumz .mcowmmmgmmm mmcwcgmm x—gzoxuu.~.< m4mgwm Ago—..v :m¢__.v “mp. emo~.- Poo. omm_.- mgmgoamA A_m__.v :omo_.v m¢_. mP~_.- “on. mmmo.- mm>wpmgmgo Amm__.v Ammop.v MNN. NNNO. mom. mmo_. swampkmgu Amo,_.v Apmp_.v _mo. mm_P.- mmo. m_¢o. _m0wgmpu gm_Pou gmp_ou _F< Lm__ou qu mg_;3 m=.m m=_m manmem> mammz mcwgzuum$ meow: mcwgzuumw meow: mm»flmflwm: mcmmz mcwgau mcmmz -uwchumu agoumcmpaxm -acmscoz uzcmE:oz z nummzcmz $ 2 |' I‘Ol'l'-.. uwzcwucounn.m.< mgm