Fortin Econ 56 Lecture 3D III. Wage Inequality and Labour Market Institutions D. Labour Market Institutions 1. Overview 2. Effect of Minimum Wages 3. Effect of Unions on Wage Inequality
Fortin Econ 56 Lecture 3D 1. Overview Labour market institutions are another potential explanation for differences in inequality across countries. Further there have been changes over time in the strength of these institutions. Two labour market institutions that have received considerable attention in the United States, are labour unions and the federal minimum wage. In Canada, minimum wages are set at the provincial level and generate more variations. Some countries (namely the United States and the United Kingdom) experienced sizeable decline in unionization rate. These institutions and their close relatives are still quite important in many other countries.
Fortin Econ 56 Lecture 3D Collective bargaining in many European countries (and Australia) is conducted at the industry or sectoral level, and the provisions are formally or informally extended to most of the labor force. Moreover, in many countries, unions exert considerable influence on political decisions (such as minimum wages) that directly affect labor market outcomes. Union coverage has also changed much over time. Union formation really got going in the 194s, increasing steadily into the 5s and 6s. By the mid 197s, about 3% of the workforce was a member of a union in some way or another. This amount has declined considerably since 1981. The U.S. has considerably lower union coverage than Canada s this difference in coverage happened fairly recently. Up until the 197s, both Canada and the U.S. had high coverage, then the proportion of the U.S. workforce unionized fell markedly, from about 3% to 15% by 1998.
Fortin Econ 56 Lecture 3D
11 Figure 6c: Nominal Minimum Wage in the Four Largest Provinces 1 9 8 7 6 5 4 1997 1999 21 23 25 27 29 211 Quebec Ontario Alberta British Columbia 11 Figure 6d: Nominal Minimum Wage in Other Provinces 1 9 8 7 6 5 4 1997 1999 21 23 25 27 29 211 Newfoundland PEI Nova Scotia New Brunswick Manitoba Saskatchewan 33
Appendix Figure A1: The Ratio of Minimum Wages to Average Wages, Canada and the United States, 1975-21 5.45.4.35.3 1975 198 1985 199 1995 2 25 21 Canada 35 United States
Fortin Econ 56 Lecture 3D Source: OECD, ETUI
Source: Borjas (1996)
Source: Autor, Manning, and Smith (21) Figure 1A: Trends in state and federal minimum wages, and log(p1)-log(p5) Figure 1B: Trends in state and federal minimum wages, and log(p9)-log(p5) Log real minimum wage, 27 dollars 1.6 1.7 1.8 1.9 2 2.1 2.2.6.7.8.9 1 1.1 Log(p9)-log(p5) Log real minimum wage, 27 dollars 1.6 1.7 1.8 1.9 2 2.1 2.2 -.8 -.7 -.6 - -.4 -.3 -.2 Log(p1)-log(p5) 1974 198 1985 199 1995 2 25 Real value of average state/federal mins log(p1)-log(p5), female Real value of fed. min. log(p1)-log(p5), male Note: Annual data on state and federal minimum wages and log percentiles. Minimum wages are in 27 dollars. 1974 198 1985 199 1995 2 25 Real value of average state/federal mins log(p9)-log(p5), female Real value of fed. min. log(p9)-log(p5), male Note: Annual data on state and federal minimum wages and log percentiles. Minimum wages are in 27 dollars.
Source: Card, Lemieux and Riddell (24) Table 1 Union density and collective agreement coverage in selected OECD countries, 198 and 1994 Union membership as a percentage of paid workers Collective agreement coverage as a percentage of paid workers 198 1994 198 1994 Canada 36 34 37 36 United Kingdom 5 34 7 47 United States 22 16 26 18 Other Countries Australia 48 35 88 8 Austria 56 42 98 98 Belgium 56 54 9 9 Denmark 76 76 69 69 Finland 7 81 95 95 France 18 9 85 95 Germany 36 29 91 92 Italy 49 39 85 82 Japan 31 24 28 21 Netherlands 35 26 76 81 New Zealand 56 3 67 31 Norway 57 58 75 74 Portugal 61 32 7 71 Spain 9 19 76 78 Sweden 8 91 86 89 Switzerland 31 27 53 5 Source: Organisation for Economic Co-operation and Development, Employment Outlook. Paris: OECD, July 1997.
Source: Lemieux (23) Appendix Table 2: Union coverage rate by region, 1984 and 21. ---------------------------------------------------------------------------- Men Women ----------------------------------- ---------------------------------- 1986 21 1984 21 ---------------- ---------------- ---------------- ---------------- W S Central.13 W S Central.82 W S Central.89 W S Central.71 S Atlantic.164 S Atlantic.99 S Atlantic.111 S Atlantic.78 Mountain.186 Mountain.19 Mountain.125 E S Central.84 E S Central.28 E S Central.124 E S Central.127 Mountain.91 New England 7 New England.165 W N Central.152 W N Central.118 W N Central.269 W N Central.179 New England.18 New England.142 Pacific.296 Pacific.191 E N Central.195 E N Central.156 Alberta.331 Alberta.219 Pacific.221 Pacific.179 E N Central.342 E N Central.232 Middle Atlantic 8 Middle Atlantic.218 Middle Atlantic.364 Middle Atlantic 3 Ontario.315 Ontario.266 Ontario.431 Ontario.289 Alberta.34 Alberta.269 Prairies.442 Maritimes.315 Maritimes.376 Maritimes.33 Maritimes.468 Prairies.324 B.C..389 B.C..34 B.C. 3 B.C..348 Prairies.44 Quebec.377 Quebec 38 Quebec.415 Quebec.449 Prairies.381 Canada.463 Canada.323 Canada.369 Canada.311 United States.26 United States.166 United States.171 United States.133
HENRY S. FARBER and BRUCE WESTERN 46! Figure 1 Union Membership Rate. Nonagricultural Workers, 188-1998.35 -,3 - -.2 -.15 -.1 -.5 - - 188 19 192 194 I96 198 2
Fortin Econ 56 Lecture 3D 2. Effect of Minimum Wages The questions of interest are: how does the minimum wage and/or unionization affect the shape of the wage distribution including: o Overall earnings inequality o Residual earnings inequality o Inequality between observably different demographic groups: The potential importance of this set of topics was not fully appreciated prior to the work by DiNardo, Fortin and Lemieux (DFL, 1996). Figures 1 in that paper shows the density of distribution of wages by gender, particularly the piling up of women s wages at the minimum wage threshold makes the case that the minimum wage must be important for the wage distribution, particularly for women. But developing a counterfactual is intrinsically difficult.
Fortin Econ 56 Lecture 3D The minimum wage, and similarly labour unions, could affect the wage distribution by: o Displacing low wage workers. o Boosting wages of workers who were below the minimum to the minimum o Inducing wages in the uncovered sector to either rise or to fall o Causing spillover effects on the wages of workers who are close substitutes to minimum wage workers. In addition, two over-arching problems are: o The need to develop counterfactual wage densities, that is, an estimate of the entire distribution of earnings under another minimum wage or union regime. o The question of institutional endogeneity: Should we view the minimum wage as an exogenous force or an institution that is set (or influenced) by supply and demand? DFL addresses the first question but had little to say on the second. The stability of nominal federal minimum wages over the 198s in the U.S. made the second point less troublesome in that peculiar case, but more generally the role of supply and demand might certainly come into play
Fortin Econ 56 Lecture 3D o For example, Alberta was among the first province to raise its minimum wage in the mid-2s, amidst general increases in wages Before discussing this paper, we need to be familiar with Decomposition Methods. See notes on the topic. The assumptions required to form a counterfactual minimum wage estimate in DFL are quite stringent, but are conservative: 1. Minimum wages have no spillover effects on the distribution of wages above the minimum. This is a conservative assumption since any spillover effects (which are plausible slightly higher in the distribution) would augment the impact of the minimum wage. 2. The shape of the conditional density of wages at or below the minimum depends only upon the real minimum. 3. The minimum wage has no impact on employment probabilities, hence there is no need to develop counterfactual wage densities for workers who lose employment due to imposition of a binding minimum. This assumption is also conservative since removal of low wage observations from the distribution (due to job loss) would tend to further decrease inequality.
1 1973 1 1974 1 1975 1 1976 1 1977 1 1 1 1 1 ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) 1 1978 1 1979 1 198 1 1981 1 1982 1 1 1 1 1 ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) 1 1983 1 1984 1 1985 1 1986 1 1987 1 1 1 1 1 ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) 1 1988 1 1989 1 199 1 1991 1 1992 1 1 1 1 1 ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) Figure 1b. Kernel Density Estimates of Women's Real Log Wages 1973-1992 ($1979)
1 1973 1 1974 1 1975 1 1976 1 1977 ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) 1 1978 1 1979 1 198 1 1981 1 1982 ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) 1 1983 1 1984 1 1985 1 1986 1 1987 ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) 1 1988 1 1989 1 199 1 1991 1 1992 ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) ln(2) ln(5) ln(1) ln(25) Figure 1a. Kernel Density Estimates of Men's Real Log Wages 1973-1992 ($1979)
Fortin Econ 56 Lecture 3D These assumption allow DFL to graft the lower tail of the earnings distribution below the minimum wage (e.g., from 1979) directly onto another era s wage distribution (e.g., 1989) when imposing the counterfactual minimum wage. This is not entirely satisfactory, but it is difficult to improve upon. The easiest way to see the results of the DFL analysis is to study Figures 6 and 7. Key results of DFL: The decline in the minimum wage was important in the growth of lower tail inequality, especially for women The decline in unionization was a significant factor in increasing wage inequality especially for men, and corresponds to a declining middle effect. The effect of both labor market institutions explain more than a third the total change in the standard deviation for men. For males, supply and demand are likely to have been quite important to the growth of earnings inequality.
Fortin Econ 56 Lecture 3D Some remarks: o The exercise appears to have much more explanatory power for earnings inequality among women than men. This is not surprising because the minimum wage had been much more binding for women who compose a disproportionate share of the lower tail of the wage distribution. o Because the decomposition is sequential, the choice of the ordering of factors is not innocuous. o The decomposition does not deal with residual wage inequality. Despite its limitations (such as partial equilibrium analysis, assumption of exogeneity of minimum wages), the DFL results caused economists to take the minimum wage more seriously as an explanation for rising wage inequality.
Fortin Econ 56 Lecture 3D Several years later, David Lee s (1999) paper offered a potentially more compelling empirical case for its importance. Lee s idea was to use cross-state differences in the effective minimum wage to study the impact of the erosion of the minimum wage on earnings inequality. The effective minimum wage is defined as the log difference between the nominal value of the minimum wage and a state s median wage. DFL (1994) had also included an analysis of high/low minimum wage states. He then examines the cross-state relationship between the decline in the minimum wage over 1979 to 1989 and the growth in lower-tail inequality. The key identifying assumption for his approach is that cross-state variation in the effective minimum is not systematically related to underlying variation in the states wage distributions (i.e., the distribution that would prevail in the absence of a minimum wage).
Source: DiNardo, Fortin and Lemieux (1994)
Source: Lee (1999)
Fortin Econ 56 Lecture 3D Figure III which plots the 1-5 and 75-5 differentials for three high wage and three low wage states makes a compelling case that the effective minimum wage construct has empirical traction. The drop in the lower tail of the distribution is much more pronounced in low wage states, but this is not so for the 75-5 differential. If the minimum wage has no disemployment effect on workers whose latent wages w are below the minimum and has no spillover effect on workers whose wages are above the minimum, then (Figure IVa): w w = w w if mw w < w w = mw w otherwise These observations suggest a simple empirical approach: a regression of the change in state level wage inequality on the change in the state s effective minimum wage. But a practical problem immediately arises in estimating the following regression: w w = α + β mw w + ε which is that there is a potential mechanical relationship between the dependent and independent variables induced by the presence of w in both expressions.
Source: Lee (1999)
Source: Lee (1999)
Fortin Econ 56 Lecture 3D This relationship may induce an upward bias in estimates of β, giving rise to the spurious conclusion that increases in minimum wages reduce inequality (note: the dependent variable is by construction negative). So the actual estimating equation is w w = α + β mw + γ mw + ε Lee estimates that more than the entire rise of lower tail earnings inequality between 1979 and 1989 was due to the falling federal minimum wage. Problem 1: Violation of the identifying assumption. What if the median wage is correlated with latent wage dispersion? Then, states with a higher effective median will have higher 1-5 inequality and lower 9-5 inequality. Problem 2: Measurement error. While this might not be a serious problem in the crosssection, it may lead to severe biases after adding state fixed effects. Lee s approach (violation of the mean-variance independence assumption and simultaneity bias due to measurement error) are discussed and taken-on in a subsequent paper by Autor, Manning and Smith (29).
Fortin Econ 56 Lecture 3D AMS provided an updated assessment of the impact of the minimum wage on the wage distribution by using a longer panel (incorporating many additional years of data and including significantly more variation in state minimum wages) and an instrumental variables specification that purges estimates of simultaneity bias stemming from errors in variables. They estimate that 3-6% of the growth in lower tail inequality in the female wage distribution between 1979 and 1988 as measured by the differential between the log of the 5th and 1th percentiles is attributable to the decline in the real value of the minimum wage. While around one-third of the growth in male lower tail inequality over this period could be attributed to the minimum wage. In the end, their findings are very close to those of DFL.
Fortin Econ 56 Lecture 3D 3. Effect of Unions on Wage Inequality The impact of unions on the wage structure -- the way in which wages vary systematically with characteristics such as education, age, gender, or occupation has long been a topic of study among social scientists. For example, o do unions widen or narrow pay differentials between the skilled and unskilled, between men and women, or between blue-collar and white-collar workers? o Is the net effect of unions to increase or decrease overall wage inequality? The interest for this topic has been renewed in recent years as analysts have struggled to explain the rise in earnings inequality in several industrialized countries. The fact that two of the countries with the largest declines in unionization -- the US and the UK -- also experienced the biggest increases in wage inequality raises the question of whether these two phenomena are linked. o If so, how much of the growth in earnings inequality can be attributed to the fall in union coverage?
Fortin Econ 56 Lecture 3D The impact of unions on the wage structure depends on the industrial relations system -- the social, political, legal, institutional and economic environment in which unions operate. o Thus the mechanisms through which unions alter the wage structure and the magnitude of these impacts are likely to vary across countries. As usual, the difficult issue in assessing the impact of unions on the wage structure in finding an appropriate counterfactual. This can be done either (or both) by comparing changes the wage structure over time as union density decline (DiNardo, Fortin and Lemieux, 1996) by cross-country-comparison (DiNardo and Lemieux, 1997, Card, Riddell and Lemieux, 23). Most progress has arguably been made where the non-union wage structure provides a sensible benchmark for the wage structure in the absence of unions.
Fortin Econ 56 Lecture 3D In countries (such as the US, Canada, and the UK) where highly decentralized firmby-firm bargaining is the norm, the non-union wage structure can provide such an adequate counterfactual. The fact the non-union sector is relatively large also reduces concerns about union spillover effects. This contrasts with countries (such Australia and many European countries), where centralized bargaining between unions and groups of employers in an industry or region is the usual case. In decentralized economies, the impact of union on the wage structure depends on which workers are covered (self-selection, screening) by how much the union alters the pay of those who are covered. For example, we will find that for male workers, union coverage tends to be concentrated in the middle of the skill distribution and union wages tend to compress returns to skill differential relative to non-union wages, so for males, unions will have an equalizing effect on dispersion.
Fortin Econ 56 Lecture 3D For female workers, union coverage is concentrated near the top of the skill distribution and there is less compression of returns to skill differential. So for females, unions tend to increase wage dispersion. Early studies identified two channels by which this disequalizing effect was operating: o one was a between-sector effects, o the other was a hypothesized positive correlation between the union wage gain and the level of wage in the absence of unions. A break-through followed the important contribution by Freeman (198), which first laid out the two-sector framework which incorporated between and within effects. He found that unions compressed male wages within and across firms and establishments, and substantially narrowed the wage differential between blue-collar and more highly paid white-collar employees within the organized sector. o These two equalizing effects more than offset the between-sector effect that runs in the other direction.
Single Factor Analysis of Variance Let each wage W ij denote observation j in group i. Deviations from the grand mean W can be written as (W ij W ) = (W ij W i ) + ( W i W ). Taking the sum of squares, Next, expand the squared term SSS = i ( Wij W ) 2 [ = (W ij W i ) + ( W i W ) j i j [(W ij W i ) 2 + 2(W ij W i )( W i W ) + ( W i W ) 2] ] 2 SSS = i Distribute the summation operatiors SSS = (W ij W i ) 2 + 2(W ij W i )( W i W ) + i j i j i j ( W i W ) 2 Now, because deviations from the mean always sum to zero, the middle term in the above expression is equal to zero. Because j does not index anything in the third term above, we can make the following simplications: SSS = ( Wij W ) 2 = (W ij W i ) 2 + n i ( W i W ) 2, i j i j i where n i is the number of observation in group i and n = n i is the total number of observations. In population terms, the variance of wages across groups i is the sum of the within-group and between-group variances V ar(w ij ) = V ar i [W ij ] + E i [V ar( W i )] If there are only two groups, i = 2, one gets ( Wij W ) 2 = (W j1 W 1 ) 2 + (W j2 W 2 ) 2 i j j j + n 1 ( W 1 W ) 2 + n 2 ( W 2 W ) 2, Letting π = n 1 /n, since nˆσ 2 = (W ij W i ) 2, one obtains the following estimate of the variance Var(W ) = π Var(W 1 ) + (1 π) Var(W 2 ) + π(1 π) ( W 1 W 2 ) 2. j
Fortin Econ 56 Lecture 3D a. Two-Sector Model A convenient framework for analyzing the effect of unions on wage inequality is the potential outcomes model used in treatment effects literature. Assume that each worker faces two potential wages: U o W i, a log wage in the union sector, and o a log wage in the nonunion sector. NU Wi Letting U i denote union status, the observed wage for individual i is U NU W U W ( 1 U ) W (1) i i i i i Averaging across individuals would get the means of the potential wage outcomes in the two sectors, U U U U NU NU NU NU W E W ) E( W ) and W E W ) E( W ) ( i, and their corresponding variances i U V and NU V. ( i, i
Fortin Econ 56 Lecture 3D From a single factor analysis of variance (ANOVA) decomposition into a withinsector and between-sector components, we can write the overall variance of wages as U NU U NU 2 V UV (1 U) V U(1 U)[ W W ] V V V V NU NU U( V U V NU ) U(1 U )[ W U( V ) U(1 U ) W 2 U W NU ] 2 (2) So that the effect of union on the variance of wages, relative to what would prevail if all workers were paid according to the current wage structure in the non-union sector, would be V NU 2 V U( V ) U(1 U) W (3) o where the first term on the right hand side of this equation is a within-sector effect associated with the difference in wage dispersion ΔV in the two sectors o and the second term is a between sector effect, arising because unions insert a wedge ΔW between the average pay of union and nonunion workers that is always disequalizing.
Fortin Econ 56 Lecture 3D This simple model can be used to simulate the impact of changing the proportion of unionized workers, but it has limitations. One problem is that the observed variances may not correspond to the potential NU variances. For example, V may depend on the size on the union sector, if there are spill-over effects of unionism. NU In the absence of unionization we would observe V (), but in reality all we can observed in V NU (U ), where U is the average union coverage. This imply that there NU may be some bias in our estimate of the effect of union: V V (U ). b. Heterogeneous Effects by Skill Levels The simple two-sector model does not incorporate differences in the extent of union coverage or the size of the union wage effect across different workers.
Fortin Econ 56 Lecture 3D Second generation studies (DiNardo, Fortin and Lemieux, 1996, DiNardo and Lemieux, 1997 and others) begin to develop a more complete framework where union rates depend on individual characteristics. DiNardo, Fortin and Lemieux (1996) develop a reweighing methodology know as DFL and analogous the propensity score reweighing. They argue that de-unionization was a substantial factor: for men, the union density went from.317 in 1979 to.29 in 1988; for women from.17 in 1979 to.129 in 1988. Card, Lemieux and Riddell (23) provide some simple evidence on differences in unionization rate by wage rates and union/non-union wage gap by skill level, as well as changes over time. Given important differences in union density and union wag gap by skill groups, Card, Lemieux and Riddell (23) perform a variance-decomposition that address these differences across skill groups.
Effect of the Decline in the Unionization Rate onthevarianceofwages The affected group of workers comprise unionized workersandthegroupofnon-affected workers comprise non-unionized workers. Var C 88 = p 79(1 p 79 )(W I88 W N88 ) 2 + p 79 Var I88 +(1 p 79 )Var N88 where p 79 denotes the proportion of unionized workers in 1979, W I88 denotes the average wage in 1988 in the affected sector, and V I88 denotes the variance of wages in the affected sector.
1 Non-Union Union ln(2) ln(5) ln(1) ln(25) 1 Non-Union Union ln(2) ln(5) ln(1) ln(25) 1 Non-Union Union ln(2) ln(5) ln(1) ln(25)
Source: Card, Lemieux and Riddell (23) 43 Table 2 Second Generation Studies of the Impact of Unions on the Wage Structure -------------------------------------------------------------------------------------------------------------------------------------------------------------- Study Country Nature of data Findings -------------------------------------------------------------------------------------------------------------------------------------------------------------- DiNardo and US and 1981 and 1988 In 1981 unions reduced the variance of wages by 6 Lemieux (1997) Canada Men only. CPS per cent in the US and 1 per cent in Canada. data (US) and LFS In 1988 unions reduced the variance of wages by 3 data (Canada) per cent in the US and 13 per cent in Canada. Wage dispersion grew faster in the US relative to Canada for age/education groups with larger relative declines in unionization. DiNardo, Fortin US 1979-1988 Shifts in unionization explained 15-2 per cent of rising and Lemieux (1996) CPS data wage inequality for men, 3 per cent for women. Shifts in unionization explained up to one-half of the rise in the wage gap between male high school graduates and dropouts. Machin (1997) UK 1983 GHS and About 4 per cent of the rise in the variance of log wages 1991 BHPS of men was attributable to the decline in unionization. Bell and Pitt (1998) UK 1982-93 FES Approximately 2 per cent of the increase in the standard Supplemented with NCDS, deviation of log male wages during the 198s was due GHS and BHPS to declining union density. Card (21) US 1973/74 and 1993 Unionization rates fell for less educated men and women CPS data but were stable (men) or rising (women) for college graduates. Union densities rose in the public sector. Shifts in unionization explained 1-15 per cent of the rise in male wage inequality, none of the rise for women.
Source: Card, Lemieux and Riddell (23) 44 Relative shifts in unionization explained one-half or more of the greater rise in male inequality in the private sector. Gosling and US and 1983 and 1998 Unionization fell faster in the UK than the US. Shifts in Lemieux (21) UK CPS data (US), unionization explained up to one-third of the rise in male GHS and LFS wage inequality in the UK and up to 4 per cent of the rise data (UK) in male inequality in the US. Shifts in unionization explained very little of the rise in wage inequality for women in US or UK.
Source: Card, Lemieux and Riddell (23) a. Canadian Men.7 Union Coverage Rate.6.4.3.2.1 1984 1993 21. 4.71 6.5 7.77 9.97 12.81 16.44 21.12 27.11 34.81 Hourly Wage (21$, Log Scale) b. Canadian Women.7 Union Coverage Rate.6.4.3.2.1 1984 1993 21. 4.71 6.5 7.77 9.97 12.81 16.44 21.12 27.11 34.81 Hourly Wage (21$, Log Scale) Figure 2. Unionization Rate by Wage Level, Canada
Source: Card, Lemieux and Riddell (23) a. Canadian Men, 1991-95 3.3 3.1 Mean Union Wage 2.9 2.7 2 2.3 2.1 Solid line = 45 degrees Dotted line = linear fit 1.9 1.9 2.1 2.3 2 2.7 2.9 3.1 3.3 Mean Non-union Wage b. Canadian Women, 1991-95 3.3 3.1 Mean Union Wage 2.9 2.7 2 2.3 2.1 Solid line = 45 degrees Dotted line = linear fit 1.9 1.9 2.1 2.3 2 2.7 2.9 3.1 3.3 Mean Non-union Wage Figure 5: Union Relative Wage Structure in Canada, 1991-95
Source: Card, Lemieux and Riddell (23) Table 3: Effects of Unions on Wage Structure of Canadian Workers, 1984-21 1984 1991-95 21 male female male female male female Fraction Union Workers.467.369.48.353.33.317 Mean Log Wages (21$) Non-union Workers 2.658 2.365 2.661 2.452 2.728 2.495 Union Workers 2.987 2.793 2.972 2.851 2.964 2.853 Union Gap (unadjusted).33.428.311.398.236.358 Union Gap (adjusted) 1.321.24.275.153.226 Standard Deviation Log Wages: Non-union Workers 28.446 14.465 1.463 Union Workers.343.368.362.38.386.395 Union Gap -.185 -.78 -.152 -.84 -.115 -.68 Variance Decomposition: Overall Variance.231.218.233.227.229.224 Two sector model Within-sector effect - -.23 -.54 -.25 -.34 -.19 Between-sector effect.27.43.23.36.12.28 Total effect -.48.19 -.31.11 -.21.9 Model with skill groups Within-sector effect -.41 -.27 -.33 -.28 -.25 -.22 Between-sector effect.17.22.1.17.6.12 Dispersion across groups -.14.14 -.2.14.1.13 Total effect -.37.9 -.25.2 -.17.3 Sample size 17,737 15,356 17,981 18,323 24,3 23,73 Number of skill groups 25 25 25 25 25 25 Note: Samples include wage and salary workers age 15-64 with allocated hourly or weekly pay (except in 1991-95), and hourly wages between $2 and $44. per hour in 21 dollars.
Fortin Econ 56 Lecture 3D Take-Away of the Effect of Union of Wage Dispersion Unions always tend to reduce wage dispersion among union workers Whether this effect implies that unions reduce overall wage inequality depends on the relative size of the union wage gap and the union coverage rate at different points of the skill distribution Thus the anticipated effect of unions of wage inequality may vary by group of workers, or countries Among men, union coverage and union wage gap is highest in low-middle of the wage distribution, unions tend to reduce wage inequality among men Among women, where unionization is more evenly distribution, but were wage gaps are highest in the top-middle of the wage distribution, unions tend to increase wage inequality