Fringe Benefits and Inequality in the Labor Market
Chung, Wankyo, Economic Inquiry
Wage inequality has been the topic of much recent empirical research. Similarly, inequality in the distribution of fringe benefits has also been studied. (1) On the whole, the vast majority of research on inequality has examined wages and fringe benefits separately, even though the cross-sectional relationship between the two has been emphasized and analyzed in terms of the theory of equalizing differences. Rosen (1986) states, "The equalizing difference model is built upon the simple and intuitively compelling idea that it is the combination of wages and job attributes that constitute the relevant 'price' of labor for market analysis of jobs." Thus, inequality analysis based on either wages or fringe benefits alone can overlook the interaction between them and be misleading.
Given equal productivity, workers would trade fringe benefits for higher wages and vice versa. However, across different levels of productivity, skilled workers will demand more fringe benefits simply because the benefits are normal goods. Moreover, fringe benefits, unlike wages, are not taxed, and skilled workers who face higher marginal tax rates are likely to demand more fringe benefits. Overtime, one can expect the discrepancy between wages and compensation to grow, because wages represent a smaller fraction of compensation. (2)
A few studies show potential biases in measured inequality when wages are used instead of compensation. (3) However, the inequality of fringe benefits at the level of the individual workers can be analyzed by focusing on the coverage rate of health insurance and pensions without knowing the monetary value of them due to data availability. (4) Moreover, even though compensation inequality can be analyzed when the monetary values of fringe benefits are available, it can be analyzed only at the level of the industry or job, not at the level of the individual worker.
This research is different in that it uses individual-level data with imputed monetary values for fringe benefits to analyze inequality in the labor market at a point in time and its growth over time. Deeper investigation has been conducted within the distribution to ascertain how each group is affected when fringe benefits are accounted for. Several alternative explanations of the observed discrepancies between wage inequality and compensation inequality are further assessed.
The remainder of this article is organized as follows. Section II describes in detail the data used in this study. Section III supplies the empirical results, possible explanations for which are explored in section IV. Section V provides a conclusion.
The lack of suitable data has precluded analysis of compensation inequality despite widespread interest in the topic. Until now, no data set has contained information on both the demographic characteristics of workers and the monetary value of fringe benefits. However, such a data set has been created by merging individual-level data from the Current Population Survey (CPS) and industry-level data from the Chamber of Commerce's Employee Benefits Survey (BBS).
The March supplement of the CPS for 1988 to 1995 supplies information on age, gender, years of schooling, occupation of employment, industry of employment, wages, weeks worked, health insurance and pension coverage, and imputed monetary value of health insurance for the year preceding the March survey year. This annual survey is the largest source of information on fringe benefits and first included questions on health insurance and pension coverage in 1980. Although the questionnaire items regarding health insurance were changed twice (in survey years 1988 and 1995), I mitigate the problem by restricting the sample to workers ages 25 to 54, the group that the Census Bureau reports is the least likely to be affected (Gruber and Poterba, 1994) and by restricting the survey years to 1988 onward to focus on employer-provided own health insurance coverage, which is also less likely to be affected by the changes in the CPS questionnaire on health insurance coverage. (5) The other essential variable in the CPS is t he estimated value of employer contributions for health insurance, which is provided by the Census Bureau. (6)
The other data source, EBS, an annual survey of more than 900 firms, provides information on average fringe benefit costs as a percentage of payroll by industry. This survey oversamples large-sized firms (for example, in 1994, 305 out of 929 firms had fewer than 100 employees), which are more likely to provide fringe benefits. The firms in the survey together employed 2.4 million full-time equivalent workers, which makes the survey the largest of its kind and more indicative of what the typical worker was getting rather than what the typical employer was giving (U.S. Chamber of Commerce, 1995). The survey is also more indicative, considering the fact that about 80% of the workers who are covered by pensions in my CPS sample are employed in those firms with more than 100 employees, to whom data from the EBS is matched. (7)
Data from the EBS are combined with data in the CPS to estimate the pension and legally required benefits, for example Social Security, unemployment insurance, and workers' compensation. Even though an industry-average pension benefit is assigned to all the CPS workers who are employed in a particular industry, benefits as a percent of payroll, rather than absolute dollar values, are used to minimize related measurement errors, as in the work of Hersch and White-Means (1993). (8) Moreover, pension benefits have been limited to defined contribution and defined benefit plans alone, to make this variable in the EBS consistent with the coverage variable available in the CPS.
Therefore, the pension benefit is calculated as a percentage of payroll in the EBS multiplied by both the yearly earnings and the pension coverage variable in the CPS, conditional on the industry of employment. Similarly, legally required benefits are estimated by multiplying those benefits, as a percent of payroll in the EBS, by the yearly earning variables in the CPS, conditional on the industry of employment. (9) Note that this methodology restores the variance of fringe benefits among those employed in the same industry. Total compensation is estimated by adding the estimates of fringe benefits to yearly earnings, and voluntary compensation is estimated by adding the estimates of voluntary fringe benefits alone, excluding legally required benefits, to yearly earnings.
The sample is limited to workers with a strong attachment to the labor force, which allows a focus on the price structure in the labor market. The sample is white men, aged 25 to 54, who work full-time and full-year in the private sector, are not self-employed, have positive potential labor market experience, and are not living in group quarters. Furthermore, yearly earnings, top coded at the census maximum, are replaced by the estimates of the conditional mean of earnings above the top-coded value, computed using the upper 10% of data in the distribution and under the assumption that they follow a pareto distribution. (10) Finally, all of the values are deflated to 1994 values by the personal consumption expenditure deflator from the National Income and Product Accounts.
III. EMPIRICAL ANALYSIS
Levels of Inequality
Table 1 summarizes the distributions of wages, fringe benefits, and compensation in 1994. (11) Wages average $775.68 per week, and health insurance and pensions average $49.63 and $34.90 per week, respectively. Legally required benefits, such as Social Security, workers' compensation, and unemployment insurance, average $62.54. As expected, legally required benefits have the smallest variance and health insurance has the second smallest because it is less proportional to wages than are pensions and is more likely to be a costly fixed benefit. Adding up pension and health insurance benefits, voluntary fringe benefits average $84.54 per week compared to $62.54 for legally required benefits.
In terms of coverage, 76.3% of workers aged 25 to 54 and full-time and full-year private sector employees are covered by health insurance, and 59% are covered by pensions. When the sample is limited to the workers who are covered by health insurance and pensions, respectively, average health insurance benefit increases from $49.63 to $65.08 and average pension benefit from $34.90 to $59.16. Of the 76.3% of workers covered by health insurance, around 70% are also covered by pensions, and among the 59% covered by pensions about 90% also have health insurance. Therefore, 82.1% of all workers have either health insurance or pensions, but only 53.1% have both.
Because the absolute value of fringe benefits is more likely to be subject to measurement error, the average share of fringe benefits in the total compensation is also contained in Table 1. On average, health insurance constitutes 5.6% of total compensation, and pensions constitutue 3.1%. Summing up, voluntary fringe benefits constitute 8.7% of total compensation as compared to 7.5% for legally required benefits. (12) When the sample is limited to workers who are covered by each type of fringe benefit, the average share for health insurance increases to 7.4% and that for pensions increases to 5.2%. Overall, wages constitute 83.8% of total compensation, and all fringe benefits constitute 16.2%.
To indicate the effect of fringe benefits on inequality, Figure 1 provides the share of fringe benefits and voluntary fringe benefits in the total compensation across the distribution of weekly compensation in 1994. The statistics used are the mean values of the shares for each 10 equidistant intervals of the weekly compensation distribution. Both series show a smooth rising pattern, except for the upper extreme range of the compensation distribution, which means that highly skilled workers receive more fringe benefits. Specifically, voluntary fringe benefits rise from around 4% to more than 10% across the distribution, and fringe benefits rise from 13% to more than 17%. Therefore, fringe benefits appear to be so distributed to increase the level of inequality at a point in time. Unsurprisingly, as measured by the difference between the two series, the share of legally required benefits declines smoothly across all the ranges of compensation distribution, from approximately 8% to about 5%.
To clarify the individual effect of each benefit, Figure 2 shows separate graphs for health insurance and pensions. Because the benefit share can covary positively with compensation as benefits increase, separate series, one based on the wage distribution and the other based on the compensation distribution, are overlaid for comparison. As in the work of Pierce (2001), the wage-based series can be regarded as a lower bound to the true benefit share and compensation relationship, because one would expect a negative relationship between benefit share and wages. If anything, the wage-based series shows a less pronounced rising pattern in all of the figures that follow. However, the fact that it also maintains a smooth rising pattern confirms that fringe benefits are so distributed to increase inequality in the labor market.
Figure 2A shows that pension share increases by 3.9 percentage points, from about 0.7 to 4.6 across the compensation distribution. Figure 2B describes the coverage rate profile, thus illustrating that the pension coverage increases from 13.8% to 83.3%. A comparison of both parts indicates that pensions appear to increase inequality across the entire range of the compensation distribution with both coverage rate and monetary value.
The story changes, however, when the focus shifts to health insurance. Figures 2C and D show that its share rises abruptly toward the median and then decreases mildly after that. Specifically, the share increases from 3.2 to 6.6 toward the median and then decreases to 4.1 toward the end of the compensation distribution. However, its coverage rate rises from 27% to 93%. Comparison of both series indicates that the dollar value of health insurance appears to grow disproportionately with wages, thus confirming the fact that health insurance is more likely than pensions to be a fixed benefit. Thus, health insurance increases inequality in the lower range of the compensation distribution through both its value and coverage rate but decreases it in the upper range of the compensation distribution through its monetary value alone.
It should be noted that health insurance coverage is higher than pension coverage over the entire range of the compensation distribution but forms a more quadratic pattern than pensions, indicating greater income elasticity of coverage in the lower range of compensation and lower income elasticity of coverage in the upper range of compensation relative to pensions. Moreover, health insurance coverage among those who are covered by pensions increases from 57.2% to 95.3% across the compensation distribution, and the pension coverage among those who are covered by health insurance increases from 29.1% to 85.7%. Thus, the percent of workers who are covered by both health insurance and pensions rises significantly from 7.9% to 79.4% across the compensation distribution.
Table 2 quantifies the size of inequality in 1994. Statistics are calculated based on three different series. Those that are based on the wage distribution (sorted by wages) measure changes in the size of the measured inequality when variables change from wages to voluntary compensation and compensation for the same workers at particular points of the wage distribution. However, statistics that are based on the voluntary compensation distribution and those based on the compensation distribution are calculated after resorting each series.
Above all, the first column of Table 2 shows that there is significant inequality measured by wages at a point in time. For example, wages at the 90th percentile of the distribution are greater by 1.54 log points than those in the 10th percentile. (13) Within the wage distribution, inequality below the median, as measured by the 10th and median log differential (0.78), is slightly greater than that above the median, as measured by the median and 90th log differential (0.76). However, no monotonic increase in inequality is found toward the lower end of the wage distribution. For example, the 25th and median log differential (0.41) is greater than the 25th and 10th log differential (0.37).
Columns (2) and (3) of Table 2 shows the differences in measured inequality when the proxy changes from wages to voluntary compensation and compensation for the same workers controlled by wage distribution. Compensation inequality is greater than wage inequality by 0.013 log points when measured by the 90th and 10th log differential. When legally required benefits are excluded, the voluntary compensation inequality is greater than the wage inequality by 0.031 log points. Though these differences appear small, this result is due to the fact that wages underestimate inequality below the median (especially the 10th and 25th percentile differential), but overestimate inequality above the median (especially the 75th and 90th percentile differential).
Columns (4) and (5) of Table 2 provide similar statistics after resorting the data based on the voluntary compensation and compensation distributions, respectively. As expected, sorting exacerbates the size of inequality. Specifically, it increases the 10th and median differential by 0.049 log points for voluntary compensation and by 0.042 log points for compensation. Moreover, this increase is so heavily skewed toward the lower end of the distribution that a monotonic increase in inequality now appears toward the lower end of the distribution, a feature not found in the statistics based on wage distribution. However, sorting decreases the 90th and median differential by 0.018 and 0.029 log points for voluntary compensation and compensation, respectively, thus moderating the inequality above the median of each distribution.
Therefore, wages underestimate inequality in the labor market relative to compensation. Though the discrepancy appears to be small, it results from the fact that wages underestimate inequality below the median and overestimate it above the median of the distribution.
Trends in Inequality
As background, Table 1 describes the distributional changes of wages, fringe benefits, and compensation from 1987 to 1994. Average weekly wages declined slightly from $777.14 to $775.68. However, health insurance and pensions increased by 27% from $39.05 to $49.63 and from $27.49 to $34.90, respectively. When the sample is limited to those workers with both health insurance and pensions, health insurance increased by 37% (from $47.65 to $65.08) and pensions increased by 23% (from $48.24 to $59.16). Therefore, voluntary fringe benefits increased considerably during this period, and health insurance rose more than pensions. Adding these benefits to the legally required benefits, total fringe benefits increased by 20%, from $122.45 to $147.08. However, total compensation increased rather slightly as wages, the largest component of compensation, decreased and legally required benefits, the second largest, rose relatively little. Although the distributional pattern appears to remain unchanged, with health insuranc e alone having a larger median value than mean, all of the variables became more dispersed. Nonetheless, health insurance, being more equally distributed than pensions and wages, experienced the most growth of inequality, as measured by the standard deviation, which increased by 47% in comparison to 19% for wages and 18% for pensions. This means that health insurance might have played the major role in exacerbating the inequality in the labor market between 1987 and 1994.
In terms of share, the wage share in the total compensation declined from 86.3% to 83.8%. However, the health insurance share increased from 4.5% to 5.6% and the pension share also increased from 2.5% to 3.1%. When the fringe benefits are totaled, their share jumps from 13.7% to 16.2% over the eight years. As indicated, health insurance is responsible for most of this growth. Meanwhile, the dispersion of the shares increased for both wages and health insurance but declined for both pensions and legally required benefits.
In terms of coverage, health insurance coverage declined by 5.7 percentage points, but pension coverage increased by 2 percentage points. Because there is little decline in the percent of workers who were covered by both health insurance and pensions (0.6 percentage points) most of the decline in the percent of workers who were covered either by health insurance or pensions (3.1 percentage points) results from the decline in health insurance coverage. There was a 5-percentage-point decline for the workers who were covered by health insurance but not by pensions (from 28.2% to 23.2%). However, there was a rise of 2.7 percentage points for the workers covered by pensions but not by health insurance, from 3.2% to 5.9%.
In another dimension, Figure 3 describes the change in fringe benefit share from 1987 to 1994 across the entire range of the compensation distribution. Evidently, the share rose across the whole range of the compensation distribution. On average, it increased by about 2.5 percentage points, with a greater rise centered on the median range of the compensation distribution. The greater increase in fringe benefits for skilled, highly compensated workers indicates that fringe benefits changed over time to exacerbate inequality in the labor market.
Figures 4A and B decompose these changes into those in the pension share and those in the health insurance share. Figure 4A shows an overall increase in the pension share with a more significant increase around the median range of the compensation distribution. However, Figure 4B shows that the health insurance share has changed to increase inequality mainly below the median range of the compensation distribution, more so when compared with the pension share. During this period, the pension share rose an average of 0.6 percentage points and the health insurance share increased by 1.1 percentage points, reflecting rising medical costs.
Figures 4C and D provide additional information regarding how much of these changes resulted from changes in the monetary values conditional on coverage or changes in coverage rates. Although there has been little change in pension coverage, or a slight increase, health insurance coverage has declined, with a sharper decline in the lower range of the compensation distribution. Therefore, Figure 4 shows that health insurance has increased inequality mainly below the median of the compensation distribution through its change in coverage rates, but pensions have slightly increased inequality across the compensation distribution through their change in monetary values.
This inequality growth has been summarized in Table 3, which reports changes in log differentials across various points in the distributions of wages, voluntary compensation, and compensation, respectively. For comparison purposes, inequality growth has been measured for the same workers in the wage distribution using wage-sorted series in the first place. Briefly, inequality between those at the 10th percentile and those at the 90th percentile of wage distribution was measured separately for the years 1987 and 1995 by wages, voluntary compensation, and total compensation, respectively, and then their growth was measured by their differences from 1987 to 1995.
Significant inequality growth was measured without regard to the variables used. However, comparing the statistics from wages with those from voluntary compensation and compensation tells that wages slightly underestimate overall inequality growth, which was measured as increasing by 0.154 log points (from 1.385 to 1.538) by wages but by 0.16 log points for voluntary compensation and by 0.158 log points for total compensation. As expected, this small difference is due to the fact that wages underestimate inequality growth in the lower range of the wage distribution but overestimate it in the upper range of the wage distribution in relation to compensation. Specifically, wage inequality below the median of the wage distribution, as measured by the 10th and 50th percentile differentials, declined from 0.783 to 0.781, whereas compensation inequality increased from 0.806 to 0.814, conditional on the same wage distribution. In the meantime, wage inequality above the median of the wage distribution, as measured by the 50th and 90th percentile differentials, grew by 0.155 log points, whereas both voluntary compensation and compensation inequality rose by 0.150 log points. Delving further into the distribution, therefore, tells that wages underestimate inequality growth at the lower end of its distribution and overestimate inequality growth at the upper end of its distribution.
Finally, each series of wages, voluntary compensation, and total compensation was resorted and used to measure inequality growth separately. Statistics from the resorted series show that sorting exacerbates inequality growth below the median and moderates it above the median of the distribution. Therefore, though the size differs, the fact that both results--one based on series conditional on the wage distribution and the other based on resorted series--tell the same story, confirms the result.
As shown, there are significant discrepancies between wage and compensation inequality at a given point in time and in their changes overtime. This is shown to be mainly due to the disproportionately larger decline in health insurance coverage for the less skilled workers. (14) The search for an explanation should begin by analyzing the determinants of its distribution and their changes over time. (15)
To explore the role of these factors--union coverage, employment in a large firm, and income--in explaining the greater decline in health insurance coverage for the less skilled workers, Figure 5 describes the distributional changes in these factors from 1987 to 1994. (16)
Figures 5A and B show that though union coverage and employment in a large firm help in explaining the lower health insurance coverage for the less skilled workers at a point in time, their changes over time do not appear to be responsible for the larger decline of health insurance coverage for the less skilled workers, because there is no disproportionately larger decline in union coverage and employment in a large firm for them. Similar patterns result when the graphs are drawn based on the wage distribution.
However, Figure 5C shows that there has been a disproportionately greater decline in measured income either by compensation or wages for the less skilled workers. Therefore, this disproportionately greater decline in income for the less skilled, compounded by the greater income effect on their demand for health insurance, contributes to the greater decline in health insurance coverage for the less skilled workers. (17)
Another hypothesis, by Cutler and Gruber (1996), is that Medicaid expansion for the 1987-92 period crowded out private health insurance. Because Medicaid is a substitute for private health insurance, especially for the less skilled workers, this hypothesis is examined in Figure 5D. Although it suggests that the crowding out effect exists, it is limited mainly to the far lower end of the compensation distribution, thus not fully explaining the declining health insurance trend over time. (18)
It is worthwhile to note in passing that pension coverage remains almost the same except for the slight increase around the median of the compensation distribution. Even and Macpherson (1994) show that 401(k)-defined contribution plans grew significantly in the late 1980s, and these offer many of the same attractions as individual retirement accounts and also allow employees to make voluntary contributions. Though not testable, this increase in 401(k) pension plans seems to moderate the decline in pension coverage and is consistent with the increase in pension coverage for those workers who are in the median range of the compensation distribution.
The discrepancy between wage and compensation inequality has grown because wages represent a smaller fraction of compensation over time and because the trade-offs between fringe benefits and wages have changed.
At any point in time, analysis that is based exclusively on wages tends to overestimate inequality among the skilled, underestimate inequality among the less skilled, and understate inequality in the labor market. Furthermore, it also appears to understate inequality growth in the labor market, especially for the less skilled workers. Statistics that are conditional on wage distribution have yielded similar results with the difference that those from resorted series intensify the degree of the overestimation and underestimation of wages.
Although the monetary value of health insurance, which is more likely to be less variable than wages, tends to moderate inequality, especially for those workers who are covered by health insurance, the disproportionately lower coverage of less skilled workers exacerbates inequality at a given point in time. In contrast, pensions appear to increase inequality across the entire range of the compensation distribution through both coverage and monetary value. However, greater inequality growth over time results primarily from the disproportionately greater decline in health insurance coverage for the less skilled workers.
A primary explanation is that less skilled workers, who face declining income from unfavorable inequality increasing market forces, are driven away from health insurance and toward wages alone, a move that is compounded by the greater income effect on their demand for health insurance. Other possible causes, such as rising medical costs and institutional changes, remain to be studied.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
[FIGURE 5 OMITTED]
TABLE 1 Changes in the Distribution of Wages, Fringe Benefits, and Compensation 1987 1994 Mean SD Median Mean Weekly value (1994$) Wages 777.14 519.87 680.79 775.68 Health insurance (HI) 39.05 26.69 43.03 49.63 HI, the covered 47.65 21.43 48.33 65.08 Pensions (PEN) 27.49 38.93 16.27 34.90 PEN, the covered 48.24 40.73 35.73 59.16 Voluntary fringe benefits 66.54 56.81 58.82 84.54 Legally required benefits 55.91 25.37 53.40 62.54 All fringe benefits 122.45 74.33 114.24 147.08 Total compensation 899.58 577.32 799.46 922.76 Share in total compensation Wages 0.863 0.046 0.859 0.838 Health insurance (HI) 0.045 0.034 0.046 0.056 HI, the covered 0.055 0.029 0.052 0.074 Pensions (PEN) 0.025 0.028 0.024 0.031 PEN, the covered 0.044 0.023 0.041 0.052 Voluntary fringe benefits 0.070 0.046 0.074 0.087 Legally required benefits 0.067 0.016 0.065 0.075 All fringe benefits 0.137 0.046 0.141 0.162 Coverage Health Insurance 0.820 0.763 Pensions 0.570 0.590 Either HI or PEN 0.852 0.821 Both HI & PEN 0.538 0.531 No. of observations 14,410 14,085 1994 SD Median Weekly value (1994$) Wages 615.87 615.38 Health insurance (HI) 39.17 53.96 HI, the covered 31.73 65.66 Pensions (PEN) 45.91 25.92 PEN, the covered 46.24 47.31 Voluntary fringe benefits 74.29 76.44 Legally required benefits 30.60 57.50 All fringe benefits 97.84 133.25 Total compensation 700.01 756.63 Share in total compensation Wages 0.053 0.832 Health insurance (HI) 0.047 0.057 HI, the covered 0.040 0.069 Pensions (PEN) 0.028 0.040 PEN, the covered 0.013 0.049 Voluntary fringe benefits 0.057 0.097 Legally required benefits 0.013 0.076 All fringe benefits 0.053 0.168 Coverage Health Insurance Pensions Either HI or PEN Both HI & PEN No. of observations Source: March CPS and EBS, 1988 and 1995. TABLE 2 Inequality Measures Based on Log Difference in 1994 Sorted on Wages Wages Voluntary Compensation (1) (2) Standard deviation 0.6433 Percentile differential 25-10 0.372 0.402 50-25 0.409 0.418 75-50 0.384 0.381 90-75 0.373 0.367 50-10 0.781 0.820 90-50 0.757 0.748 90-10 1.538 1.569 Sorted on Sorted on Voluntary Sorted on Wages Compensation Compensation Compensation (3) (4) (5) Standard deviation 0.6525 0.6432 Percentile differential 25-10 0.398 0.420 0.413 50-25 0.416 0.410 0.411 75-50 0.379 0.381 0.378 90-75 0.358 0.358 0.350 50-10 0.814 0.831 0.823 90-50 0.737 0.739 0.728 90-10 1.551 1.570 1.552 Source: March CPS and EBS, 1995. TABLE 3 Changes in Inequality Measures Based on Log Difference Percentile Differential 1987 1988 1989 1990 1991 1992 Sorted on wages Wages 50-10 0.783 0.763 0.809 0.795 0.769 0.782 90-50 0.602 0.625 0.649 0.649 0.661 0.688 90-10 1.385 1.388 1.458 1.444 1.430 1.470 Voluntary compensation 50-10 0.809 0.788 0.838 0.820 0.789 0.820 90-50 0.599 0.620 0.641 0.645 0.664 0.680 90-10 1.408 1.408 1.479 1.465 1.454 1.500 Total compensation 50-10 0.806 0.779 0.831 0.813 0.784 0.811 90-50 0.587 0.607 0.629 0.636 0.655 0.671 90-10 1.393 1.386 1.460 1.450 1.439 1.482 Sorted on voluntary compensation 50-10 0.795 0.779 0.802 0.814 0.819 0.828 90-50 0.618 0.620 0.645 0.660 0.665 0.676 90-10 1.412 1.399 1.447 1.474 1.484 1.504 Sorted on total compensation 50-10 0.792 0.775 0.795 0.802 0.813 0.818 90-50 0.606 0.607 0.634 0.650 0.656 0.666 90-10 1.397 1.382 1.429 1.453 1.469 1.485 Standard deviation Wages 0.577 0.585 0.601 0.599 0.601 0.622 Voluntary com. 0.585 0.593 0.609 0.607 0.611 0.631 Total com. 0.578 0.583 0.600 0.599 0.603 0.622 No. of obs. 14410 13693 14968 14765 14250 14081 Percentile Differential 1993 1994 94-87 Sorted on wages Wages 50-10 0.746 0.781 - 0.001 90-50 0.737 0.757 0.155 90-10 1.483 1.538 0.154 Voluntary compensation 50-10 0.778 0.820 0.011 90-50 0.733 0.748 0.149 90-10 1.510 1.569 0.160 Total compensation 50-10 0.771 0.814 0.008 90-50 0.723 0.737 0.150 90-10 1.493 1.551 0.158 Sorted on voluntary compensation 50-10 0.819 0.831 0.036 90-50 0.713 0.739 0.121 90-10 1.532 1.570 0.157 Sorted on total compensation 50-10 0.810 0.823 0.032 90-50 0.704 0.728 0.123 90-10 1.513 1.552 0.154 Standard deviation Wages 0.625 0.643 Voluntary com. 0.635 0.652 Total com. 0.626 0.643 No. of obs. 13904 14085 Source: March CPS and EBS, 1988-1995.
(1.) Katz and Autor (1999) summarize studies on the inequality in the labor market. Fringe benefits can be divided into two basic types, according to Ehrenberg and Smith (1997): (1) those legally required, such as Social Security, unemployment insurance, and workers' compensation; and (2) voluntary fringe benefits, which include vacation pay, holiday pay, sick leave, pensions, and health insurance. Because paid vacation and holiday and leave benefits are not separable in the Current Population Survey data this article uses, health insurance and pensions alone will be treated as voluntary fringe benefits.
(2.) Gruber (2000) shows that health insurance amounted to 7.1% of compensation in 1996, and this share has grown by more than 300% over the past 30 years.
(3.) Hamermesh (1999) shows that a larger share of the burden of workplace injuries was borne by workers in the lower part of the widening distribution of earnings, so that the increase in inequality of earnings understates the increase in the inequality of compensation in terms of earnings plus workplace injuries. Pierce (2001) also finds that at the job level as defined by the sampled establishment, wage inequality tends to understate compensation inequality, and wage inequality growth appears to understate compensation inequality growth.
(4.) Bloom and Freeman (1992) and Even and Macpherson (1994) have examined the trend of pension coverage. Similarly, the trend of health insurance coverage has been examined by Farber and Levy (2000) and Currie and Yelowitz (2000).
(5.) From 1988 on, the universe of the CPS survey questionnaire on employer-provided group health insurance coverage has become those with health insurance in their own name. With these restrictions, employer-provided own insurance coverage in the CPS has been round to have a similar time series pattern to those in Currie and Yelowitz (2000), who use an alternative data source, the Survey of Income and Program Participation. Conversely, data on pension benefits in the EBS appear to be most consistent until 1994 due to changes in the size and composition of the surveys' samples, fewer firms and more small-sized firms in the more recent years of the surveys.
(6.) The procedure to estimate employer expenditures on health insurance is explained in Appendix B, Current Population Report, U.S. Bureau of the Census (1992). Data files from the 1977 National Medical Care Expenditures Survey (NMCES, a survey of individuals in households and verified through insurance companies and employers), are merged with yearly earnings and weekly hours in the CPS using variables available from both sources. The variables are: (1) type of plan (family or individual), (2) proportion of cost paid by employer (part or all), (3) level of earnings, (4) type of worker (full-time or part-time), (5) industry, (6) occupation, (7) sector (private or government), (8) region, (9) residence, and (10) personal characteristics, such as age, race, marital status, and education. The merged NMCES file was used to estimate a model that relates employer expenditures to these variables after deflating each year's earnings to 1977 dollars, then this model was run on the March CPS files to obtain estimates of employer expenditures on health insurance. The estimates were then inflated by multiplying the 1977 level estimates by the change from 1977 to each year in employer contributions per covered employer.
(7.) Moreover, this oversampling of large firms is counterbalanced by the fact that pension benefits as a percent of payroll, conditional on the firms paying the benefits in the EBS, will be biased downward to the extent that the calculation includes workers who receive no pension benefits.
(8.) The three-digit industry code in the CPS is matched with the 21 industry groups in the EBS to merge the data in the EBS into the CPS.
(9.) The Social Security benefit is calculated by multiplying the benefit share by the yearly earnings up to a maximum tax base. The maximum tax base has increased from $43,800, $45,000, S48,000, $51,300, $53,400, $55,500, and $57,600 to $60,600 from 1987 to 1994 (Rejda, 1999). Workers' compensation is calculated by multiplying the benefit share by the yearly earnings. Similarly, unemployment insurance benefit is calculated by multiplying the benefit share by the yearly earnings up to a maximum tax base of $10,000, which is the most common unemployment insurance payroll tax base, as in Smeeding (1983).
(10.) The conditional mean of the earnings above the top coded earnings is estimated by multiplying the top coded value for each year by an estimated adjusting coefficient of 1.54, 1.52, l.56, 1.53, l.50, 1.61, 1.59, and 1.66 for the years from 1987 to 1994, respectively.
(11.) Table 1 also provides their distributions for 1987, but discussion of those changes during the period is postponed to the next subsection.
(12.) These numbers look quite comparable in size to those of Pierce (2001). Using data from the fourth quarter of the Employment Cost Index of 1997, he shows that health insurance constitute 5.4% of total compensation, pensions constitute 2.8%, and legally required benefits constitute 9.4%.
(13.) Geometric means over 5 percentage points centered on the relevant percentile are used to reduce measurement error; for example the 90th percentile of wages is the geometric mean of wages between the 88th and 92nd percentiles, and the 10th percentile is the geometric mean of wages between the 8th and 12th percentiles. The interpercentile differentials are simply the natural logs of the percentile ratios, which approximate percentage changes.
(14.) The measured greater inequality growth when fringe benefits are accounted for is not imposed by the imputation method of this article. To begin with, the pension benefit to wage ratios in the EBS have become less dispersed across industries during the period (standard deviation declined from 0.035 to 0.016, with a moderate increase in mean values from 0.053 to 0.055). Moreover, when the pension benefit to wage ratios of 1995 are used instead of those of 1987 for the imputation of pension benefits, inequality of wages plus pensions in 1987 either declines from 1.426 to 1.419, when measured by the 90th-10th log differentials, or remains the same from 0.588 to 0.589, when measured by the standard deviation of the log values. However, the measured greater inequality growth will be biased downward to the extent that the imputation methodology of this article, using data from the EBS, does not capture the positive association of the pension benefit to wage ratios with wages within workers in an industry.
(15.) Factors that are related to the workers' preference for fringe benefits and the firms' cost of providing them are well summarized in Gustman et at. (1994) and Gruber (2000). When log wages and log voluntary compensation are regressed on several relevant covariates and compared using the data pooled over the years from 1987 to 1994, the benefits to wage ratios are higher for workers who are covered by a union and/or employed by large firms (100 or more employees at all locations), but they are almost the same across more or less experienced workers. The latter is mainly due to this article's industry-average imputation method of calculating pension benefits and the fact that it will understate benefit variance with experience because experience does not predict industry affiliation well. I thank an anonymous referee for bringing this to my attention.
(16.) A separate linear probability model of health insurance coverage has been estimated with log weekly wages included to account for the income effect on the demand for health insurance. The coefficient on the income measure is 0.17, biased downward reflecting the trade-off between earnings and health insurance but still indicating a significant income effect. The coefficient increases slightly only when the firm size variable, to reflect scale economies in the provision of the benefit, is excluded from the model. Moreover, when a quadratic-function of log weekly wages is used, it ranges from 0.29 for the 1st decile to 0.04 for the 10th decile of the wage distribution. Therefore, the income effect is not only significant but more so for the less skilled workers. Woodbury and Hamermesh (1992) estimated 2.85 for income elasticity and 2.91 for substitutability elasticity between wages and fringe benefits using data from the American Association of University Professors. As expected, both union coverage and es pecially employment in a large firm are found to have a significant positive effect on health insurance coverage.
(17.) Welch (1997) has explored the substitution effect from declining income for the declining labor force participation rate among the low wage earners.
(18.) Because male workers are less likely to be covered by Medicaid but may drop their employer-provided group health insurances if their dependents are eligible for Medicaid, Medicaid coverage is measured by coverage for anyone in the household. However, male workers might have kept their own employer-provided coverage while dropping the coverage of their Medicaid-eligible dependents with the Medicaid expansion, because the value of Medicaid is below that of the employer-provided group health insurance; see Cutler and Gruber (1996). For example, about 54% of male workers whose dependents were covered by Medicaid were also covered by employer-provided group health insurance in my sample in 1994. Moreover, Swartz (1997) argues that the change in the 1995 CPS question related to Medicaid might have caused more people to respond that they had such coverage. For example, the Medicaid coverage for anyone in the household increased sharply from 5% to 7.4% from 1993 to 1994, and it increased moderately from 4.1% to 5% from 1992 to 1993 in my sample.
Bloom, D. E., and R. B. Freeman. "The Fall in Private Pension Coverage in the United States." American Economic Review, 82(2), 1992, 539-45.
Currie, J., and A. Yelowitz. "Health Insurance and Less Skilled Workers," in Finding Jobs: Work and Welfare Reform, edited by D. Card and R. M. Blank. New York: Russell Sage Foundation, 2000, 233-61.
Cutler, D. M., and J. Gruber. "Does Public Insurance Crowd out Private Insurance?" Quarterly Journal of Economics, 111(2), 1996, 391-430.
Ehrenberg, R. G., and R. S. Smith, Modern Labor Economics. New York: Addison-Wesley Educational Publishers, 1997.
Even, W. E., and D. A. Macpherson. "Why Did Male Pension Coverage Decline in the 1980s?" Industrial and Labor Relations Review, 47(3), 1994, 439-53.
Farber, H. S., and H. Levy. "Recent Trends in Employer Sponsored Health Insurance Coverage: Are Bad Jobs Getting Worse?" Journal of Health Economics, 19(1), 2000, 93-119.
Gruber, J. "Health Insurance and the Labor Market," in Handbook of Health Economics, edited by A. Culyer and J. Newhouse. Amsterdam: Elsevier Science Publishing, 2000, 645-700.
Gruber, J., and J. Porterba. "Tax Incentives and the Decision to Purchase Health Insurance: Evidence from the Self-Employed." Quarterly Journal of Economics, 109(3), 1994, 701-34.
Gustman, A. L., O. S. Mitchell, and Thomas L. Steinmeier. "The Role of Pensions in the Labor Market: A Survey of the Literature." Industrial and Labor Relations Review, 47(3), 1994, 417-38.
Hamermesh, D. S. "Changing Inequality in Markets for Workplace Amenities." Quarterly Journal of Economics, 114(4), 1999, 1085-123.
Hersch, J., and S. White-Means. "Employer Sponsored Health and Pension Benefits and the Gender/Race Wage Gap." Social Science Quarterly, 74(4), 1993, 851-66.
Katz, L., and D. H. Autor. "Changes in the Wage Structure and Earnings Inequality," in Handbook of Labor Economics, edited by O. Ashenfelter, and D. Card. Amsterdam: Elsevier Science Publishing, 1999, 1463-555.
Pierce, B. "Compensation Inequality." Quarterly Journal of Economics, 116(4), 2001, 1493-525.
Rejda, G. E. Social Insurance and Economic Security. New Jersey: Prentice Hall, 1999.
Rosen, S. "The Theory of Equalizing Differences," in Handbook of Labor Economics, edited by O. Ashenfelter and R. Layard. Amsterdam: Elsevier Science Publishing, 1986, 641-92.
Smeeding, T. M. "The Size Distribution of Wage and Nonwage Compensation: Employer Cost versus Employee Value," in The Measurement of Labor Cost, edited by J. E. Triplett. Chicago: University of Chicago Press, 1983, 237-77.
Swartz, K. "Changes in the 1995 Current Population Survey and Estimates of Health Insurance Coverage." Inquiry, 34(1), 1997, 70-79.
U.S. Bureau of the Census. Current Population Reports. Washington, DC, 1992.
U.S. Chamber of Commerce. Employee Benefits. Washington, DC, 1995.
Welch, F. "Wages and Participation." Journal of Labor Economics, 15(1) part 2, 1997, S77-S103.
Woodbury, S. A., and D. S. Hamermesh`. "Taxes, Fringe Benefits and Faculty." Review of Economics and Statistics, 74(2), 1992, 287-96.
RELATED ARTICLE: ABBREVIATIONS
CPS: Current Population Survey
EBS: Employee Benefits Survey
NMCES: National Medical Care Expenditures Survey
WANKYO CHUNG *
* I would like to thank Finis Welch, Manuelita Ureta, Wayne Strayer, and two anonymous referees for their helpful comments. This paper is based on research that I undertook for my Ph.D. dissertation at Texas A&M University.
Chung: Assistant Professor, Department of Economics, National University of Singapore, 1 Arts Link, Singapore 117570. Phone 65-6874-3963, Fax 65-6775-2646, E-mail email@example.com…
Questia, a part of Gale, Cengage Learning. www.questia.com
Publication information: Article title: Fringe Benefits and Inequality in the Labor Market. Contributors: Chung, Wankyo - Author. Journal title: Economic Inquiry. Volume: 41. Issue: 3 Publication date: July 2003. Page number: 517+. © 2003 Western Economic Association International. COPYRIGHT 2003 Gale Group.
This material is protected by copyright and, with the exception of fair use, may not be further copied, distributed or transmitted in any form or by any means.