Journal Of Political Economy Fang

IS WAGE INEQUALITY LIMITED THROUGH FIRM HUMAN RESOURCE MANAGEMENT PRACTICES?1

John C. Dencker University of Illinois at Urbana-Champaign School of Labor and Employment Relations 504 E. Armory Avenue Champaign, IL 61820 Tel: (217) 333-2383 e-mail: [email protected] Chichun Fang University of Illinois at Urbana-Champaign School of Labor and Employment Relations 504 E. Armory Avenue Champaign, IL 61820 Tel: (217) 265-0954 e-mail: [email protected]

January 15, 2009

1

Submitted to Academy of Management Annual Meeting, 2009

Electronic Submission ID: 12291 IS WAGE INEQUALITY LIMITED THROUGH FIRM HUMAN RESOURCE MANAGEMENT PRACTICES?

ABSTRACT We examine the dynamics of wage patterns in a large firm, focusing on the role of HRM systems and compensation practices in shaping wage patterns and inequality through corporate restructuring. We conclude that the firm’s HRM system and its internal labor market mechanism governed the wage patterns and limited wage inequality considerably prior to the onset of restructuring. However, the firm’s institutional power on wage determination was undermined after corporate restructuring as the wage inequality within the firm increased. After restructuring, the firm became more inclined to reward high performance workers, and reduced entry-level wages for a number of workers.

Key Words: Compensation, Corporate Restructuring, Wage Inequality

1

Electronic Submission ID: 12291 INTRODUCTION The increase of wage inequality in U.S. labor market during 1980s is well documented. In brief, the wage structure polarized (Morris, Bernhardt, & Handcock, 1994), as the upper-tail (the difference between the 90th and 50th percentiles) inequality increased steadily and the lower tail (the difference between the 50th and 10th percentiles) inequality rose sharply in early 1980s but stopped increasing thereafter (Autor, Katz, & Kearney, 2008). While the increase of wage inequality in 1980s is studied early and a lot in labor economics (Autor et al., 2008; Autor, Katz, & Krueger, 1998; Borjas & Ramey, 1995; Card & DiNardo, 2002; Chay & Lee, 2000; DiNardo, Fortin, & Lemieux, 1996; Katz & Murphy, 1992; Lemieux, 2006; Murphy & Welch, 1992), the roles of the firm and its human resource policies in contributing to such increase is surprisingly not widely studied. Since wages are paid by the firm, which makes pay decisions relying on its human resource management practices and compensation policies, such dramatic change in wage inequality should be closely related with change in human resource management practices. The relationship between human resource management practices and firm performance has early been established (Huselid, 1995; Ichniowski, Shaw, & Prennushi, 1997; MacDuffie, 1995; Youndt, Snell, Dean, & Lepak, 1996). Among all the human resource practices, compensation policies have been related with various outcomes such as individual incentives (Guzzo, Jette, & Katzell, 1985; Jenkins, Gupta, Gupta, & Shaw, 1998; Stajkovic & Luthans, 1997), turnover (Shaw & Gupta, 2007; Shaw, Gupta, & Delery, 2005), and promotion (Gerhart & Milkovich, 1989; Rosenbaum, 1979), and pay dispersion is also linked to firm performances (Cowherd & Levine, 1992; Eriksson, 1999; Main, O'Reilly, & Wade, 1993; Shaw, Gupta, & Delery, 2002). However, pay dispersion is usually discussed under the framework of equity

2

Electronic Submission ID: 12291 theory and seen as a practice rather the outcome of a firm’s compensation policy (Cowherd & Levine, 1992). In this study, we view wage dispersion as the outcome of compensation policies and examine the impact of a firm’s compensation rules on its wage dynamics. More specifically, we discuss how the change in compensation policies caused by corporate restructuring influences wage patterns and distribution. We also explore how the firm’s ability to limit wage inequality is reduced through restructuring. We find systematic and consistent wage patterns within the firm before and after corporate restructuring. Cohort and tenure effects are significant; employees entered the firm in different years may receive different entry wages depending on market conditions, but the wage growth patterns conditional on respective entering wages are very similar. The effects of firm restructuring, however, varies by types of employees. For blue collar workers, the postrestructuring wage distribution is more polarized than the pre-restructuring distribution. For white collar clericals and managers, the distribution shifts rightwards and becomes more dispersed in the right tail, indicating both real wage level and wage dispersion increase after restructuring. This suggests the firm was losing its ability to control wage inequality within the firm after corporate restructuring. THE FIRM’S ROLE IN SHAPING WAGE DISTRIBUTION We argue that the firm and its HR practices play an important role in shaping the wage distribution. Studies have shown that human resource practices modify the effect of technological changes on hourly workers (Fernandez, 2001) and influence the link between performance appraisal and pay structure (Castilla, 2008). Also, the firm’s own compensation policy reflects its own internal labor market structure and will induce both cohort and tenure effects in wage patterns (Baker, Gibbs, & Holmstrom, 1994a, 1994b) and downward wage

3

Electronic Submission ID: 12291 rigidity (Gibbs & Hendricks, 2004; Harris & Holmstrom, 1982; Seltzer & Merrett, 2000). Through its internal labor market, the firm can provide employees job security and opportunities for career development, and it also shields employees from labor market fluctuations (Doeringer & Piore, 1971; Osterman, 1984). While the employees are moved among positions, the overall wage structure and wage policies of the firm are guided by the job levels and tenure effect. The firm’s compensation policies hence play the key role of shaping wage patterns and distributions and controlling wage inequality. Nevertheless, the effectiveness of internal labor market might be weakened through a sequence of deregulation and hostile acquisitions (Cappelli, 1992; Cascio, Young, & Morris, 1997) that necessitates economic efficiency in firm operations (Schleifer & Summers, 1998; Useem, 1996). Employees will be exposed to external market conditions (Cappelli et al., 1997; Eriksson & Werwatz, 2005; Lazear & Oyer, 2004), and the firm is also more inclined to reward high performance employees (Cappelli et al., 1997; Mitchell, 1989; Zenger, 1992); both will change the role of a firm in shaping the wage distribution and undermine the firm’s power to limit wage inequality. Cohort and Tenure Effects Wages can increase with tenure for several reasons, such as the accumulation of firmspecific human capital (Becker, 1975), attaining higher position in firm hierarchy (Lazear, 1995; Lazear & Rosen, 1981), or a deferred compensation used as incentives for staying in the firm (Gibbons & Waldman, 1999, 2006). In the absence of dramatic institutional changes, tenure effect should apply to all employees no matter when an employee enters the firm and when the wage data is recorded, providing the hiring pattern is consistent so the likelihood of being promoted does not vary much by year. And hence, tenure effect by itself should not change the

4

Electronic Submission ID: 12291 overall wage patterns and wage distribution, unless the composition of employees within a firm changes (as a counter example, the overall aging of baby boom generation in 1980s is believed to be one of the many reasons why the upper tail of wage distribution thickened at that time.) However, the firm might pay different entry wages to the employee cohorts who enter in different years. This is especially likely when the firm is under the economic pressure and needs to peg its entry wages to market clearing wages or when a two-tier wage structure is implemented (Martin & Peterson, 1987). Cohort effect is expected to impact the variation of wage patterns across cohorts (Baker et al., 1994b; Heckman & Robb, 1985; Lazear, 1995), even when tenure effect applies uniformly to all employees. The discrepancy in entry wages of two adjacent cohorts could still persist after more than 15 years of tenure (Baker et al., 1994b). Downward Wage Rigidity In a neo-classical economics setting where wages are set to be equal to the value of marginal products according to spot contracts, wages should fluctuate both upwards and downwards with market condition. However, implicit contract theory suggests that a risk-averse employee will accept lower life-time earnings if their annual compensation are guaranteed not to decrease during economic downturns (Azariadis, 1975; Baily, 1974; Klaas & Ullman, 1995). And hence, once being employed, wages will be “sticky” (Okun, 1981) or downward rigid (Harris & Holmstrom, 1982). On the other hand, the employer might be more inclined to adjust compensation than employment (Craig & Pencavel, 1992) to provide job security. Empirical results about downward wage rigidity are mixed: some report nominal wage cuts are not rare (Baker et al., 1994b; Treble, Gameren, Bridges, & Barmby, 2001), while some others confirm that nominal wages are downward rigid except for demotions (Gibbs & Hendricks, 2004; Seltzer & Merrett, 2000).

5

Electronic Submission ID: 12291 Downward wage rigidity, if any, provides not only job security but also a guaranteed level of income. Comparing to the potential wage loss due to involuntary turnover and loss of firm-or industry-specific human capital (Neal, 1995) that may take more than 5 years for a displaced worker to re-attain his or her pre-displacement income level (Podgursky & Swaim, 1987)1, downward wage rigidity avoids wage loss and compresses the bottom tail of wage distribution, since the very first wage offer to an employee receives will server as an anchor in his or her future wage profile. In this sense, the firm plays an important role in limiting the increase in wage inequality (especially in the lower tail of distribution) and guarantees employee's quality of life when wages are downward rigid. Promotion and Fast-trackers Although tenure and cohort effects that apply uniformly to employees who have the same tenure and entered the firm in the same year, employees are treated differently in the firm. Since wages are largely attached to job levels, the way an employer promotes/demotes employees will also influence wage patterns. Studies have shown the existence of “fast-trackers” (Baker et al., 1994a; Rosenbaum, 1979, 1984; Seltzer & Merrett, 2000; Treble et al., 2001), who are high performance workers and are promoted at a much faster pace than others. Studies also show that conditional on wages at the time t, wages at time t+1 and time t-1 are negatively correlated (Chiappori, Salanié, & Valentin, 1999), indicating employees who receive a higher raise this year are more likely to receive a higher raise in the coming years as well. Corporate Restructuring

1

Podgursky and Swaim (1987) found median losses for workers reemployed full time were not large. They did, however, find that a sizable group—most with substantial human capital investments—experienced large and enduring wage losses due to involuntary turnover. Podgursky and Swaim (1987) could not estimate the duration it took to recover the pre-displacement level earnings, because the data they used (the Displaced Worker Survey) only contained 5-year job history. For the group who suffered large wage losses, mean wages at the fifth year after displacement were still 15% less than what they would have been paid had there been no job displacement.

6

Electronic Submission ID: 12291 Mergers and hostile acquisitions that followed the deregulation in late-1970s and early1980s (Schleifer & Summers, 1998) ignited the firm’s pursuit of organizational efficiency (Useem, 1996), which in turn led to restructuring, including reduction in forces (RIFs) and changes in wage policies (Cappelli et al., 1997; Cascio et al., 1997). Firm engaging in RIF undermined job security provided under internal labor markets, and continuing employment were more influenced by market factors (Cappelli, 1992; Eriksson & Werwatz, 2005; Lazear & Oyer, 2004). Also, changing the compensation scheme to a reward or performance based system made pay more variable than it would have been in the previous seniority-based system (Cappelli et al., 1997; Mitchell, 1989). The firm will be more inclined to reward high performance employees (Cappelli et al. 1997; Mitchell 1989; Zenger 1992). And hence, firm restructuring should lead to a more wide-spread wage distribution and decreasing returns to seniority. Corporate restructuring can limit the firm’s ability to reduce wage inequality. Wage inequality within a firm is limited through cohort and tenure effects that govern the behavior of group wages, and downward rigidity of wages, if any, anchors the lower tail of wage distribution. Oppositely, a more dispersed wage distribution due to corporate restructuring that both decreases the wages of low wage earners and increases the wages of high wage earners indicates the reducing effectiveness of a firm’s internal labor market mechanism. The firm’s ability to control wage inequality within the firm is hence weakened. The firm that we study experienced several waves of restructuring. Two waves of reduction in force (RIF) were undertaken. The first RIF tool place in early- to mid-1980s during the time of hostile takeovers.2 The second RIF occurred in early 1990s following the regulatory changes that limited takeovers. Between two waves of RIF, the firm had a hiring freeze in 1986 and also transitioned its performance management and compensation system, in which a 2

The firm itself was not taken over during the time. Neither was there any evidence that it was a target of takeover.

7

Electronic Submission ID: 12291 seniority-based pay scheme was changed to a performance-based one. The firm sought to make performance goals more measurable and relevant, which pay decisions can be made upon. Additionally, the firm destroyed the performance evaluation records once promotion and pay raise decisions were made. The purpose of this was to make sure performance evaluation would not be biased by an employee’s past performances, and hence, all promotions and pay raises would have to be “re-earned” each year. ORGANIZATIONAL SETTING, DATA, AND METHODS Organizational Setting The firm that we study is a Fortune 500 firm in energy sector. Like most other large firms studied over the same period of the time (Baker et al., 1994a), this firm also had an internal labor market composed of hierarchically-ranked salary grade levels (SGLs) to which jobs and salaries were attached. Non-exempt employees (clerical, secretarial, and support staff) ranked in SGL 1 through 9, and exempted workers (managers and professionals) were in SGL 7 through 24. Additionally, roughly 25% of employees were paid on hourly basis and did not belong to a salary grade level. We use the terms “white collar clericals”, “white collar managers” and “blue collar workers” to denote these three broad categories, respectively. An employee hired as an hourly worker could be promoted to a white collar clerical later in the career, so could a white collar clerical become a white collar manager. Neither the salary grade level system nor job requirements attached to each level were significantly changed through the time of study. Employees in the firm that we study were paid relatively well. For those who were paid by hours, mean hourly wages were two to three times higher than federal minimum wages at any time in our data. The average union coverage was not high; at most 10% blue collar workers

8

Electronic Submission ID: 12291 were covered by a labor union in any given year. Table 1 shows the descriptive statistics that summarize the employee characteristics. ----------------------------Insert table 1 about here ----------------------------Data Set Career records of 25% randomly sampled U.S. employees between 1969 and 1993 are provided by the firm. We examine the change of wage patterns and distribution by three broad categories of employees: blue-collared workers (who do not belong to any salary grade level), non-exempted white collar clericals, and exempted white collar managers. Only employees hired after 1969 are included in our analysis in order to avoid potential bias caused by left censoring due to incomplete career information of employees hired before 1968 (Petersen, 1995). We only include full time employees in our analysis. In the original data set provided by the firm, a new record is added whenever there is a “career change” (such as hiring, salary change, promotion, demotion, transfer, etc.) for each employee. To transform the data into a yearly panel, we keep only the last one record for each employee in each year. If an employee has no records in a given year, his/her record in the previous year is used to fill forward in the missing year. Although we delete all records but the last one for each employee in any given year, all information regarding promotion, demotion, and bonus awarded occurred during the year are still preserved. And hence, we have a “snapshot” of all employees in the firm in the end of each year (Gibbs & Hendricks, 2004). We perform our analysis based on these end-of-year “snapshots” rather than event histories that could possibly occur at any given time during the year. Our final sample includes a total number

9

Electronic Submission ID: 12291 of 22, 187 employees: 6,555 blue collar workers (34,808 employee-year records), 9,051 white collar clericals (46,173 records), and 6,551 white collar managers (67,276 records). Dependent Variable We look at the wage patterns and wage distribution in the twenty-five year period, and hence, wage is the major dependent variable of interest. Provided in the data are the nominal annual wages, which generates some potential problem. Since the level of annual wage depends on both the level of hourly wage and the number of hours worked for those who are paid on hourly basis, we only include full time employees in our study to eliminate labor supply effect. In order to make wage distributions observed at different time comparable, all wage data are deflated to 2007 U.S. dollars. We also take the logarithm transformation when necessary. Both real wage and log of real wages in 2007 U.S. dollars are used in this study. Independent and Control Variables We use different independent variables in our analysis. The occurrence of corporate restructuring is the major independent variable in this study, as we look at how wage patterns and distributions changed through restructuring. More specifically, we focus our discussion on both yearly wage patterns and pre- and post-restructuring comparison. We also regress wage at time t+1 on wage at time t-1 controlling for wage at time t to test the existence of fast trackers. Other control variables are defined as the following. As noted, the firm has 24 formal salary grade levels, with additional 25% employees paid on hourly basis and do not belong to any salary grade level. We divide all employees into three types: blue collar workers, white collar clericals, and white collar managers, depending on whether they belong to a salary grade level, their exempt status, and which salary grade level they are in. We analyze these three types of employees separately to allow different career and

10

Electronic Submission ID: 12291 wage patterns across types. For each employee, we have his or her gender, race, level of education, and union coverage. An employee’s firm tenure and tenure at current salary grade level can be easily constructed given the nature of our data. Also, we calculate the frequency of promotions that an employee has received. This is an important proxy for job performance since the firm destroyed all performance measures once raise and promotion decisions were made. Methods The methods we use for empirical analysis are based on a set of both group- and individual-level regressions (Baker et al., 1994b) and a semi-parametric variance decomposition (DiNardo et al., 1996). We give an overview of the quantitative techniques we use here and will explain the details when we present the results. We largely follow the strategies in Baker, Gibbs, and Holmstrom (1994b) to test for cohort effects, downward wage rigidities, and fast trackers. Similar procedures for firm-level data analysis have been widely adopted in previous studies (Gibbs & Hendricks, 2004; Lin, 2005). A model suggested in Chiappori, Salanié, and Valentin (1999) is also tested. We further use a semi-parametric variance decomposition method proposed in DiNardo, Fortin and Lemieux (1996) that can be applied to the whole distribution. Comparing to Oaxaca decomposition (Oaxaca, 1973) that assumes that the effect of treatment (in this case firm restructuring) is constant across the whole wage distribution, the method we use allows different treatment effect across different segments of the wage distribution. This enables us to assess different effects of firm restructuring at various points across wage distribution. RESULTS Before proceeding to further quantitative analysis, we show a series of graphs that presents the change of wage patterns in the firm through 1969 to 1993.

11

Electronic Submission ID: 12291 -----------------------------------------Insert figures 1, 2, and 3 about here -----------------------------------------Figures 1 through 3 replicate figure 2 in Baker, Gibbs, and Holmstrom (1994b) and show the group wage patterns of cohort entered since 1969. The solid baseline in each graph connects the mean starting wage of the cohort entered in each year. There is a different line deviating from the baseline every year, each of these lines indicates the year-by-year mean wages of that specific cohort. For example, the top-most dashed line in figure 2 indicates the year-by-year mean wages of white collar clericals who entered the firm in 1969. Two observations emerged from a quick inspection of figures 1 through 3. First, there is a significant cohort effect, as huge discrepancies existed among cohorts who entered the firm in different years. A more notable trend is the decreasing of entry wages for blue collar workers between late-1970s and mid-1980s. Second, there is also a tenure effect, shown by the nearly parallel wage profiles of cohorts entered at different years, especially for white collar employees. This suggests wage growths are guided by similar rules throughout the time of study. -----------------------------------------Insert figures 4, 5, and 6 about here -----------------------------------------Figures 4 through 6 give a graphic presentation of wage dispersion, again by three types of employees. The 10th, 30th, 50th, 70th, and 90th wages percentiles in each year are shown in each graph. Overall, it is clear that the lower tail inequality (the difference between the 50th and 10th percentiles) increased dramatically for blue collar workers. Combining the evidences from figure 1, such huge increase is likely to be caused by the decreasing entry wages. Both upper tail

12

Electronic Submission ID: 12291 (the difference between the 90th and 50th percentiles) and lower tail inequality increased at similar paces for white collar clericals. For white collar managers, upper tail inequality increased faster than the lower tail side. In the following sections, we analyze the cohort and tenure effect of wages at group level, downward rigidity, promotion and faster-trackers, and effect of firm restructuring on wage distribution. Cohort and Tenure Effects We focus only on the behavior of group wages in this section. Wages at individual level are determined by other more complicated factors which will be explored in the following sections. More specifically, we cluster the employees who entered the firm in the same year and have the same firm tenure as one group and look at the mean wages of each group. It is reasonable that group wage can be influenced by when the group entered the firm (cohort effect), how long the group has stayed with the firm (tenure effect), and in which year the wage is recorded (year effect). Statistically it is impossible to test these three effects separately, since any two of these three variables (entering year, firm tenure, and year of observation) can jointly decide the other. For example, if we observe the wage in 1990 of a cohort who entered in 1980, this cohort must have stayed with the firm for 10 years. And hence, a regression equation with wages as the dependent variable and entering year, firm tenure and year of observation as the independent variables is not identifiable (Heckman & Robb, 1985). Alternatively, we follow a similar strategy as the one in Baker, Gibbs and Holmstrom (1994b) to impose a parametric constraint for tenure effect to eliminate the perfect multicollinearity among entering year, firm tenure, and year of observation. Our identification strategy is detailed in the following.

13

Electronic Submission ID: 12291 In table 2, we test cohort effects for three types of employees separately (blue collar, white collar clericals, white collar managers) using the same identification strategy. In model 1, group wages are regressed on both tenure dummies and year of observation dummies. In model 2, we remove all tenure dummies and add polynomials of firm tenure sequentially (linear, quadratic, cubic, etc.) until model 1 is no longer statistically different than model 2. Unlike the linear trend of tenure effect found in Baker, Gibbs and Holmstrom (1994b), we fit our wage data with both linear and quadratic terms of tenure for white collar workers, and all linear, quadratic, and cubic terms are used to fit the tenure trend for blue collar workers (see table 2). Consequently, model 2 for three types of workers is slightly different. Model 2 hence controls for both year and tenure effects as model 1 but uses a single trend for tenure effect rather than a full set of tenure dummy variables. ----------------------------Insert table 2 about here ----------------------------In the final step, cohort dummies are add to model 2, and the regression results are shown in model 3. Since tenure effect is controlled without using the full set of tenure dummies, the perfect multicollinearity problem is eliminated and model 3 is identifiable. Model 3 is significantly different when tested against model 2 for all three types of workers, implying a statistically significant cohort effect. When the employees entered the firm will impact their wage levels, everything else being equal. Combining with the significant polynomial of tenure terms in model 2, we conclude that both cohort and tenure effects would influence wages at group level. Downward Wage Rigidity

14

Electronic Submission ID: 12291 Starting from this section, we switch our focus to individual wages. We have shown that wages at group level are determined by cohort effect, year effect, and a uniform tenure effect. We now examine some other factors that influence wages at individual level and could have heterogeneous effects on workers in the same cohort-tenure group. We start with testing downward wage rigidity. Percentages of blue collar workers, white collar clericals, and white collar managers who received zero nominal wage increases and negative real wage increases are tabulated in tables 3 through 5, respectively. Negative nominal increases are very rare and usually results of demotions, so the frequencies of negative nominal increases are not listed in the table. ----------------------------------------Insert tables 3, 4, and 5 about here ----------------------------------------It should be first noticed that a non-trivial amount of workers receive zero nominal increases throughout the 25-year period. Blue collar workers seem to be more likely to receive a zero increase than other white collar workers. With a positive inflation rate, a zero nominal increase actually implies a decline in real wage. As the numbers in the third columns of tables 3 through 5 would indicate, real wages are anything but downward rigid. Real wage decreases occur because of two different reasons: a zero (or even negative) nominal wage increase, or a positive nominal wage increase but with a growth rate lower than inflation. A brief inspection of tables 2 through 4 lead to an observation that percentages of workers who received real wage decrease are higher and more variable than percentages of workers who received zero nominal increases. This suggests the firm did not adjust its wage structure according to inflation rate. It is evident especially in the years (1974-75 and 1979-80) when the inflation was high, as

15

Electronic Submission ID: 12291 percentage of workers who received zero nominal wage increase did not deviate much from other years but percentage of those who suffered from a real wage decline rocketed. Overall, the firm that we study did not seem to provide its employees enough protection from inflation, especially in the years of oil crisis when inflations were high. Most of workers suffered real wage declines. The last columns in tables 3 through 5 provide another unique feature of the firm’s wage policy. These numbers are the probabilities of an employee receiving a zero nominal wage increase this year conditional on him or her receiving zero nominal increase in the previous year. It is clear than an employee who received a zero nominal increase is more likely to receive zero increase as well in the following year, since the conditional probability is usually larger than the unconditional probability of receiving zero nominal increase in the same year. In other words, wage increases are awarded to a certain group of workers repetitively rather than to all workers uniformly, and these workers are more likely to be given other increases in their tenures with the firm. This leads to our next section of analysis on promotion and fast-trackers. Promotion and Fast-trackers In the previous section we argue wage increases are not uniformly awarded to all employees. An employee who received wage increase in the previous year is more likely to receive another increase in the future. In this section we show that not only the likelihood of wage increase is serially correlated but also the amount of wage increase. There are nine sets of regression output in table 6. For each type (blue collar, white collar clericals, and white collar managers) of workers we run three different regression models. In model one, wages at time t+1 is regressed on wages at time t-1, controlling for the wages at time t. In model 2, individual background and job characteristics such as gender, race, education,

16

Electronic Submission ID: 12291 union status, current salary grade level, tenure at current salary grade level, and rate of promotion are included as control variables. In model 3, the year when an employee entered the firm and the year when the wage is recorded are also added as control variables. ----------------------------Insert table 6 about here ----------------------------It is clear from table 6 that after the wage in the current period is controlled, the wage in the next period and the wage in the previous period is negatively correlated. Conditional on an employee’s current wage, the lower his/her wage in the previous year, the higher his/her wage in the next year will be. Hence, not only are wage increases given unevenly to all employees, so are the amounts of wage increases. The evidences from table 6 and last columns in tables 3 through 5 suggest there are “fast-trackers” in the firm.3 Some employees receive wage increases more likely than others, and some among those who receive wage increases more frequently receive larger raises than the rest. Effect of Corporate Restructuring In order to separate the effect of firm restructuring on wage distribution from those of other factors, we ask the question “what the pre-restructuring wage distribution would have been had employee characteristics attained the post-restructuring level.”4 A semi-parametric approach (DiNardo et al., 1996) is used to construct a “counterfactual” pre-restructuring wage distribution

3

It is possible that there is a tradeoff between the likelihood of wage increase and the amount of wage increase. The firm can award big and one-time wage raise to some employees and small and repetitive raises to others. This is also consistent with our findings in tables 3 through 6. If there were a tradeoff between likelihood and amount of wage increases, there would be no “fast-trackers” as big and one-time raise can offset small and repetitive raises. However, a simple correlation between rate of promotion and rate of wage growth suggests that there is no such tradeoff. 4 This question can be asked in an opposite way as “what the post-restructuring wage distribution would have been had employee characteristics remained at the post-restructuring level”. This, however, only influences which distribution is weighted and will not alter the final result.

17

Electronic Submission ID: 12291 in which each individual employee is weighted in a way that overall employee characteristics, such as age, gender, race, schooling, tenure, and union coverage, resemble those in the postrestructuring level. Such “counterfactual” distribution is then compared to the actual postrestructuring wage distribution to obtain the pure effect of firm restructuring on wages. And hence, the discrepancy between the “counterfactual” and the actual distributions can be interpreted as the change in wage distribution due to factors other than employee characteristics. -----------------------------------------Insert figures 7, 8, and 9 about here -----------------------------------------Figures 7 through 9 graphically show the results of DFL decomposition. We use the wage distribution in 1981 as “pre-restructuring” distribution and the one in 1990 as “postrestructuring” one. The selection of timing is somewhat arbitrary, and we choose 1981 and 1990 as the starting and end points of the whole process of corporate restructuring, which took place in 1980s. Using other start points (such as 1980 or 1982) and ending points (1989 or 1991) yields similar results. There are three panels in each graph: the dashed line in the top-left panel denotes the actual 1981 real wage distribution, and the solid line in the same panel denotes a “weighed” real wage distribution, which is the distribution that would have prevailed if the employee characteristics in 1981 were the same as those in 1990. The solid line in top-right panel is the same as the solid line in top-left panel, and the dashed line is the actual real wage distribution in 1990. Finally, the difference of the two lines in top-right panel is shown in the bottom-left panel. If there is no difference in the two lines in the top-right panel, a horizontal line at zero should be observed in the bottom-left panel; this means employees across the whole wage distribution are paid in the same way in 1981 and 1990 conditional on their characteristics. However, we do not

18

Electronic Submission ID: 12291 observe a straight line in the bottom-left panel in figure 7. We observe some positive difference when logarithm of real wage is around 10 (equivalent to $22,026 in 2007 dollars) and around 11 (equivalent to $59, 874 in 2007 dollars), and we also observe some negative difference when logarithm of real wage is between 10 and 11. This indicates, conditional on worker characteristics, less people are paid between $22,026 and $59,874 (in 2007 dollars) in 1990 compared to 1981; more people are paid around $22,026 and around $59,874 in 1990 than in 1981. Briefly, the wage distribution in 1990 is more polarized than in 1981, holding employee characteristics constant. Similar inspections in the bottom-left panels of figures 8 and 9 suggest the wage distributions for white collar workers are more skewed to the left in 1990 than in 1981—the firm is more likely to pay a higher wage than a lower wage to white collar workers when employee and job characteristics are held constant. Moreover, the discrepancy between weighted 1981 and actual 1990 distributions is larger for white collar managers (figure 8) than for white collar clericals (figure 9), implying the change in compensation system that inclines to reward top performers impacted white collar managers more than white collar clericals. Overall, the change of wage patterns for white collar workers under the semi-parametric decomposition show similar mechanisms as the well-documented skill-biased technological change (Autor et al., 1998; Fernandez, 2001). After corporate restructuring, wage inequality increased for all types of employees. DISCUSSION Summary We examine the role of the firm’s compensation practices in influencing wage patterns, wage distributions, and wage inequality. The firm can shape wage patterns and limit wage

19

Electronic Submission ID: 12291 inequality through systematic wage polices. Such ability to limit wage inequality with the firm was undermined through corporate restructuring. We find significant cohort and tenure effects. Employees entered the firm in different years receive very different wages. Conditional on entering wages, wage growth follow similar patterns no matter when the employees entered the firm. Real wages were not downward rigid, and the firm seemed not be able to keep wage growth with inflation rates especially when inflations were high. Employees who received zero nominal wages increases in the past were more likely to receive other zero nominal increases later in their career. The amount of wage raises an employee received was serially correlated, suggesting the existence of fast-trackers. A semi-parametric strategy is used to compare wage distribution before and after corporate restructuring and implementation of performance-based compensation system. Holding individual backgrounds and job characteristics consistent, higher wage earners were more likely to be paid even higher after restructuring. The effect of restructuring on low wage earners differ by types of employees—low-paid blue collar employees were paid even lower after restructuring, while the whole wage distribution shifted rightward with a flatter upper tail for white collar employees. In brief, the wage inequality increased after restructuring, indicating the firm’s power to reduce wage inequality was declining through corporate restructuring and market pressure. Limitations The issue of external validity and generalizability inevitably rises when we draw conclusions from analysis of a single firm in such a special economic context. Two reasons help strengthen our results. First, our findings are largely consistent with other studies of large firms in the same period (Gibbs & Hendricks, 2004; Lin, 2005). Second, there certainly are firm

20

Electronic Submission ID: 12291 idiosyncrasies in wage policies and how wage patterns are influenced; we do not intend to conclude what will happen in a certain context, alternatively, our findings provide possible explanations to what happened in the past and also can be used to infer what may happen under similar circumstances in the future. As noted before, the lack of performance measures in our data also handicapped our ability to quantitatively test how much of the change in the wage patterns can be attributed to the introduction of performance-based compensation. Nevertheless, we tried two indirect strategies. First, preliminary results not reported here indicate that the impact of firm tenure on wages decreased after restructuring. This confirms that wages were less dependent on firm tenure after the firm introduced a performance-based compensation system. Also, holding explanatory powers of other control variables constant, a decreasing impact of firm tenure implies an increasing impact of performance on wage determination. Alternatively, we tried to implement a proxy of performance. Assuming wage changes are caused by the change in job characteristics (such as salary grade level) and change in performances. The residual in the regression of wage changes on change of job characteristics should be composed of both changes in performance and regression errors, which then can be seen as a noisy measure of performance change. We do find the returns to this noisy measure increased after corporate restructuring. This is our second indirect way to test the effect of the new compensation system on wages. Conclusion In this study we address the changing of a firm’s role in shaping wage patterns, especially through corporate restructuring times. We conclude that, in additional to economic factors, there are institutional reasons why wage inequality increased in 1980s. Market deregulations and takeovers necessitated corporate restructuring and to some extent weakened the internal labor

21

Electronic Submission ID: 12291 market within the firm, which in turn increased the effect of market forces on wage structure. While some features of internal labor market were preserved through corporate restructuring, we also observe a more polarized wage structure. The firm was more inclined to reward high performers and pay the new hires at lower wages. We conclude that the ability of the firm to limit wage inequality through its internal labor market mechanisms was substantially weakened through restructuring, as market power prevailed and altered the firm’s wage policies. Employment outcomes relied more on market conditions than on firm rules. Although firm hierarchies still was and continued to be an important factor that determines wage structure, we suggest there are increasing individual idiosyncrasies in wage patterns and employees were less likely to shielded from market risks and chances by the firm. Our findings also have potential implications on current economic crisis. Downsizing and wage cuts happened since mid-2007 resemble what occurred to the firm we study through reduction in forces and wage restructuring in the pressure of economic efficiency. On the one hand, we would hence expect an increasing wage inequality among the workers who manage to stay with the same employer through current economic crisis. On the other hand, albeit the public opinion on the media to re-establish and re-regulate economic order, we doubt if firms can regain its ability to limit wage inequality after market forces have penetrated into firm structures for so many years.

22

Electronic Submission ID: 12291 REFERENCES Autor, D. H., Katz, L. F., & Krueger, A. B. 1998. Computing Inequality: Have Computers Changed the Labor Market? Quarterly Journal of Economics, 113(4): 1169-1213. Autor, D. H., Katz, L. F., & Kearney, M. 2008. Trends in U.S. Wage Inequality: Revising the Revisionists. The Review of Economics and Statistics, 90(2): 300-323. Azariadis, C. 1975. Implicit Contracts and Underemployment Equilibria. Journal of Political Economy, 83(6): 1183-1202. Baily, M. N. 1974. Wages and Employment under Uncertain Demand. Review of Economic Studies, 41(1): 37-50. Baker, G. F., Gibbs, M., & Holmstrom, B. 1994a. The Internal Economics of the Firm: Evidence from Personnel Data Quarterly Journal of Economics, 109(4): 881-919. Baker, G. F., Gibbs, M., & Holmstrom, B. 1994b. The Wage Policy of a Firm. Quarterly Journal of Economics, 109(4): 921-955. Becker, G. S. 1975. Human Capital. New York, NY: Columbia University Press. Borjas, G. J. & Ramey, V. 1995. Foreign Competition, Market Power and Wage Inequality. Quarterly Journal of Economics, 110(4): 1075-1110. Cappelli, P. 1992. Examining Managerial Displacement. Academy of Management Journal, 35(1): 203-217. Cappelli, P., Bassi, L., Katz, H. C., Knoke, D., Osterman, P., & Useem, M. (Eds.). 1997. Change at Work. New York, NY: Oxford University Press. Card, D. & DiNardo, J. E. 2002. Skill-Biased Technological Change and Rising Wage Inequality: Some Problems and Puzzles. Journal of Labor Economics, 20(4): 733-783.

23

Electronic Submission ID: 12291 Cascio, W. F., Young, C. E., & Morris, J. R. 1997. Financial Consequences of EmploymentChange Decisions in Major U.S. Corporations. Academy of Management Journal, 40(5): 1175-1189. Castilla, E. J. 2008. Gender, Race, and Meritocracy in Organizational Careers. American Journal of Sociology, 133(6): 1479-1526. Chay, K. Y. & Lee, D. S. 2000. Changes in Relative Wages in 1980: Returns to Observed and Unobserved Skills and Black-White Wage Differentials. Journal of Econometrics, 99(1): 1-38. Chiappori, P.-A., Salanié, B., & Valentin, J. 1999. Earlier Stayers versus Late Beginners. Journal of Political Economy, 107(4): 731-760. Cowherd, D. M. & Levine, D. I. 1992. Product Quality and Pay Equity Between Lower-Level Employees and Top Management: An Investigation of Distributive Justice Theory Administrative Science Quarterly, 37(2): 302-320. Craig, B. & Pencavel, J. 1992. The Behavior of Worker Cooperatives: The Plywood Companies of the Pacific Northwest. American Economic Review, 82(5): 1083-1105. DiNardo, J. E., Fortin, N. M., & Lemieux, T. 1996. Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach. Econometrica, 64(5): 1001-1044. Doeringer, P. B. & Piore, M. J. 1971. Internal Labor Markets and Manpower Analysis. Lexington, MA: Heath. Eriksson, T. 1999. Executive Compensation and Tournament Theory: Empirical Tests on Danish Data. Journal of Labor Economics, 17(2): 262-280.

24

Electronic Submission ID: 12291 Eriksson, T. & Werwatz, A. 2005. The Prevalence of Internal Labor Markets-- New Evidence from Panel Data. International Journal of Economics Research, 2(1): 1-20. Fernandez, R. M. 2001. Skill-Biased Technological Change and Wage Inequality: Evidence from a Plant Retooling. American Journal of Sociology, 107(2): 273-320. Gerhart, B. & Milkovich, G. T. 1989. Salaries, salary growth, and promotions for men and women in a private firm. In R. T. Michael & H. I. Hartmann & B. O'Farre (Eds.), Pay Equity: Empirical Inquiries. Washington, D.C.: National Academic Press. Gibbons, R. & Waldman, M. 1999. A Theory of Wage and Promotion Dynamics inside Firms. Quarterly Journal of Economics, 114(4): 1321-1358. Gibbons, R. & Waldman, M. 2006. Enriching a Theory of Wage and Promotion Dynamics inside Firms. Journal of Labor Economics, 24(1): 59-107. Gibbs, M. J. & Hendricks, W. 2004. Are Formal Salary Systems a Veil? Industrial and Labor Relations Review, 25(1): 71-93. Guzzo, R. A., Jette, R. D., & Katzell, R. A. 1985. The Effects of Psychologically Based Intervention Programs on Worker Productivity: A Mete-Analysis. Personnel Psychology, 38(2): 275-291. Harris, M. & Holmstrom, B. 1982. A Theory of Wage Dynamics. Review of Economic Studies, 49(3): 315-333. Heckman, J. & Robb, R. 1985. Using Longitudinal Data to Estimate Age, Period, and Cohort Effects in Earnings Equations. In W. M. Mason & S. E. Feinberg (Eds.), Cohort Analysis in Social Research: Beyond the Identification Problem. New York, NY: Springer Verlag.

25

Electronic Submission ID: 12291 Huselid, M. A. 1995. The Impact of Human Resource Management Practices on Turnover, Productivity, and Corporate Financial Performance. Academy of Management Journal, 38(3): 635-672. Ichniowski, C., Shaw, K., & Prennushi, G. 1997. The Effects of Human Resource Management Practices on Productivity: A Study of Steel Finishing Lines. American Economic Review, 87(3): 291-313. Jenkins, G. D., Gupta, N., Gupta, A., & Shaw, J. D. 1998. Are Financial Incentives Related to Performance? A Meta-Analytic Review of Empirical Research. Journal of Applied Psychology, 83(5): 777-787. Katz, L. F. & Murphy, K. M. 1992. Changes in Relative Wages, 1963-1987: Supply and Demand Factors. Quarterly Journal of Economics, 107(1): 35-78. Klaas, B. S. & Ullman, J. C. 1995. Sticky Wages Revisited: Organizational Responses to a Declining Market-Clearing Wage. Academy of Management Review, 20(2): 281-310. Lazear, E. P. & Rosen, S. 1981. Rand-Order Tournaments as Optimum Labor Contracts. Journal of Political Economy, 89(5): 841-864. Lazear, E. P. 1995. Personnel Economics. Cambridge, Massachusetts: The MIT Press. Lazear, E. P. & Oyer, P. 2004. Internal and External Labor Markets: A Personnel Economics Approach. Labour Economics, 11(5): 527-554. Lemieux, T. 2006. Increasing Residual Wage Inequality: Composition Effects, Noisy Data, or Rising Demand for Skill? American Economic Review, 96(3): 461-498. Lin, M.-J. 2005. Opening Black Box: The Internal Labor Market of Company X. Industrial Relations, 44(4): 659-706.

26

Electronic Submission ID: 12291 MacDuffie, J. P. 1995. Human Resource Bundles and Manufacturing Performance: Organizational Logic and Flexible Production Systems in The World Auto Industry. Journal of Management, 48(2): 197-221. Main, B. G. M., O'Reilly, C. A., & Wade, J. 1993. Top Executive Pay: Tournament or Teamwork? . Journal of Labor Economics, 11(4): 606-628. Martin, J. E. & Peterson, M. M. 1987. Two-Tier Wage Structures: Implication for Equity Theory. Academy of Management Journal, 30(2): 297-315. Mitchell, D. J. B. 1989. Wage Pressures and Labor Shortages: The 1960s and 1980s. Brookings Papers on Economic Activity, 1989(2): 191-232. Morris, M., Bernhardt, A. D., & Handcock, M. S. 1994. Economic Inequality: New Methods for New Trends. American Sociological Review, 59(1): 205-219. Murphy, K. M. & Welch, F. 1992. The Structure of Wages. Quarterly Journal of Economics, 107(1): 285-326. Neal, D. 1995. Industry-Specific Human Capital: Evidence from Displaced Workers. Journal of Labor Economics, 13(4): 653-677. Oaxaca, R. 1973. Male-Female Wage Differentials in Urban Labor Market. International Economic Review, 14(3): 693-709. Okun, A. M. 1981. Prices and Quantities: A Macroeconomic Analysis. Washington D.C.: The Brookings Institution. Osterman, P. 1984. Introduction: The Nature and Importance of Internal Labor Markets. In P. Osterman (Ed.), Internal Labor Markets. Cambridge, MA: The MIT Press.

27

Electronic Submission ID: 12291 Petersen, T. 1995. Analysis of Event Histories. In G. Arminger & C. C. Clogg & M. E. Sobel (Eds.), Handbook of Statistical Modeling for the Social and Behavioral Sciences: 453517. New York: Plenum Press. Podgursky, M. & Swaim, P. 1987. Job Displacement and Earnings Loss: Evidence from the Displaced Worker Survey. Industrial and Labor Relations Review, 41(1): 17-29. Rosenbaum, J. E. 1979. Tournament Mobility: Career Patterns in a Corporation. Administrative Science Quarterly, 24(1): 220-241. Rosenbaum, J. E. 1984. Career Mobility in a Corporate Hierarchy. New York, NY: Academic. Schleifer, A. & Summers, L. H. 1998. Breach of Trust in Hostile Takeovers. In A. J. Auerbach (Ed.), Corporate Takeovers: Causes and Consequences. Chicago, IL: University of Chicago Press. Seltzer, A. & Merrett, D. 2000. Personnel Policies at the Union Bank of Australia: Evidence from the 1888–1900 Entry Cohorts. Journal of Labor Economics, 18(4): 573-613. Shaw, J. D., Gupta, N., & Delery, J. E. 2002. Pay Dispersion and Work Performance: Moderating Effects of Incentives and Interdependence. Strategic Management Journal, 23(6): 491-512. Shaw, J. D., Gupta, N., & Delery, J. E. 2005. Alternative Conceptions of the Relationships between Voluntary Turnover and Organizational Performance. Academy of Management Journal, 48(1): 50-68. Shaw, J. D. & Gupta, N. 2007. Pay System Characteristics and Quit Patterns Of Good, Average, and Poor Performers. Personnel Psychology, 60(4): 903-928.

28

Electronic Submission ID: 12291 Stajkovic, A. D. & Luthans, F. 1997. A Meta-Analysis of the Effects of Organizational Behavior Modification on Task Performance, 1975-95. Academy of Management Journal, 40(5): 1122-1149. Treble, J., Gameren, E. v., Bridges, S., & Barmby, T. 2001. The Internal Economics of the Firm: Further Evidence from Personnel Data. Labour Economics, 8(5): 531-552. Useem, M. 1996. Investor Capitalism: How Money Managers Are Changing the Face of Corporate America. New York, NY: Basic Books. Youndt, M. A., Snell, S. A., Dean, J. W., & Lepak, D. P. 1996. Human Resource Management, Manufacturing Strategy, and Firm Performance. Academy of Management Journal, 39(3): 836-866. Zenger, T. R. 1992. Why Do Employers Only Reward Extreme Performance? Examining the Relationships among Performance, Pay, and Turnover. Administrative Science Quarterly, 37(2): 198-219.

29

Electronic Submission ID: 12291

Salary Paths for Each Cohort (Blue Collar Workers)

20000

30000

40000

50000

60000

Figure 1: Salary Paths by Cohort, Blue Collar Workers

1970

1975

1980

1985

1990

1995

Year

Figure 2: Salary Paths by Cohort, White Collar Clericals

30000

35000

40000

45000

50000

55000

Salary Paths for Each Cohort (White Collar Clericals)

1970

1975

1980

1985 Year

30

1990

1995

Electronic Submission ID: 12291 Figure 3: Salary Paths by Cohort, White Collar Managers

60000

80000

100000

120000

Salary Paths for Each Cohort (White Collar Managers)

1970

1975

1980

1985

1990

1995

Year

Figure 4: Wage Dispersion of Blue Collar Workers, 1969-1993

9.5

Log Real Annual Wage 10 10.5 11

Wage Dispersion of Blue Collar Workers

1970

1975

1980

1985

1990

Year 10th Percentile 50th Percentile 90th Percentile

31

30th Percentile 70th Percentile

1995

Electronic Submission ID: 12291 Figure 5: Wage Dispersion of White Collar Clericals, 1969-1993

10

Log Real Annual Wage 10.2 10.4 10.6 10.8

11

Wage Dispersion of White Collar Clericals

1970

1975

1980

1985

1990

1995

Year 10th Percentile 50th Percentile 90th Percentile

30th Percentile 70th Percentile

Figure 6: Wage Dispersion of White Collar Managers, 1969-1993

10.8

Log Real Annual Wage 11 11.2 11.4 11.6 11.8

Wage Dispersion of White Collar Managers

1970

1975

1980

1985

1990

Year 10th Percentile 50th Percentile 90th Percentile

32

30th Percentile 70th Percentile

1995

Electronic Submission ID: 12291 Figure 7: DFL Decomposition for Blue Collar Workers, 1981 versus 1990

3 0

0

1

Density 1 2

Density 2 3

4

DFL Decomposition for Blue Collar Workers

9

10 11 Log Real Wages

9

Weighted 1981

10 11 Log Real Wages 1990

12

Weighted 1981

Difference in Densities -.4 -.2 0 .2 .4

1981

12

9

10 11 Log Real Wages

12

Figure 8: DFL Decomposition for White Collar Clericals, 1981 versus 1990

1.5 Density .5 1 0

0

Density .5 1

1.5

DFL Decomposition for White Collar Clericals

9.5

10

10.5 11 Log Real Wages

9.5

Weighted 1981

9.5

10

10.5 11 Log Real Wages

10

10.5 11 Log Real Wages

1990

Difference in Densities -.4 -.2 0 .2 .4

1981

11.5

11.5

33

11.5

Weighted 1981

Electronic Submission ID: 12291 Figure 9: DFL Decomposition for White Collar Managers, 1981 versus 1990

1.5 Density .5 1 0

0

Density .5 1

1.5

DFL Decomposition for White Collar Managers

10

11 12 Log Real Wages

10

Weighted 1981

10

11 12 Log Real Wages

11 12 Log Real Wages 1990

Difference in Densities -.4 -.2 0 .2

1981

13

13

34

13

Weighted 1981

Electronic Submission ID: 12291 Table 1: Descriptive Statistics

Year 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993

Blue Collar Workers Real Tenure Wage (Year) $44,072 0.62 $45,630 1.13 $49,870 1.81 $50,899 2.51 $51,593 3.01 $49,960 3.27 $53,014 3.57 $54,422 3.73 $54,279 3.86 $55,512 4.21 $59,398 4.33 $50,749 4.41 $50,319 4.53 $51,019 5.03 $51,510 5.73 $49,928 6.19 $46,155 6.17 $45,417 6.69 $44,816 7.06 $43,748 7.08 $43,179 7.44 $43,446 7.74 $44,140 8.36 $45,180 9.14 $46,207 9.32

Age 42 42 41 41 39 36 35 34 33 33 33 32 33 33 34 35 34 35 35 35 36 36 37 38 39

White Collar Clericals Real Tenure Age Wage (Year) $35,426 0.60 31 $35,177 1.16 31 $38,045 1.81 32 $38,551 2.43 32 $38,121 2.60 31 $37,681 2.70 31 $38,302 3.28 31 $38,825 3.56 31 $37,856 3.78 31 $37,690 3.92 31 $36,807 4.01 31 $34,915 3.91 31 $35,607 3.86 31 $37,660 4.43 32 $38,589 5.05 32 $37,648 5.52 33 $37,691 5.85 34 $38,739 6.68 34 $38,551 7.05 35 $38,212 6.55 35 $37,822 6.49 35 $38,620 6.71 35 $39,557 6.95 36 $40,189 7.70 37 $40,894 7.83 37

35

White Collar Managers Real Tenure Age Wage (Year) $73,869 0.62 40 $73,926 1.38 39 $80,341 2.22 39 $82,410 3.01 39 $82,533 3.60 39 $85,092 3.96 38 $82,517 4.63 37 $84,393 4.85 37 $83,222 5.20 36 $83,174 5.42 35 $80,569 5.58 35 $75,768 5.49 34 $78,053 5.47 34 $83,540 5.96 34 $86,327 6.72 35 $86,503 7.62 35 $87,692 8.02 36 $89,830 8.84 36 $89,264 9.35 36 $89,406 9.43 37 $88,416 9.49 37 $88,265 9.70 37 $90,218 10.06 37 $91,902 10.79 38 $93,341 11.04 38

Electronic Submission ID: 12291 Table 2: Testing Cohort Effect5 Dependent Variable: Mean Wage by Entering Cohort and Tenure Groups Blue Collar Workers White Collar Clericals White Collar Managers Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 44072.1** 44072.1** 44072.1** 35426.0** 35426.0** 35426.0** 73869.5** 73869.5** 73869.5** Intercept (3017.72) (3022.17) (1922.20) (2284.89) (2249.96) (1729.68) (3803.29) (3724.93) (2288.59) 3411.10** 3573.34** 1619.23** 1534.66 3470.67** 3080.84** Tenure ---(208.06) (150.79) (72.92) (91.09) (120.73) (120.53) -214.74** -206.62** -28.38** -31.57 -46.42** -50.40** Tenure Squared ---(23.72) (15.98) (3.54) (2.81) (5.87) (3.72) 4.86** 4.94** Tenure Cubed -------(0.74) (0.50) Year of Record Yes Yes Yes Yes Yes Yes Yes Yes Yes Dummies Tenure Yes No No Yes No No Yes No No Dummies Entering Cohort No No Yes No No Yes No No Yes Dummies R-Squared Change in R2 Compared to Model 2 F-statistics Compared to Model 2

5

0.8381

0.8253

0.9348

0.8934

0.8884

0.9391

0.9453

0.9433

0.9803

0.0128

--

0.1095**

0.0050

--

0.0507**

0.0020

--

0.0370**

1.04

--

18.13

0.59

--

9.97

0.45

--

22.37

Standard errors are in the parentheses. ** if p<0.01.

36

Electronic Submission ID: 12291 Table 3: Probability of Zero or Negative Wage Increases, Blue Collar Workers Percent receiving a: Year Zero nominal increase Negative real increase 1970 13.73 49.54 1971 10.39 11.54 1972 10.33 11.96 1973 11.92 13.12 1974 11.03 71.42 1975 13.20 23.38 1976 17.20 18.26 1977 25.17 32.38 1978 19.00 22.94 1979 18.76 81.23 1980 19.28 89.92 1981 20.44 31.87 1982 21.99 26.52 1983 14.40 17.73 1984 17.25 60.45 1985 37.08 74.38 1986 50.11 53.86 1987 23.19 79.17 1988 13.52 69.35 1989 17.36 70.85 1990 12.73 40.27 1991 26.22 31.55 1992 23.26 25.64 1993 6.86 13.62

6

The probability of receiving a zero increase this year conditional on having received a zero nominal increase last year.

37

Conditional Probability6 -46.93 76.66 69.73 60.24 70.12 78.63 70.25 56.22 76.09 83.90 73.44 66.66 39.71 57.67 52.63 37.69 31.51 31.25 57.00 32.50 51.54 19.56 8.49

Electronic Submission ID: 12291 Table 4: Probability of Zero or Negative Wage Increases, White Collar Clericals Percent receiving a: Year Zero nominal increase Negative real increase 1970 11.95 42.62 1971 8.57 8.92 1972 17.06 18.90 1973 11.73 13.97 1974 8.97 39.48 1975 11.53 60.68 1976 12.56 13.27 1977 11.31 24.28 1978 9.47 16.25 1979 9.41 55.32 1980 9.28 75.20 1981 8.99 20.46 1982 8.50 10.27 1983 7.86 11.41 1984 22.79 53.02 1985 8.21 23.95 1986 13.50 15.28 1987 13.12 40.31 1988 7.72 19.32 1989 9.41 33.41 1990 6.67 23.40 1991 7.01 11.07 1992 14.15 24.46 1993 7.69 21.04

7

The probability of receiving a zero increase this year conditional on having received a zero nominal increase last year.

38

Conditional Probability7 -72.97 38.96 24.44 43.66 82.00 56.36 66.33 66.66 70.65 75.30 54.71 69.13 61.16 76.04 24.52 43.02 42.85 30.95 46.42 33.65 62.02 52.32 25.49

Electronic Submission ID: 12291 Table 5: Probability of Zero or Negative Wage Increases, White Collar Managers Percent receiving a: Year Zero nominal increase Negative real increase 1970 10.49 18.13 1971 11.59 11.99 1972 11.06 15.17 1973 11.88 15.76 1974 4.59 14.29 1975 11.21 56.75 1976 7.61 8.38 1977 6.97 16.95 1978 6.90 11.44 1979 5.34 42.58 1980 6.10 65.26 1981 5.26 13.27 1982 5.62 6.83 1983 6.57 10.32 1984 25.92 44.60 1985 4.03 15.93 1986 11.23 11.76 1987 9.75 37.84 1988 4.47 13.50 1989 6.17 33.58 1990 4.44 18.40 1991 4.49 7.28 1992 12.30 19.54 1993 5.64 16.70

8

The probability of receiving a zero increase this year conditional on having received a zero nominal increase last year.

39

Conditional Probability8 -58.06 30.76 48.93 42.22 88.00 47.85 70.00 64.76 65.76 87.00 50.00 51.35 45.31 44357 7.87 40.95 26.92 18.24 37.50 27.68 54.68 50.40 22.71

Electronic Submission ID: 12291 Table 6: Testing Fast Trackers9

Logarithm of Wage at time t-1 Logarithm of Wage at time t Individual Backgrounds10 Job Characteristics11 Entering Cohort Dummies Year of Record Dummies Number of Records

Dependent Variable: Logarithm of Wage at time t+1 Blue Collar Workers White Collar Clericals Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 -0.092** -0.076** -0.034** -0.082** -0.053** -0.037** (0.011) (0.010) (0.012) (0.009) (0.009) (0.010)

White Collar Managers Model 1 Model 2 Model 3 -0.124** -0.069** -0.057** (0.007) (0.006) (0.007)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

No

Yes

Yes

No

Yes

Yes

No

Yes

Yes

No

Yes

Yes

No

Yes

Yes

No

Yes

Yes

No

No

Yes

No

No

Yes

No

No

Yes

No

No

Yes

No

No

Yes

No

No

Yes

23,296

23,296

23,296

28,312

28,312

28,312

50,926

50,926

50,926

R-Squared

0.9390

0.9407

0.9485

0.9536

0.9544

0.9609

0.9697

0.9713

0.9758

Change in R2

--

0.0017**

0.0078**

--

0.0008**

0.0065**

--

0.0016**

0.0045**

F-statistics

--

85.09

72.80

--

34.86

79.73

--

135.57

200.47

9

Clustered Huber-White standard errors are in the parentheses. ** if p<0.01. Individual backgrounds include gender, age, race (white versus minority), job tenure (tenure in the current salary grade level; white collar clericals and white collar managers only), union status (blue collar workers and white collar clericals only), highest level of education (less than high school, high school, college, graduate degree), and rate of promotion. 11 Job characteristics include current salary grade level and starting salary grade level when hired, all controlled as dummy variables. 10

40