An Examination of the Wage Productivity Gap. June 1, Nikhil Sachdev

Transcription

1 Sachdev 1 An Examination of the Wage Productivity Gap June 1, 2007 Nikhil Sachdev Department of Economics Stanford University Stanford, CA nsachdev@stanford.edu ABSTRACT Economists have been puzzled by the recent trends in real wage and productivity growth. While productivity has risen steadily, the popular press has lamented that real wages have failed to keep pace. Some argue that the inclusion of benefits explains the disparity. Others suggest that labor has lost much of its bargaining power with the firm, as evidenced by a decline in unionization rates, and as such, a lower portion of firm profits are going to pay wages and salaries. Still others have suggested that the puzzle of real wage and productivity growth might not be much of a mystery at all real wages have kept pace with productivity depending on how one measures these trends. This paper seeks to examine more closely the dynamics of real wage and productivity growth. Using data from various US government agencies, including the Bureau of Labor Statistics and the Bureau of Economic Analysis, it takes an empirical approach to determine the extent of the divergence between real wage and productivity growth and discusses factors influencing these relationships. Keywords: Wages, compensation, productivity, CPI, PCE, PPI Acknowledgements: I am deeply grateful to my thesis advisor, Professor John Taylor, for his encouragement, guidance, and patience throughout this process. It has been a great privilege to have him as an advisor and teacher. I also grateful to Professor John Pencavel and Professor Geoffrey Rothwell for their insight and support.

2 Sachdev 2 Table of Contents Introduction 3 Literature Review 4 Data and Study Design 15 Results 24 Discussion 27 Conclusion 31 References 33 Appendices 35

3 Sachdev 3 1. Introduction Recently, scholars have begun to ask more questions about trends in wages and labor productivity. The popular press has reported that the current economic expansion has a chance to become the first sustained period of economic growth since World War II that has not offered a prolonged increase in real wages for most workers. The media also claims that the median hourly wage for American workers has declined 2 percent since 2003 after accounting for inflation. These numbers are particularly onerous as real wages appear to have fallen while productivity or the amount of worker output per hour has risen steadily over this same period (Greenhouse and Leonhardt 2006). Economists generally believe that productivity growth is an avenue through which to raise living standards, and wage growth is expected to track increases in productivity (Cashell 2004). A few explanations have been offered to explain why productivity growth has not been fully realized in real wage increases. To a certain extent, stagnating wages have been offset by the rising value of benefits. So while overall compensation for most Americans has continued to increase, the rising share of benefits has altered the mix of fringe benefits and take-home pay. However, it is unlikely that benefits completely explain the disparity between changes in productivity and real wages. Economists have also suggested that although the economy continues to add jobs, global trade, immigration, layoffs, and technology as well as the insecurity caused by them have eroded the bargaining power of workers. Labor unions have been weakened and have thus been unable to demand the wages and salaries they believe should be allocated to them (Greenhouse and Leohardt 2006). Still others have suggested that the puzzle of real wage and productivity growth might not be much of a mystery at all. The extent to

4 Sachdev 4 which a wage productivity gap exists depends on how one measures wages such as the type of deflator one applies to put nominal wage values in real terms (Pencavel 2007). This paper seeks to examine more closely the dynamics of wage and productivity growth. Focusing on the period, it leverages data on labor productivity and wages from a number of domestic government agencies, including the Bureau of Labor Statistics and the Bureau of Economic Analysis. It takes an empirical approach to determine the extent the wage productivity gap and discusses factors influencing this phenomenon. 2. Literature Review It is first important to discuss the theoretical mechanism through which wage increases are expected to follow productivity increases. Cashell provides a good overview of the theory behind productivity and wage growth. He begins by considering the behavior of an individual firm operating in a competitive economy that has little influence on market conditions, sells its good at prevailing prices, and hires workers at prevailing wages. His model assumes diminishing marginal productivity each additional worker hired is less productive that those hired before. This is because it in the best interest of the firm to hire the most productive workers first, implying that each additional worker is not as capable. Also, without additional investment in capital each additional worker reduces the ratio of capital per worker. A profitmaximizing firm will continue to add to its labor force as long as the contribution to output produced by the last worker hired (the price of the good times the quantity produced) exceeds the cost of his labor (the wage rate times hours worked). Of course, if the productivity of each additional worker declines, then so will the value of the worker's additional production. The profit-maximizing firm will stop adding to its labor force when the value of the output of the last

5 Sachdev 5 worker hired equals the cost of the additional labor. However, if some event, such a technological innovation, raises the productivity of all the workers at a firm, then each worker is able to produce more than before and thus the total value of the output each worker can produce will increase. As such, the last worker hired produces more than just enough to cover the cost of his labor (Cashell 2004). If the firm continues hiring as long as the value produced by each additional worker is greater than the additional labor cost then increases in productivity will increase the firm s demand for labor as hiring more labor is profitable for the firm. And ceteris paribus, an increase in the demand for labor will tend to push up the wage rate. Therefore, increases in labor productivity increase labor income. Of course, once the firm again reaches the inflection point at which additional labor cost is more than value of the incremental goods produced, it will stop hiring (Cashell 2004). It is important to note that under some circumstances an increase in productivity might not necessarily lead to an increase in employment. Take for instance a case in which worker productivity rises faster than does the demand for the goods that the workers produce. The supply of the good being produced rises relative to the demand for it and therefore the price of the good tends to fall. This fall in price will offset the effect of higher productivity on the value of goods produced by workers. Depending on the extent of the price changes, there may be no change in the value of additional workers and the firm s demand for labor will not increase the firm neither hires more workers nor increases wages. In this case the benefits of higher productivity are captured by the consumer, who enjoys goods of the same quality and quantity but at a discounted price. However, it is also important to remember that because of these price dynamics consumers will have more to spend on all other goods and service. This will lead to an

6 Sachdev 6 increase in demand and will tend to push up the prices of those other goods. Interestingly enough, when those prices rise, the demand for labor at firms producing those goods rise, which pushes up employment and wages at those firms (Cashell 2004). It may also be the case that the gains from productivity simply increase the profits of the firm. If there is no way through which the firm can increase revenues and if wages are inflexible, the firm may reduce the number of workers it employs and thus reduce overall labor costs. In such a case productivity increases are simply captured by the firm in the form of higher profitability (Cashell 2004). A number of scholars have observed that, while in theory wage increases should match productivity increases, this has not always been the case in the United States. Research in this area has produced several explanations for the presence of a wage productivity gap and sparked a lot of debate about the best way to measure trends in labor productivity and wage growth. Cashell notes that productivity varies considerably over short periods of time, but has some notable long-run trends. For instance, productivity slowed significantly in 1973 and accelerated somewhat in 1995; this can be attributed at least in part to investment in computers and information technology. For the most part he also finds that compensation growth trends track productivity growth. However, Cashell does note that the gap between compensation growth and productivity growth increased after He explains that productivity might be outpacing compensation growth because, in theory, the rate of compensation is based on the contribution to productivity of the last worker added to the labor force. His observations of real compensation and productivity growth is based on average labor productivity and it may be that the average productivity growth in not representative of marginal productivity growth. However, he admits that this is not a satisfactory explanation as economists normally assume that average

7 Sachdev 7 and marginal productivity behave similarly (Cashell 2004). Some of have suggested that the disparity between real wage and productivity growth can be explained by the inclusion of benefits in measures of wages. Wages are different from compensation because they are usually a measure of take-home pay and do not include employer contributions for social insurance, pension contributions, and employer payments for health and other fringe benefits (Bosworth and Perry 1994). Scholars have noted that the value of employee benefits is becoming an increasingly significant portion of total compensation. Total benefits have risen from 26.8 percent to 28.9 percent between 1987 and 1994 from the employers' cost point of view (Schwabish 2004). Moreover, the BLS compares its measure of real hourly compensation to its measure of real hourly earnings of non supervisory employees. The second measure excludes employer payments for pension, health care, employment taxes, and other nonwage costs that are counted in real hourly compensation. This real hourly earnings measure has declined by a dramatic 15 percent since 1973 compared with relatively modest declines in the real hourly compensation measure, and has, subsequently, led people to attribute most of the wage and labor productivity disparity to the rising cost of wage supplements (Bosworth and Perry 1994). Scholars generally agree that over the long run, wages and salaries have been declining as a share of total compensation, as benefits have increased (Cashell 2004). However, many believe that the role of benefits is overstated in the wage productivity gap discussion. They argue that using total compensation as opposed to just take-home pay to track the relationship between wages and productivity slims the gap but it does not close it (Greenhouse and Leonhardt 2006). In addition to deciding whether to work with a measure of wages that includes the cost of benefits, many scholars have challenged the way nominal wage values are put into real terms.

8 Sachdev 8 Many commonly cited BLS figures, for instance, have been challenged because of how the data has been presented, the presence of biases in the survey of hourly earnings, and how price indices are used to deflate nominal values. Some of their figures are misleading partly because the Bureau of Labor Statistics uses the CPI to adjust hourly compensation. The CPI has experienced a major change in its construction that severely distorts wage and productivity trends. Bosworth and Perry explain that prior to 1983 all versions of the CPI measured homeownership costs by assigning a large weight to changes in the mortgage interest rates, even though such changes would affect only a small number of households. But, starting in 1983 for the CPI-U (for urban households) and in 1985 for the CPI-W (for urban workers) the BLS changed the housing component of its index to a concept of rental equivalency, in which costs to homeowners change in line with the rental rates for comparable housing. Because this changeover occurred at a cyclical peak in mortgage interest rates, the CPI, when compared with alternative indices, increases much more during the period of rising nominal interest rates the 1970s and early 1980s and does not decline during subsequent period when interest rates fell. As such, the CPI greatly overstates the rise in the price level and understates real wage gains since the late 1970s (Bosworth and Perry 1997). For this reason, Bosworth and Perry advocate using the personal consumption expenditures implicit deflator (PCE deflator) from the Bureau of Economic Analysis to measure consumer price changes this measure is conceptually equivalent to the output deflator used to measure productivity, and it provides a historically consistent measure of price changes. The PCE also has an advantage over the CPI in the way that it attaches weights to specific consumption items. For instance, the CPI only reflects out-of-pocket costs for health care whereas the PCE deflator treats employer and government payments as income transfers to

9 Sachdev 9 households and reflects the full cost of health care as consumption. In addition, the individual expenditure weights used to construct the CPI are adjusted infrequently, whereas the PCE deflator is constructed to reflect shifts in the composition of spending among major expenditure categories. Even though using the PCE deflator shrinks much of the apparent gap between compensation and labor productivity growth, Perry and Bosworth still note that using this alternate deflator does not close the gap (Bosworth and Perry 1994). Perry and Bosworth also note that labor s terms of trade the prices that workers pay relative to the prices of the products they produce has also been adversely affected. Two key developments explain the changes in labor s terms of trade or the relative rise in consumer prices compared to output prices. First, there has been a dramatic decline in the prices of computers. These goods are produced in the nonfarm business sector but are a very small part of consumption, so consumers have not captured as much benefit from the sharp drop in these prices relative to producers. The second is the dramatic rise in 1982 in the cost of owneroccupied housings. This is a large element in consumption but is excluded from nonfarm business output, and as such impacts consumers more than producers. Price indices that exclude both computers and housing costs demonstrate that these two items account for almost half of the relative price increases of consumption products since Bosworth and Perry report that of the total 7 percentage point difference between the growth in consumption prices and the growth in nonfarm output prices between 1982 and 1993, computers account for 1.8 percentage points and rent accounts for another 1.4 percentage points (Bosworth and Perry 1994). While many focus on the index of hourly wage or total compensation rates released by the BLS, Perry and Bosworth also indicate that the ECI might be a viable and interesting series to use in measuring wage and labor productivity trends. The employment cost index (ECI)

10 Sachdev 10 reports measures of hourly compensation since 1980 and wage rates since Using the ECI is advantageous because it is a fixed-weight Laspeyres index, with a June 1989 base period, that corrects for the effects of shifts in employment between high-wage and low-wage jobs. The ECI also is an index of wages per hours worked, whereas other measures are based on hours paid. This difference ends up shifting some payments from wages to benefits in the ECI. For instance, increases in vacation time and overtime premiums are reported as gains in benefits, not wages, in the ECI. Finally, compensation costs in the ECI refer only to costs associated with current employees, excluding items such as retiree health costs. One problem with the ECI is that while over the long term it relatively tracks measures of hourly compensation it diverges significantly over shorter periods. Part of this can be explained by shifts in the mix of high-wage and lowwage jobs in the economy this is known to have significant effects on average compensation. The disparity is also caused some by the greater weight placed on fringe benefit costs in the ECI than in the national accounts, which may reflect a shifting of employment toward workers with lower fringe benefit costs. As a result, there is larger discrepancy between hourly compensation and the ECI when fringe benefits are excluded. The ECI also poses a problem because it estimated aggregate wages and hours from different data sources, which is likely an important source of short run volatility in these data (Bosworth and Perry 1994). In addition to the debate over how to measure the alleged wage productivity gap, scholars have also come up with several theories as to why such a gap might exist. One explanation is that the gap between compensation and productivity growth might be partially due to a decline in unionization. In her article Unions and the Wage-Productivity Gap, Madeline Zavodny explores trends in productivity, compensation, and the unionization rate and examines whether the decline in unionization rates has contributed to the gap between compensation and

11 Sachdev 11 productivity growth. Zavodny measures trends from in productivity, using a standard measure of nonfarm business sector output per hour, and trends in compensation, using PCE deflated wages and salaries and total compensation per hour in the nonfarm business sector. She finds that overall increases in real earnings and total compensation have not kept pace with productivity gains. The gap between real hourly earnings and productivity has been widening since the 1970s, although increases in real compensation per hour have more closely matched productivity gains than have wages and salaries alone. Even still, total compensation increases have lagged behind productivity increases since the mid-1980s (Zavodny 1999). The author explains that unions typically use their bargaining power to raise wages for their members. In fact, in the 1960s, union members earned about 25 percent more than nonunion workers this number is called the union wage premium. The union wage premium has declined over the 1970s and 1980s to about 15 percent but still remains positive, and union members tend to receive higher benefits than do nonunion workers. Zavodny indicates that the decline in the unionization premium is likely linked to declining union membership, or the unionization rate, across most industries in the United States. In fact, in 1996 she explains that only about 10 percent of workers in the nonfarm business sector were union members. This fall in the unionization rate may have impeded the ability of unions to raise wages for their members (Zavodny 1999). Zavodny focuses her article on the empirical relationship between unionization and the difference between changes in productivity and changes in compensation and wages in the manufacturing sector. She tests the relationship using data from and specifically focuses on the manufacturing industry as scholars tend to place more faith in manufacturing

12 Sachdev 12 productivity numbers than measures of productivity that include output that is difficult to quantify, such as services. Her study examines changes in variables over five-year intervals so as to avoid noise contained in year-to-year differences between year-to-year growth in productivity and compensation (Zavodny 1999). Her key variables are the difference between the five-year percentage change in productivity measured as real value added divided by total hours of production workers and the five year percentage change in average hourly earnings of productions workers corrected for inflation using the PCE deflator. The compensation productivity gap is the difference between the five-year percentage change in productivity and the five-year percentage change in compensation per hour measured as compensation divided by total hours of production workers and corrected for inflation using the PCE deflator. The variables rise as productivity gains outstrip wage and compensation increases. To measure unionization Zavodny creates two variables, the unionization rate at the beginning of the five-year period and the percentage change in unionization over the five-year period interval. Her argument is that if unions exert pressure on firms to raise wages and total compensation when productivity increases, the unionization rate should be negatively associated with the wage productivity gap and the compensation productivity gap. In addition the change in the unionization rate should be negatively associated with the wage productivity gap and the compensation productivity gap as a decline in unionization is predicted to cause wage and compensation growth to be smaller than productivity growth (Zavodny 1999). Other variables controlled for in the regression include changes in employment and capital stock and the degree of international competition that an industry faces. Percentage change in total employment should be negatively associated with the wage productivity gap and

13 Sachdev 13 firms with higher levels of capital tend to have higher labor productivity, so industries with larger increases in the stock of capital may have larger increases in the wage productivity gap. Also, the degree of international competition that an industry faces may influence the wage productivity gap. Increases in the import penetration ratio, defined as the ratio of imports to total domestic shipments plus imports, are expected to be positively correlated with the wage productivity gap, and increases in the export penetration ratio, defined analogously, are expected to be negatively correlated with the wage productivity gap (Zavodny 1999). Zavodny estimates her regression, focusing on how her union related variables influence the wage productivity and compensation productivity gap, using ordinary least squares (OLS) with data on sixty-two manufacturing industries in the wage-productivity gap regression and sixty-five industries in the compensation-productivity gap regression. Her results demonstrate that the difference between productivity gains and wage or compensation increases is smaller in industries with higher initial unionization rates. The coefficients from her regression indicate that if the unionization rate at the beginning of a five-year period were to fall by 1 percentage point, the difference between productivity growth and wage growth over the next five year would increase by 0.23 percentage points. Also, a 1 percentage point decline in the initial unionization rate is associated with a 0.21 percentage point increase in the difference between productivity growth and compensation growth over the next five years. These results suggest that more-unionized industries experience smaller increases in the wage-productivity gap, and consequently, that the lower unionization rates today might account for the failure of increases in wages and compensation to match productivity growth. Yet, the wage-productivity gap and the compensation productivity gap do not rise significantly faster in industries with declining unionization rates. As such, Zavodny concludes that the decline in the unionization rate plays at

14 Sachdev 14 most a minor role in the rise in the gap between productivity and wages or compensation (Zavodny 1999). While this paper will focus on the relationship between labor productivity and wages in the US, it is interesting to also draw international comparisons in these trends. Surprisingly, growth in US real compensation has been relatively low compared to that in other industrialized countries. Comparing the US with Canada, Great Britain, France, Italy, Germany, and Japan for the real compensation per worker in the private sector rose more slowly in the United States than in any of the other six economies using either the PCE deflator or the GDP deflator to calculate real compensation. Most people regard the GDP deflator as the most appropriate measure for deflating the cost of labor across countries as it takes into account exchange rate movements. When using GDP deflators to compare wages and productivity across countries Bosworth and Perry find that compensation per employee grows at a rate of 1.2 percent slower in the US compared to other countries over the over the period and 0.8 percent slower over the period. Bosworth and Perry explain that differences in the real product wage hourly compensation deflated by the output deflator between the US and the average of the other countries can be explained by productivity growth. Their analysis shows that in comparisons of the US with other individual countries, it is only in the case of the US and Canada that productivity does not account for all or most the differential in real product wage growth (Bosworth and Perry 1994). Furthermore, Bosworth and Perry demonstrate that while the US seems to fall behind most other countries in terms of real wage increases comparing strictly the real product wage growth rates across countries sampled it is joined by many other industrialized countries in its experience of a wage productivity gap. From they show that there is little difference

15 Sachdev 15 between the growth rates of productivity and real product wages in any country. However, in the there is a divergence between labor productivity and real wage growth in Canada and Great Britain wages grew faster than productivity, but the reverse was true in all of the other industrialized countries sampled (Bosworth and Perry 1994). In thinking about wages and productivity gains it is also important to consider other measures of performance in the labor markets. For instance, while the US seems to fare worse in terms of wage increases, employment, which is another measure of employee welfare, rose faster in the United States than in any other countries between 1979 and US performance was substantially better measured on the basis of unemployment rates and changes in the ratios of employment to the working-age population than any of the seven other industrialized countries sampled except Japan. However, it is unlikely that slower job creation in other countries can be explained by the faster increase in the real compensation of their labor force rather differences in productivity trends largely explain differences in real compensation growth across countries (Bosworth and Perry 1994). 3. Data and Study Design This section discusses the data and methodology used to examine the wage productivity relationship. The literature suggests some interesting explanations for the presence of the gap employee benefits becoming an increasingly important part of total compensation, the decline in unionization, and the method of arriving at a real wage measure. This paper will briefly examine the question of whether the increasing share of benefits can explain the gap between wages and productivity. For this analysis I use the ECI data from the Bureau of Labor Statistics. The ECI is a quarterly measure of the change in the price of

16 Sachdev 16 labor, defined as compensation per employee hour worked. The ECI is closely watch by many economists as it is an indicator of cost pressures within companies that could lead to price inflation for finished goods and is for wages what the CPI is for prices (Ruser 2001). I focus on the ECI series for private industry workers from for both hourly wages and salaries and total compensation. The ECI is helpful in getting a better understanding of how benefits impact the wage productivity gap because it is one of the few measures of the cost of labor that easily distinguishes wages and salaries from the total compensation, which includes benefits such as paid leave, supplemental cash payments (overtime pay), insurance benefits, retirement and savings benefits (employer contributions to retirement and stock ownership plans), legally required benefits (Social Security, Federal and State unemployment insurance, workers compensation insurance, and Medicare), and severance pay (Ruser 2001). Moreover, from the previous literature section, we know that the ECI tends to overstate benefits, and therefore if a gap is noted between labor productivity and the ECI total compensation measure, we will have more of a case to assert that benefits cannot fully explain the gap. Zavodny does a thorough job in addressing the unionization issue and her analysis demonstrates that while a decline in unionization has occurred it cannot completely explain the presence of a wage productivity gap (Zavodny1999). As such, this paper will not expand on Zavodny s analysis and simply updates overall unionization trends (Zavodny only reports on trends in unionization up to 1999). Unionization data is collected from the Bureau of Labor Statistics and is included in Appendix A of this paper. The paper will primarily focus on examining how measuring the wage productivity gap effects the extent to which such a phenomenon exists, whether there has been a change in the

17 Sachdev 17 relationship between wages and productivity over time, and whether productivity shocks take a few periods to be adopted in the form of real wage increases. For most of my analysis I focus on total compensation, and am not necessarily concerned with comparing total compensation to take-home pay, so I abandon the ECI and adopt an annual measure of wages from the Economic Report of the President published by the Council of Economic Advisors. This measure is actually total compensation of workers per hour in the business sector and includes wages and salaries of employees plus employers contributions for social insurance and private benefit plans and an estimate of wages, salaries, and supplemental payments for the self-employed (Pencavel 2007). To measure productivity, or worker output per hour, I use a data series from the Bureau of Labor Statistics. Specifically I use non-farm business labor productivity, which is a broad based series that includes manufacturing as well as services, and is the measure that seemed most complementary to productivity measures used throughout the majority of the previous literature. To demonstrate how different deflators affect the presence of a wage productivity gap I use the Implicit Deflator for Personal Consumption Expenditures (PCE), the CPI for Urban Consumers (CPI), and Producer Price Index (PPI) to put nominal wage data into real terms. The PCE data comes from the Bureau of Economic Analysis (Commerce Department) and the CPI and PPI data are collected from the Bureau of Labor Statistics. The PCE data is reported quarterly and the CPI and PPI are reported monthly so I calculate simple averages of 4-periods and 12-periods respectively to find annual data. The following is a figure demonstrating how the PCE, CPI, and PPI have trended over time.

18 Sachdev 18 Figure 1. PRICE INDEX NUMBERS FROM 1947 to 2007, with 1947 = CPI PCE PPI Year Putting nominal wage data into real terms is simply a matter of dividing the nominal wage value in year t by the respective deflator for year t and repeating this calculation on all wage-deflator pairs in our sample, which in our case is the period To compare productivity and wage data across the board I construct indices whereby the value of the respective measure in year 1959 always equals 1.0. I also worked off of the indices to construct measures of annual percent changes in productivity and wages and scatter plots of the data. The following figures outline trends in productivity and real wage indices, present scatter plots of the data, and display annual percent changes in productivity and real wages.

19 Sachdev 19 Figure 2A. INDICES OF PRODUCTIVITY AND REAL PCE DEFLATED TOTAL HOURLY COMPENSATION, INDEX 1959 = 1.0, Productivity PCE deflated total hourly compensation Year Figure 2B. SCATTER PLOT - INDICES OF PCE DEFLATED TOTAL HOURLY COMPENSATION AND PRODUCTIVITY 2.50 PCE deflated total hourly compensatio Productivity

20 Sachdev 20 Figure 2C. ANNUAL PERCENT CHANGES IN PRODUCTIVITY AND PCE DEFLATED TOTAL HOURLY COMPENSATION, % 5.00% 4.00% 3.00% 2.00% 1.00% 0.00% -1.00% -2.00% Year Productivity PCE deflated hourly compensation Figure 3A. INDICES OF PRODUCTIVITY AND REAL CPI DEFLATED TOTAL HOURLY COMPENSATION, INDEX 1959=1.0, Productivity CPI deflated total hourly compensation Year

21 Sachdev 21 Figure 3B. SCATTER PLOT - INDICES OF CPI DEFLATED TOTAL HOURLY COMPENSATION AND PRODUCTIVITY 2.00 CPI Deflated Total Hourly Compensatio Productivity Figure 3C. ANNUAL PERCENT CHANGES IN PRODUCTIVITY AND CPI DEFLATED TOTAL HOURLY COMPENSATION, % 4.00% 3.00% 2.00% 1.00% 0.00% -1.00% -2.00% -3.00% Year Productivity CPI Deflated total hourly compensation

22 Sachdev 22 Figure 4A. INDICES OF PRODUCTIVITY AND PPI DEFLATED TOTAL HOURLY COMPENSATION, 1959 = 1.0, PPI deflated total hourly compensation 2.00 Productivity Year Figure 4B. SCATTER PLOT - INDICES OF PPI DEFLATED TOTAL HOURLY COMPENSATION AND PRODUCTIVITY PPI deflated total hourly compensatio Productivity

23 Sachdev 23 Figure 4C. ANNUAL PERCENT CHANGES IN PRODUCTIVITY AND PPI DEFLATED TOTAL HOURLY COMPENSATION, % 8.00% 6.00% 4.00% 2.00% 0.00% -2.00% -4.00% -6.00% -8.00% Year Productivity PPI deflated hourly compensation These charts give a general sense of how wages and productivity have moved over time and point to an increase in the wage productivity gap around I use regression analysis to more closely analyze how different measures of wages track productivity over different intervals with the data available. The two regressions that I employ focus on how annual percent changes in productivity impact annual percent changes in wages. The first regression is of form: PCRWAGE t = α + β 1 PCPROD t (referred to as the simple form) The second regression is of form: PCRWAGE t = α + β 1 PCPROD t + β 2 PCPROD t-1 + β 3 PCPROD t-2 (referred to as the lagged form)

24 Sachdev 24 Where PCRWAGE is the percent change in the respective real wage measure over the time period t and PCPROD is the percent change in productivity over the period t. I motivate using the lagged regression as part of an effort to understand whether increases in productivity are automatically realized in the form of real wage increases, as predicted by the simple form regression, or in fact productivity increases might need a few periods to be fully adopted in the form of real wage increases. I apply the two regression forms to the annual percent changes in the previously described measures of total hourly compensation (PCE deflated, CPI deflated, and PPI deflated) and productivity over the periods , , and The 1980 break point is significant because this seems to be the point at which many of the price indices begin to diverge and the gaps between wages and productivity start to form. Finally, to link the wages and productivity to national trends in corporate profits I delved in the National Income and Product Accounts to find data on movements in labor s share of national income and corporate profits as a share of national income. 5. Results Regarding the fringe benefits question, the figure below details trends in productivity and PCE deflated ECI for total compensation and wages and salaries.

25 Sachdev 25 Figure 5. INDICES OF PRODUCTIVITY AND PCE DEFLATED REAL ECI, UNITED STATES, 1980 = 1.0, Productivity ECI, Total compensation ECI, Wages and salaries Year The figure demonstrates there has been a gap increasing between total compensation and take-home pay. As such, fringe benefits are making up an increasing share of total compensation, and, as such, total compensation, which includes fringe benefits, is the more accurate measure of wages to use when examining the wage productivity gap. However, adopting this convention does not explain away the wage productivity gap because at least when using the ECI as a measure of the cost of labor, increases in compensation have still failed to match increases in productivity. The results of my regressions, which examine the annual percent changes in total hourly compensation and annual percent changes in productivity are presented below. All confidence intervals are at the 95% level.

26 Sachdev 26 Table 1A WAGE PRODUCTIVITY GAP SIMPLE FORM REGRESSION RESULTS PCRWAGE PCE deflated CPI deflated PPI deflated Standard Conf. Standard Conf. Standard Coeff. Error P-value Interval Coeff. Error P-value Interval Coeff. Error P-value / / Intercept / / PCPRODt / / Intercept / / PCPRODt / / Intercept / / PCPRODt Conf. Interval / / / / / / Table 1B WAGE PRODUCTIVITY GAP LAGGED FORM REGRESSION RESULTS PCRWAGE PCE deflated CPI deflated PPI deflated Standard Conf. Standard Conf. Standard Coeff. Error P-value Interval Coeff. Error P-value Interval Coeff. Error P-value / / Intercept / / PCPRODt / / PCPRODt / / PCPRODt / / Sum PCPROD / / Intercept / / PCPRODt / / PCPRODt / / PCPRODt / / Sum PCPROD / / Intercept / / PCPRODt / / PCPRODt / / PCPRODt / / Sum PCPROD Conf. Interval / / / / / / / / / / / / / / / 2.762

27 Sachdev 27 One of the challenging aspects of coming up with the regression results for the lagged form regression is working with the sum of the lagged coefficients Sum PCPROD. Professor Taylor explains that when performing a lagged regression analysis it is often useful to look at the sum of the lagged coefficients in order to assess how the lagged variable impacts the dependent variable, PCRWAGE in this case. Finding the coefficient Sum PCPROD is simply a matter of summing the coefficients for PCPROD t, PCPROD t-1, and PCPROD t-2. Determining the standard errors, however, is significantly more difficult because of the covariances between the variables. To find the variance for the summed coefficient I use the formula: VAR(PCPROD t + PCPROD t-1 + PCPROD t-2 ) = VAR(PCPROD t ) + VAR (PCPROD t-1 ) + VAR(PCPROD t-2 ) + 2*COV(PCPROD t, PCPROD t-1 ) + 2*COV(PCPROD t, PCPROD t-2 ) + 2*COV(PCPROD t-1, PCPROD t-2 ) And then make use of the formula: SE(PCPROD t + PCPROD t-1 + PCPROD t-2 ) = VAR(PCPROD t + PCPROD t-1 + PCPROD t-2 )^(1/2) / N^(1/2) After finding the coefficient Sum PCPROD and its standard error for the different lagged regression cases it is straightforward to calculate the corresponding P-values and relevant 95% confidence intervals. 5. Discussion This paper has touched on the role of fringe benefits and declining unionization rates as it affects the wage productivity gap, but has mainly focused on issues surrounding the measurement of real wage, the change in the relationship of wages and productivity over time, and how productivity growth is realized in the form of real wage increases.

28 Sachdev 28 Figures 2A, 3A, and 4A document how the deflator one uses to put nominal wage data into real terms seems to have a significant effect on the presence of a wage productivity gap. As mentioned in the discussion of the previous literature, commonly cited BLS figures about real hourly wages refer to data that has been deflated by the CPI for urban consumers, which overstate the presence of a wage-productivity gap. Looking at figure 1, which details the trends in the CPI, PCE, and the PPI, one can see that the CPI has increased the most, then the PCE, and then the PPI primarily because consumer indexes were least effected by the falling prices of computers but were impacted more so by a sharp-rise in owner occupied housing in recent years (Bosworth and Perry 1994). Price indexes that trend up relative to others tend to dampen real wage growth, and as such Figures 2A, 3A, and 4A suggest that wage productivity gap is most significant in the CPI case, followed by the PCE case, and then the PPI case. Figures 2C, 3C, and 4C show trends in annual percent changes of CPI deflated, PCE deflated, and PPI deflated total hourly compensation against annual percent changes in nonfarm business labor productivity. While from figures 2A, 3A, and 4A it seems like PPI deflated total hourly compensation better tracks labor productivity, the PPI data appears to be more volatile compared to CPI and PCE deflated total hourly compensation. The results of the regression are interesting but overall somewhat inconclusive. With regard to the relationships of the different measures of real wage with productivity it seems like the CPI deflated real wage is more explanatory than the PCE deflated real wage across all time periods and across both forms of the regression. By more explanatory I mean that the PCPROD or sum PCPROD coefficient is significantly greater than zero and more closely predicts a onefor-one percentage point increase in productivity and wages. The coefficients and summed coefficients in the lagged case for the PCE case were less than the corresponding coefficients in

29 Sachdev 29 the CPI case, although all sets of coefficients were significant. Comparing the PCE and CPI coefficients with the PPI coefficients is slightly more difficult. In the simple form case the coefficient for PCPROD in the PPI regression is greater than the corresponding coefficients in the PCE and CPI regressions across all time periods. Thus, the simple form regression results suggest that PPI deflated wages best match increases in productivity. The lagged form regression gives a more muddled picture. Essentially the sum of the lagged coefficient in the PPI regression for the time period is more explanatory than the corresponding coefficients in the PCE and CPI regressions. However, the summed coefficients for PPI for the and periods are less explanatory the corresponding measures in either the PCE or CPI deflated case. Moreover, none of the summed coefficients across any time period reported in the PPI lagged regression results is statistically significant as none of the corresponding p-values is less than 0.01, 0.05, or It is also uncertain to what extent the relationship between wages and productivity changed over the to time periods. Comparing the coefficients or summed coefficients in the CPI simple form, PPI simple form, CPI lagged form, and PPI lagged form, the coefficients in the earlier time period are more explanatory than the coefficients in the later time period, indicating that productivity increases have less of an impact on real wages in the second period. However, the coefficients of the PCE simple form and the PCE lagged form indicate that the coefficients are more explanatory in the later period than in the earlier period. To more accurately determine whether there exists some structural change in the regressions across the two time periods I tried to make use of the Chow test for structural change (the F-values from the Chow test are reported in appendix C). The test indicates that none of the F-values are significant, and therefore we are not able to assert that a structural change in the

30 Sachdev 30 relationship between the variables occurred across the two time periods. It is important to note, however, that there are some issues with using the Chow test to assess structural change. One of the assumptions of the Chow test is that there is no heteroskedacity in the data, meaning that the error variances of the data across the two time periods are roughly constant. Because heteroskedacity does exist in our data, it might have been advantageous to use some sort of modified Chow test, as advocated by Gujarati, as well to test for structural change (Gujarati 2005). Finally, the regression results seem to demonstrate that lags might be important in understanding the adoption of productivity increases into real wages, but, again, the results are not entirely conclusive. Comparing the simple form coefficients with the lagged form coefficients in the PCE and CPI cases across all time periods indicate that the summed lagged coefficients are more explanatory than the corresponding coefficients in the simple form case. However, it is difficult to compare the simple form coefficients with the summed lagged coefficients in the PPI case, as none of summed lagged coefficients are significant. As such, the PPI regressions in both cases offer little evidence regarding the importance of lags. The PCE and CPI regressions do however suggest that it takes a few periods for labor productivity increases to be adopted in the form of real wage growth. The charts displaying national income trends (in appendix B) are also compelling. They indicate that labor s share of national income has stayed relatively static, although corporate profits, while varying through the period, have hit a high in recent years. This could suggest that firms have increased profits by not rewarding increased productivity with increase payments to labor, and, therefore, corporate profits have climbed while labor share has stayed relatively constant.

31 Sachdev 31 Furthermore, the trends in the price indices discussed indicate that consumer prices (CPI and PCE) have risen faster relative to producer prices, so labor s terms of trade the difference between the rate of change in prices of things workers buy and the rate of change in the prices of things workers make have fallen. Workers should be paid on the basis of the goods they produce, not on the goods they consume. As such, if prices of goods that workers produce rise more slowly than the prices of the good they consume, which is what the price trends demonstrate, then it is unlikely that labor income will keep pace with rising productivity (Cashell 2004). The trends in consumer prices and producer prices also connect back to the argument about labor s static share of national income and increasing corporate profits. If these price trends indicate that producer prices the prices that firms pay for the products they need to make their final goods are rising slower than consumer prices, which are essentially the amount people are paying for goods, then one could argue that firms profit margins are increasing. Consequently, these trends further suggest that the increases in productivity are being realized in the form of increased corporate profits as opposed to going back into rewarding labor. 6. Conclusion This paper has focused on exploring what the popular press has labeled as a new phenomenon in the labor markets that of a widening disparity between the growth in productivity and the growth in real wages. The section on the relevant literature summarizes much of the work that has already been done on the topic and many of the theories for why this wage productivity gap exists and the extent to which it exists.