Can Productivity Growth Be Used to Reduce Working Time and Improve the Standard of Living of the 99 Percent? The Future of Work in the 21st Century A Report of the Economic Analysis Research Network Future of Work Project Dean Baker, Center for Economic and Policy Research
There have been many accounts of a future economy where tens of millions of workers face little prospect of employment after losing their jobs to robots or other labor saving technology. This paper discusses this risk and provides further insight into the potential benefits and dangers of new technologies. The first part puts the prospect of labor replacing robots in the context of the technological change we have seen in past decades. Based on past experience, it is unlikely that we will see rates of productivity growth that are as high as what the economy experienced during its golden age from 1947-1973. Furthermore, productivity growth will in general be a positive factor in the future as it was in the past. If most of the gains from growth go to those at the top end of the income distribution it will have been the result of policy choices, not the inevitable course of productivity growth or the development of the market. The second part is a set of projections that show future living standards under alternative assumptions on productivity growth and income distribution. The projections highlight the fact that higher productivity growth, as in the robot replacing workers story, is likely to be associated with much higher living standards in the future. Productivity growth in the future --whether due to robots or other new production technologies-- will greatly facilitate the reduction in annual working time through a combination of shorter work weeks, paid vacation and holidays, paid sick days, paid parental leave, and a conversion of our national unemployment insurance system to one that focuses on compensating workers for reductions in their hours of work, not the loss of their jobs.
Technological Change Past and Present The displacement of workers by technology is not new, nor is the concern that it will lead to mass unemployment. There were numerous books and articles written in the fifties and sixties expressing concern that automation would lead to mass unemployment. In fact the House of Representatives Committee on Labor and Education had a subcommittee on “Unemployment and the Impact of Automation” which held widely publicized hearings in 1961 (House Committee on Labor and Education 1961). If the replacement of labor by technology is a cause for workers’ concern there was much to be worried about in this period. The annual rate of economy-wide productivity growth averaged 2.8 percent over the years from 1947 to 1973. This compares to an average of just 1.6 percent in the years since
1973. From the mid-1990s onward there was a pickup in productivity growth, although it has slowed sharply since the years leading into the downturn that began at the end of 2007. TABLE 1 Productivity Over the Last Six Decades (annual rate of increase) Net Productivity Productivity 1948-1973 2.80% 2.74% 1973-2013 1.60% 1.40% 1973-1995 1.29% 1.11% 1995-2013 1.93% 1.74% 1995-2005 2.48% 2.27% 2005-2013 1.24% 1.06% Source: Bureau of Labor Statistics and author's calculations.
“Usable Productivity,” (Net with CPI measure 2.96% 1.20% 0.86% 1.44% 1.99% 0.72%
Table 1 shows the average rates of economy-wide productivity growth for six periods. As can be seen, the average rate of productivity growth for the period 1947 to 1973 exceeded that in even the peak growth period since 1973 (from 1995 to 2005). The gap is somewhat larger if we use a more refined measures of productivity growth, “usable productivity,” shown in the last column of Table 1.1 As can be seen, over the long period from 1947 to 1973 usable productivity growth was twoand-a-half times as fast as the average growth rate in the period after 1973. The 1947-73 growth rate was still more than twice as rapid as the growth rate in the period since the 1995 speed-up. And it was more than four times as fast the pace of productivity growth we have seen since the 2005 when growth began to slow again. If productivity growth is inherently harmful to workers then we should have seen some serious negative effects in the 1947-73 percent. Instead, this was a boon time for the country’s workers. Productivity growth was largely passed on in higher wages and compensation. Average real hourly compensation rose at a 2.6 percent annual rate as most of the gains in productivity were passed on to workers.2 There also was little problem with unemployment. The unemployment rate averaged 4.5 percent in the 1950s and 4.8 percent in the 1960s, averaging just 3.5 percent in 1968 and 1969. The number of jobs grew at a 2.1 percent average rate over this period, the equivalent of roughly 3 million jobs a year in the current labor market. Obviously productivity growth by itself is not a problem for workers. In fact, quite the opposite seems to be the case. The strong productivity growth of the decades immediately 1
2
“Usable productivity” uses a net measure of output recognizing that the output going to depreciation by definition cannot be consumed or used to expand investment. It also adjusts for differences in price indices across different periods. This effectively allows us to compare across periods the extent to which gains in productivity can be translated into increases in living standards. The concept of “usable productivity” is explained in more detail in Baker (2007). The data in this section are taken from the Bureau of Labor Statistics website. The compensation data come from the productivity series showing compensation for the non-farm business sector.
2
following World War II was associated with rapid improvements in living standards that were widely shared by the bulk of the population. Figure 1 shows income gains for families at the 20th, 50th, and 95th percentiles. When productivity growth slowed in the period after 1973 income growth slowed for the country. An upward redistribution in come after 1973 meant that most families saw little benefit from the slower productivity growth that we did experience. FIGURE 1 Low, Middle, and High Incomes, 1947-2011, (20th, 50th, and 95th percentiles)
Source: U.S. Census Bureau, Current Population Survey, Annual Social and Economic Supplements.
The rapid productivity growth of the early post-war years did not translate into unemployment precisely because the gains from productivity growth were passed on in higher wages. This gave workers increased purchasing power, which increased consumer demand, and allowed the economy to remain near full employment even as each worker was able to produce far more in each hour of work. There are a wide range of projections about the course of productivity going forward, but few analysts anticipate growth anywhere near as rapid as what we saw in the golden age following World War II. The Congressional Budget Office, which tries to situate its projections in the middle of the range put forward by the experts in the area, projects that productivity growth will average just 1.8 percent in the decades ahead (Congressional Budget Office 2014). This would translate into roughly 1.3 percent per year in terms of usable
3
productivity.3 Some experts, like Northwestern University Professor Robert Gordon, are considerably more pessimistic, projecting growth of roughly 1.0 percent for the century ahead (Gordon 2012). The main outliers from this pessimism are Erik Brynjolfsson and Andrew McAfee, whose book The Second Machine Age, describes a world in which technological innovations sharply reduce the need for human labor in the near future. In this world we have computer driven cars, trucks, and busses, navigated by GPS systems. Robots handle most retail transactions, including restocking shelves and unloading new shipments. Innovations in computers allow workers with relatively little training to be better at diagnosing illness than the best physicians. This would be a world of immense abundance in which we could count on technology to meet the vast majority of our needs with little or no human labor. It is of course possible that Brynjolfsson and McAfee will be proved correct, but even in their optimistic story it is not clear that we would expect productivity growth in excess of what the economy saw in the early post-war decades. In other words, there is little reason to believe that productivity growth in the decades ahead will be more rapid than what we have seen in the recent past and certainly not qualitatively more rapid than what we saw in the boom period after World War II. Therefore productivity growth should not pose a problem for employment and living standards for the bulk of the population. If the pace of productivity growth will not be the problem, there is an implicit or even explicit concern raised by many that productivity growth will have an inherent bias in the sense that it will disproportionately take away the jobs of less-educated workers, while increasing opportunities for the highly educated and skilled segments of the workforce. Alternatively, technology could shift income from labor of all types to capital in the form of a rising profit share of income. In both these cases, future productivity growth could be associated with a worsening standard of living for large segments of the workforce. However, there is little hard data to support the case that such shifts in income distribution have been or will be driven by technology. These arguments are most often put forward as a set of anecdotes pointing out relatively low-skilled jobs that have been lost or will likely be lost in the foreseeable future due to technology. The professional research that has examined this issue has been at best inconclusive.4 The shifts in wage income don’t correspond well with developments in technology at all. The biggest shift from non-college to college educated workers occurred in the 1980s, before the productivity upturn of the mid-1990s and the advent of the Internet. While there has been some increase in the college premium in the later period, it has been relatively modest. Most of the growth in wage inequality 3 4
This assumes a gap between gross output and net output of 0.2 percentage points annually and an annual difference between the consumer price index and the GDP deflator of 0.4 percentage points, which is an explicit assumption in the projections. See Mishel, Shierholz, and Schmitt (2013) for a recounting of the debate on skill biased technical change.
4
cannot be explained by educational attainment. In fact, since 2000 workers with only a college degree have been among the losers in the U.S. economy, with little or no growth in real wages. Only workers with advanced degrees have seen wage gains (Mishel, Shierholz, and Bivens, 2012, Figure 4AJ). An alternative version of the technology story is “hollowing out of the middle” occupational argument advanced most prominently by David Autor. In this view, the failure of middle-class workers to share in the benefits of productivity growth is due to a loss of jobs in occupations that used to offer middle-class wages. This alternative view also does not fit the data well, as shown in Mishel et al (2013). There is no hollowing out story in the decade of the 00s, even though wage inequality continued to grow as fast as it had in prior decades. In the 00s, the share of employment in occupations at both the high end and the middle of the wage distribution declined, with only low-end occupations seeing an increase in their employment share. Mishel et al also shows there is no correlation between changes in relative shares of occupations and changes in wages by occupation, casting further doubt on the “job polarization” hypothesis. With neither the traditional skills-biased technical change theory nor the job polarization occupational shift theory fitting the data very well, there is little basis for assertions that technology can be blamed for the upward redistribution of wage income. As noted before, there has been a shift from labor income to capital income in the United States in the years since 2000. In other countries, this shift began in earlier decades (see Karabarbounis and Neiman, 2012). However this shift from labor to capital has been a much less important cause of the rise in inequality than the shift of wage income to high-end earners. Furthermore, it can also be readily explained by institutional factors rather than any inherent bias in technology.
An Institutional Account of Upward Redistribution While the data may not support the argument that technology is driving the upward redistribution of income over the last three decades, it is possible to present an argument that changes in the institutional structure can explain this trend. Many of the institutional supports for the wages of workers at the middle and bottom of the income ladder have been eliminated or weakened in this period. Most obviously the share of workers represented by unions has fallen sharply since the early 1980s. It declined from close to 20 percent of the private sector workforce at the start of the period to less than 7.0 percent by 2014. Deregulation of several major industries such as trucking, telecommunications, and airlines also put downward pressure on the wages of millions of workers who had previously held middle class jobs. The structure of trade agreements, which exposed manufacturing workers to competition with low-paid workers in the developing world also had a negative 5
impact on the wages of tens of millions of workers.5 And the failure of the minimum wage to even keep pace with inflation also put downward on the wages of less-educated workers. In the three decades following World War II the minimum wage rose in step with productivity growth, far exceeding the rate of inflation. In addition, the decision to have much higher levels of unemployment was also an important factor in the upward redistribution of income. There are two important points here. First, the level of unemployment is a policy decision. The unemployment rate is determined by macroeconomic policies such as the size of the budget deficit, the size of the trade deficit (which is in turn primarily dependent on value of the dollar measured in other currencies), and the Federal Reserve Board’s interest rate policies. There are tradeoffs involved in deciding to have lower unemployment. For example, the Fed would be risking more inflation if it allows the unemployment rate to get to relatively low levels. However, a tradeoff doesn’t change the fact that there is a choice. In the three decades after World War II, the Fed was willing to allow the unemployment rate to fall to much lower levels than in the years since 1980. Using the Congressional Budget Office’s estimate of the non-accelerating inflation rate of unemployment (NAIRU) as a benchmark, the unemployment rate was on average 0.5 percentage points below the NAIRU in the years from 1949 to 1979. In the period from 1980 to 2013 the unemployment rate was on average 1.0 percentage points above the NAIRU. This is a difference of 1.5 percentage points. Even pulling out the years since the beginning of the recession, the unemployment rate averaged 0.6 percentage points above the NAIRU from 1980 to 2007, a difference of 1.1 percentage points from the earlier period.6 And low unemployment matters a great deal for those at the middle and bottom end of the wage ladder. The unemployment rate for African Americans is on average almost twice the overall unemployment rate. This means that when the overall unemployment rises by 1 percentage points it typically rises for African Americans by close to 2 percentage points. The story is similar, if somewhat less extreme, with Hispanics. The most disadvantaged workers benefit most from lower overall rates of unemployment. A sustained one percentage point decline in the rate of unemployment is associated with a 12.4 percent increase in the hourly wage for men at the 20th percentile of the wage distribution and increase of 8.9 percent for women. For men at the median of the wage distribution it is associated with a 3.8 percent increase in the hourly wage and for women the increase is 4.2 percent. The impact for workers at the 90th percentile is statistically insignificant.
5 6
See Baker (2007), Schmitt (2009), Bivens (2011), and Mishel, Schmitt, and Shierholz (2014). On the impact of unions and de-regulation see Mishel et al, 2012. For the impact of trade on wages see Autor, Dorn, and Hanson, 2013. This discussion draws from Baker and Bernstein (2013).
6
In addition, low unemployment also allows workers at the bottom end of the wage ladder to work more hours, since many of these workers are involuntarily working part-time or at least put in fewer hours than they would like. The low unemployment of the late 1990s was associated with a 17 percent increase in hours worked by households in the bottom quintile of the income distribution and 5 percent for households in the second-lowest quintile of the income distribution. The opportunity that high levels of employment create for increased employment, increased work hours, and higher pay mean that there is a strong relationship between the unemployment rate and the income of those at the bottom end of the income distribution. The low unemployment of the late 1990s was associated with a 13.4 increase in the average family income for the bottom quintile of households in the four years from 1996 to 2000.7 This evidence should drive home the importance of pursuing high-employment policies. It is also important to note that there is no economic constraint in 2014 that prevents the United States economy from operating with a lower rate of unemployment and higher levels of output. The problem is purely political.
The Policies that Protect High End Earners and Corporate Profits As detailed above, there have been major efforts over the last three decades to undermine the institutional structure that supported the incomes for workers at the middle and bottom of the wage distribution. However there has been no comparable effort directed at the structures that support the income of most workers at the top of the income distribution, even when these structures lead to major inefficiencies in the market. Starting with an obvious example, the United States has 840,000 doctors.8 Their average annual earnings are over $240,000 per year.9 More than 20 percent of doctors are in the top one percent of highest earning households.10 Virtually all doctors would be in households in the top five percent of income earners. There are many institutional barriers that ensure that doctors’ pay stays far above that of ordinary workers. First and foremost doctors have 7 8
Mishel, Bivens, and Sheirholtz, 2012, data for Figure 2C. Kaiser Family Foundation, “Total Professionally Active Physicians,” http://kff.org/other/stateindicator/total-active-physicians/ 9 This number is taken from the Medscape Physician Compensation Report, 2014, slide 4 [http://www.medscape.com/features/slideshow/compensation/2014/public/overview#4]. The calculation averages earning reported by male and female physicians assuming a 60 percent/40 percent split. 10 This figure is obtained by summing the number of physicians shown as living in households in the top one percent of the Income distribution (192,275). Jeremy White, Robert Gebeloff, Ford Fessenden and Shan Carter, “The Top One Percent: What Jobs Do They Have?” New York Times, Jan.15, 2012, http://www.nytimes.com/packages/html/newsgraphics/2012/0115-one-percentoccupations/index.html?ref=business.
7
enormous control over the licensing procedures that allow people to be certified as doctors. They have enormous control over the practice of medicine, requiring that many procedures that could be performed by much lower-paid medical professionals are instead performed by doctors. They also set standards of care whereby many conditions that are routinely treated by general practitioners in other countries are instead treated by highly paid medical specialists. As a result, the average pay for doctors in the United States is more than twice as high as the average in other wealthy countries.11 While this would imply enormous potential for gains from trade (either bringing foreign doctors to the United States or encouraging people to travel to other countries for major medical procedures), trade in physicians’ services has been almost completely absent from the agenda in trade negotiations over the last three decades. This is the result of the political power of doctors, not the development of technology. In fact, if technology were the determining factor in the labor market, doctors could anticipate a big hit to their pay in the decades ahead. The development of the Internet makes it possible for a much lower-paid physician in India or elsewhere to read MRI results or other scans. In addition, the advancement of diagnostic technology will make it ever easier for people with much lower levels of expertise to recognize medical conditions and recommend appropriate treatments. If doctors are able to prevent a massive reduction in the demand for their services and a corresponding reduction in their pay it will be due to their political power, not the nature of technology. The same would apply to other highly paid professionals like lawyers, dentists, architects, and engineers. In all of these cases, technology is likely to develop in ways that should both allow workers with much less expertise to perform many of their tasks and diminish the need for their work. In the case of university faculty, the development of on-line teaching in various forms could replace some percentage of college professors in the next two decades. Again, if this turns out not to be the case, and most university courses are still taught in the traditional manner, it will have been because of the power of university professors in controlling the development of the technology, not the inherent nature of the technology itself. Many of the country’s highest earners are in the financial sector. This is a sector that also benefits enormously from its ability to use the government to shield it from market outcomes. While there are a long list of ways in which the financial sector benefits from government support the two biggest are the implicit guarantees provided to the largest banks through “too big to fail” insurance and the special tax treatment accorded the sector. 11 “How Much Do Doctors Make in Other Countries?” New York Times, July 15, 2009, http://economix.blogs.nytimes.com/2009/07/15/how-much-do-doctors-in-other-countries-make/
8
The former benefit is effectively a free insurance policy that the government provides large banks that allows them to borrow at below-market interest rates. Because lenders assume that the government will intervene to bail out large banks if they find themselves in trouble, creditors are willing to accept a lower interest rate than if they believed the government would actually allow them to fail. The I.M.F. recently estimated the size of this subsidy at $50 billion a year in the United States (I.M.F., 2014). (It estimated the subsidy at $300 billion a year in the euro zone countries.) The benefits from this subsidy will be divided between shareholders and the top executives in the major banks. The other special benefit received by the industry is its special tax treatment. The financial sector is exempted from most state sales taxes in the United States and value-added taxes in Europe. There is no obvious rationale for this exemption. Effectively, other sectors must pay higher taxes so that the financial sector can evade taxation. This is part of the motivation behind recent efforts to impose financial transactions taxes. Such taxes would both raise substantial revenue and reduce the amount of money going to high earners in the financial sector.12 In addition to the exemption from many forms of taxation that the industry enjoys, it also benefits from a wide variety of clauses in the tax code which it routinely exploits. For example, private equity companies typically borrow heavily against the assets of firms they acquire because the interest is tax deductible. Since the government has no reason to promote excessive leverage, this scheme could be undermined by limiting the amount of interest that could be deducted. There are also enormous tax savings from relocating – at least in name – corporate headquarters in a state or country with a lower tax rate. The financial sector is quite adept in making such arrangements. Finally, many of the top actors in the sector (hedge fund and private equity fund managers) benefit personally from the carried interest tax deduction. This deduction effectively allows these managers, which include some of the wealthiest people in the country, to pay taxes at the capital gains tax rate (currently 20 percent), rather than the 39.6 percent rate they would be forced to pay if their wages were taxed in the same way as ordinary workers. In other words, it is not technology or the natural development of the economy that has allowed people in the financial sector to get extremely wealthy even as most workers see few of the gains from economic growth. They have rigged the rules of the game so that they work in their favor. The financial sector has enormous political power so it will be difficult to alter the rules and institutional structures that allow it to profit at the expense of the rest of the economy. But it is important to recognize that the source of much of its profits and the extraordinary income of the high earners in the sector has little to do with the natural development of 12 The I.M.F. recommended as the best form of taxation a “financial activities tax,” effectively a value-added tax imposed on the financial sector. It suggested a tax equal to 0.2 percent of GDP (@ $35 billion annually in the United States) as a reasonable target (I.M.F, 2010).
9
technology. It is simply a case of a political powerful group rigging the rules for its own benefit. Finally, there has been a notable increase in profit shares in the U.S. economy since the turn of the century. The profit share of corporate income rose from an average of 13.3 percent over the period from 1996-2000 to 19.0 percent in 2013.13 In popular discussions this shift from wages to profits is often blamed on technology -- more as an argument by anecdote than evidence – and that the trend will continue long into the future. The basic argument is that the people who own the robots will make all the money in a world where robots are replacing workers. While the pace at which robots or any form of capital is displacing labor is questionable, as noted above, there is a further problem with the idea that the people who will benefit are the ones who own the robots. It is not likely that robots (or other forms of capital) will be expensive to build. Presumably robots of the future will largely be able to build the robots of the future. This means that there will be relatively little labor involved. If robots end up being expensive it will likely be due to the fact that we assigned strong and long patent protection to the procedures needed to build a robot. If that case, it is not technology that is causing a redistribution from labor to capital, it is the decision to have strong patent rules.14 This makes no economic sense. The strength and length of patents are matters determined by policy, not technology. If technology is advancing so rapidly that it is leading to serious disruptions in the economy and society, then the obvious response would be to reduce the incentives for innovation. That would mean shorter and weaker patents. The result would be that the price of robots and other technology would fall more quickly. To take the extreme case, we could all have robots that would clean our houses, cook our meals, wash our clothes, and even install the solar panels to generate the electricity that powers their work. Since robots are built by robots, they would be plentiful and cheap, requiring no more than a few hours of labor to pay for one. In this world, the profits for robot makers and designers would also be relatively modest. Since everyone could copy their design and programming after a short period of time, the robot industry would be like the television industry is today: low markups, modest profits, and cheap televisions. The basic point is simple but important, if the development of robots and other important innovations lead to an upward redistribution of income it would be the 13 National Income and Product Accounts, Table 1.14, Lines 9 plus 11, divided by line 3 minus line 7. 14 Sovaldi, a drug that provides an effective treatment for Hepatitis C, provides an excellent illustration of this point. The patent holder, Gilead Sciences, sells the drug in the United States for $84,000. A generic version is available in Egypt for less than $1,000 (SOURCE). This huge price difference is explained by the decision to grant Gilead Sciences an unchecked patent monopoly in the United States. It has nothing to do with technology. The pharmaceutical industry alone had sales of $386.0 billion in 2013 (National Income and Product Accounts, Table 2.4.5U, Line 120). This comes to more than 45.0 percent of after-tax corporate profits for the year.
10
result of rules that provided excessive protection for patents and/or other forms of intellectual property. It will not be the fault of the technology. To sum up, there is no basis for believing that the natural development of technology will lead to mass unemployment and declining income shares for the bulk of the working population. If we face a threat from a continuation and possible acceleration of the upward redistribution of income that we have seen in the last three decades it is due to institutional structures that have been put in place for this purpose. These structures can be altered to ensure that the gains from technology are broadly shared, if there is the political will.
The Working World in 2044 The first part of this paper argued that there is no reason that workers need to fear productivity growth as a cause of unemployment or a factor driving the upward redistribution of income. There is nothing intrinsic to the development of technology that would lead to these outcomes. The real question is rather how we shape markets to deal with new technologies. This section will outline alternative scenarios depending on the extent to which the trends in inequality of the past three decades continue into the future. It applies three basic scenarios, the Reagan Redux scenario in which those at the top end of the income distribution enjoy the bulk of the gains from productivity growth over the next three decades as they did over the last three decades. The Status Quo scenario in which the income distribution stays constant over the next three decades, and the Populist Revival Scenario in which the upward redistribution of the post-1980 era is reversed. (These scenarios are described in more detail in the appendix.) With each scenario, there is a low-, middle-, and high-productivity case in which the annual rates of usable productivity growth are 1.0 percent, 2.0 percent, and 4.0 percent, respectively. The last case is a rate of productivity growth that is a full percentage point more rapid that the rate of the golden age. This is presumably the high end of the plausible range for productivity growth over the next three decades. In each scenario, income gains for workers can be taken in the form of a mix of higher annual wages, more leisure in the form of paid parental and family leave, sick days, vacation days, or shorter workweeks. Historically workers have taken the benefits of productivity improvements in the form of higher wages and more leisure time, including shorter workweeks and more paid time off. In the years since 1980 there has been little gain to the workers in the United States in the form of more leisure, although workers in other wealthy countries have seen reductions in work time as shown in Figure 2.
11
FIGURE 2 Reductions in average annual hours 1980-2012 OECD
Source: OECD
Interestingly, there is also a negative correlation between average hours worked and employment rates. The countries that have seen sharper reductions in average annual hours have also seen the largest increases in the percentage of their population that is employed as shown in Figure 3. While this simple correlation hardly establishes causality, it is certainly consistent with the idea that by having each worker put in fewer hours, more workers have been given the opportunity to work. FIGURE 3 Average annual hours worked and EPOP 1990-2012
Source: OECD.
12
Policies that reduce average work hours, such as shorter work weeks or work days, work sharing, paid family leave, paid sick days, and mandated vacations should be an important part of a full-employment agenda. These policies both offer direct benefits to workers, but also by keeping labor markets tight they will help to ensure that workers have the bargaining power to secure their share of the gains from productivity growth.
The Slow Productivity Growth World In the slow productivity case it is assumed that usable productivity will increase at an average annual rate of 1.0 percent over the next three decades. This is roughly the pace of productivity growth during the slowdown period from 1973 to 1995. The compounded rate of productivity growth in this slow growth scenario is 34.8 percent in the 30 years through 2044. Tables 2a, 2b, and 2c, show wage indexes over the next three decades alternatively under Reagan Redux scenario, the Status Quo Scenario, and the Populist Revival Scenario. TABLE 2A Reagan Redux – Hourly Compensation 2014 2024 Bottom 90% 100 97.6 90-99 100 120.8 Top 1% 100 137.8
2034 95.2 141.6 175.6
2044 92.8 162.3 213.3
TABLE 2B Status Quo – Hourly Compensation 2014 Bottom 90% 100 90-99 100 Top 1% 100
2034 123.2 123.2 123.2
2044 134.8 134.8 134.8
2024 111.6 111.6 111.6
TABLE 2C Populist Revival – Hourly Compensation Bottom 90% 90-99 Top 1%
2014
2024
2034
2044
100 100 100
125.6 102.4 85.4
151.2 104.8 70.8
176.8 107.2 56.2
13
TABLE 2D Change in Annual Hours Consistent with Constant Income 2014 2024 2034 Reagan-Redux 0 49.4 101.3 Status Quo 0 -207.8 -376.5 Populist Revival 0 -407.6 -677.3
2044 155.9 -516.2 -868.8
The Reagan Redux scenario shown in 2a is the only one in which most workers can anticipate seeing a decline in their living standards over the next three decades. This point is important since it a scenario that both assumes a bad outcome in terms of productivity growth and a continuing upward redistribution of income. If we see either a middle case for productivity growth or a stabilization of the income distribution, then incomes for most workers will rise in the decades ahead. Since the Reagan Redux scenario implies that most workers will have a decline in hourly wages over the next three decades, there will not be a growth dividend to divide between leisure and income however in both the Status Quo and the Populist Revival scenarios even in the case of slow productivity growth there will be a substantial growth dividend. In the Status Quo scenario hourly compensation would be 34.8 percent higher in 2044 than in 2014. As an alternative to taking the benefits of productivity growth in the form of higher wages, workers could take the benefits in the form of more leisure. Table 2d shows the change in work hours, starting from a 2000 hour work year base, that would be consistent with constant annual income for the bottom 90 percent of the population. In the Reagan Redux scenario, since real wages are falling, workers would have to work more hours to keep their income constant. In 2024 they would have to work an additional 49 hours per year. By 2034 the reduction in hourly pay would require 101 additional hours to keep annual income constant and by 2044 it would require another 156 hours of work. In both the Status Quo and Populist Revival scenarios workers would be able to see substantial reductions in work hours and still keep their annual income constant. By 2024, they would be able to have a reduction of 208 hours per year in the Status Quo scenario and 408 hours in the Populist Revival scenario. The former would be sufficient to allow for 4 weeks a year of paid vacation, and 8 days of paid sick leave. Alternatively, if a typical worker wanted to take 6 months of paid family leave over a 35-year working lifetime that would translate into 3.5 days per year, still leaving the possibility of an additional 4 weeks of annual vacation time, plus 4.5 days per year for sick days. The possibilities for more time off from work are even more impressive in the Populist Revival scenario. It would be possible to have a 36-hour workweek, 4 additional weeks of paid vacation, 6 months of paid family leave over a working life-time and 5 days per year of paid sick leave. 14
Going out to 2044, the potential amount of additional leisure increases hugely. In the Status Quo scenario it would come to 516 hours per year, while in the Populist Revival scenario it would be 869 additional hours of leisure each year. The former would allow for a 35-hour work week, 5 additional weeks of vacation each year, 6 months of paid family leave, and 7 paid sick days. In the Populist Revival scenario it would be possible to reduce the work week to 30 hours, have 8 additional weeks of vacation, 12 months of paid family leave over a working lifetime, and 8 days of paid sick leave each year. Most likely workers would opt to take the benefits of productivity growth with a mix of higher income and more leisure. In the Status Quo scenario they would be able to work 10 percent fewer hours in 2044 than 2014 and still enjoy an annual income that was 21.3 percent higher (0.9 *1.348 =1.213). Ten percent fewer hours for a full-time full year worker would imply roughly 200 hours less work over the course of a year. This would translate into 5 weeks of additional vacation time. Alternatively, the reduction in hours could also be accomplished through a shorter workweek. A 10 percent reduction in hours would allow for a 36 hour workweek. All of these changes would be phased in over three decades, which would mean, for example, that workers could see another 8 days of vacation on average in 2024 and an additional 16 days by 2034. In the Populist Revival scenario hourly compensation would be 76.8 percent higher in 2044 than it is today. That would allow for 25 percent reduction in hours and still leave workers with a 32.6 percent increase in living standards (0.75* 1.768 = 1.326). Starting from a 2000 hour work year this would imply an additional 500 hours a year of leisure time. That would be the equivalent of a 12.5 additional weeks of vacation a year or cutting back a typical workweek to 30 hours. This scenario would mean that workers could both see large increases in living standards (far more than what they have seen over the last 30 years) and enjoy a huge increase in the amount of leisure time they have available.
The Middle Productivity Growth Case If the economy can sustain a 2.0 percent annual growth rate in usable productivity (roughly the rate of speed-up years from 1995-2005) then workers would see substantial gains over the next three decades even if there was a further upward redistribution of income. Table 3d shows the potential reduction in hours, starting from a 2000 hour work year, that would be consistent with a constant real level of income.
15
TABLE 3A Reagan Redux – Hourly Compensation 2014 Bottom 90% 100 90-99 100 Top 1% 100
2024 108.2 139.4 162.2
2034 116.5 178.8 224.5
2044 124.7 218.2 286.7
TABLE 3B Status Quo – Hourly Compensation 2014 Bottom 90% 100 90-99 100 Top 1% 100
2024 127.1 127.1 127.1
2034 154.1 154.1 154.1
2044 181.1 181.1 181.1
TABLE 3C Populist Revival – Hourly Compensation 2014 2024 Bottom 90% 100 145.9 90-99 100 114.7 Top 1% 100 91.9
2034 191.7 129.4 83.7
2044 237.6 144.1 75.6
TABLE 3D Change in Annual Hours Consistent with Constant Income 2014 2024 2034 -152.0 -282.5 Reagan-Redux 0 -425.8 -702.1 Status Quo 0 -628.9 -956.9 Populist Revival 0
2044 -395.8 -895.9 -1158.3
In the Reagan Redux scenario workers would still see an increase in hourly compensation of 24.7 percent by 2034. This would allow workers to reduce work hours by 5 percent and still see an increase in income of 18.5 percent (0.95*1.247= 1.185). A 5 percent reduction in hours would translate into an additional 100 hours a year in leisure time for a person working 2000 hours a year. That amounts to 2.5 weeks in additional vacation or a reduction in the length of the average workweek to 38 hours. Alternatively, using the same calculations as in the prior section, it would allow for paid family leave coupled with an additional 9 days of vacation a year. The Status Quo scenario would imply an increase in hourly compensation of 81.1 percent by 2044. This would allow for a 25 percent reduction in work hours coupled with a 35.8 percent increase in income. That implies a 500 hour reduction in annual hours for a full time fullyear worker.
16
The Populist Revival scenario would imply an increase in compensation per hour of 137.6 percent by 2044. In this scenario a 30 percent reduction in hours would still allow for an increase in annual income of 66.3 percent (0.7*2.376 = 1.663). A 30 percent reduction in hours would translate into 600 fewer hours of work for someone working a full-time fullyear job. That would allow for a 32 hour work week, coupled with 6 additional weeks of vacation a year. In an extreme case, where workers decided to take all of the benefits of productivity in the form of more leisure, by 2044 they would be able to reduce annual work hours by 396 in the Reagan Redux Scenario, 896 hours in the Status Quo scenario, and 1196 hours in the Populist Revival scenario. The reduction in hours in the Reagan Redux scenario would be sufficient to allow for a 35 hour work week and 4 weeks a year of paid vacation. In the Populist Revival scenario it would allow for a 20 hour work week, 6 weeks a year of paid vacation, 6 months of paid family leave over a working lifetime, and 7 days a year of paid sick leave.
The High Productivity Growth Case The high productivity growth scenario assumes an extraordinary 4.0 percent rate of annual productivity growth. This is a rate of productivity growth that is one percentage point higher than during the post-war Golden Age. It would only be possible if robotics and other new technologies paid enormous dividends in terms of productivity. It is worth noting that virtually no economists believe that anything like this pace of productivity growth is likely over a sustained period, but it is certainly imaginable that these technologies will pay off more than is generally expected. TABLE 4A Reagan Redux -- Hourly Compensation 2014 2024 141.1 Bottom 90% 100 196.9 90-99 100 237.8 Top 1% 100
2034 182.2 293.7 375.6
2044 223.2 390.6 513.4
TABLE 4B Status Quo -- Hourly Compensation 2014 Bottom 90% 100 90-99 100 Top 1% 100
2034 249.6 249.6 249.6
2044 324.3 324.3 324.3
2024 174.8 174.8 174.8
17
TABLE 4C Populist Revival -- Hourly Compensation 2014 Bottom 90% 100 90-99 100 Top 1% 100
2024 208.5 152.7 111.8
2034 317.0 205.4 123.5
TABLE 4D Change in Annual Hours Consistent with Constant Income 2014 2024 2034 Reagan-Redux 0 -582.3 -902.0 Status Quo 0 -855.7 -1198.6 Populist Revival 0 -1040.7 -1369.0
2044 425.4 258.1 135.3
2044 -1104.1 -1383.4 -1529.9
In this high productivity case workers would see large gains even with substantial upward redistribution. In the Reagan Redux scenario, average hourly compensation would be more than 123.2 percent higher in 2044. This would allow for a 56.2 percent increase in income even if annual work hours were reduced by 30 percent or 600 hours a year for a full-time full-year worker. In the Status Quo scenario compensation would have risen by 224.3 percent by 2044. In this case, a 30 percent reduction in work time would be associated with a 127.0 percent increase in income. Finally, if the rapid pace of productivity growth was coupled with a reversal of the upward redistribution of the last three decades average hourly compensation would rise by 325.4 percent. If work hours were reduced by 30 percent this would allow for an increase in annual pay of 197.8 percent. In the scenarios where all the gains are taken in the form of more leisure, by 2004 workers would be able to enjoy another 1100 hours a year of leisure in the Reagan Redux scenario, 1380 hours in the Status Quo scenario, and 1530 hours a year in the Populist Revival scenario. The reduction in hours in the Reagan Redux scenario would allow for a 20 hour work week and 5 weeks a year of additional vacation. In the Populist Revival scenario it would be possible to have a 20 hour workweek and still have 26 weeks a year of vacation.
The Payoff from Higher Productivity As noted in the sections above, even with modest rates of productivity growth workers should be able to enjoy higher standards of living in the decades ahead. The only scenario in which workers will not see gains in living standards is if slow rates of productivity growth are coupled with the continuation of the pattern of upward redistribution we’ve seen over 18
the last three decades. In every other scenario the projections show substantial gains in compensation which can be divided between more income and more leisure. Tables 5a and 5b summarize the various scenarios. TABLE 5A Change in Hourly Compensation from 2014 to 2044 Slow Middle Growth Growth Reagan Redux -0.073 0.247 Status Quo 0.348 0.811 Populist Revival 0.768 1.376 TABLE 5B Potential Reduction in Hours, 2014-2044 (income constant) Slow Middle Growth Growth Reagan Redux 0.079 -0.198 Status Quo -0.258 -0.448 Populist Revival -0.435 -0.579
High Growth 1.232 2.243 3.254
High Growth -0.188 -0.554 -0.693
Table 5a shows the gains in compensation per hour. If there were no change in hours per year this would correspond to the gains in income. (Some of the gains may go to non-wage compensation that would not show up in a paycheck, such as health care benefits.) Table 5b shows the potential reduction in hours that is consistent with constant annual compensation. As shown in 5a, if we assume a 2.0 percent rate of usable productivity growth, coupled with a constant distribution of income, then hourly compensation will be 81.1 percent higher in 2044 than in 2014. In the fast productivity growth case with a constant distribution of income, hourly compensation would be 124.3 percent higher by 2044. If these projected gains are translated entirely into a reduction in work hours the impact is striking. In the slow productivity growth case, with a constant distribution of income, workers in 2044 would be able to enjoy a reduction in work hours of 25.8 percent from current levels and still have the same level of income as in 2014. This would translate into a reduction of 516 hours annually for a full-time full year worker, or an additional 12.9 weeks of vacation. In the middle productivity scenario workers would be able to work 44.8 percent fewer hours in 2044 and still keep their income constant. That would imply a reduction in work hours of 896 hours annually or 22.4 weeks of additional vacation. In the fast productivity case, workers could reduce their work hours 57.9 percent by 2044 and still maintain the same level of income as in 2014. That would imply a reduction in annual hours of 1158 hours, the equivalent of 29 weeks a year of additional vacation.
19
In sum, if the economy experiences strong productivity growth, then workers should see substantial improvements in living standards and increases in leisure time even if there were still some upward redistribution of income over the next three decades. The only scenario in which they will fare poorly is if there is combination of slow productivity growth and a continued sharp upward redistribution of income.
Conclusion Historically productivity growth has been associated with rising living standards for the bulk of the working population. There is no technological reason that this will not be the case in the future. There is no obvious basis for thinking that future technologies will be more harmful to ordinary workers than the technological developments of the prior seventy years. If technology does end up being harmful to workers it will be due to the rules put in place by the government. One key to ensuring that most workers do share in gains will be keeping the economy operating at a high level of employment. While this will depend in large part on demand side policies, such as stimulatory fiscal and monetary policies and an exchange rate policy that limits the size of the trade deficit, levels of employment will also be affected by the supply of labor as well. If we implement policies that reduce average work time, so that work is shared more evenly, such as paid family leave, sick days, vacation, and shorter workweeks, it will help to keep labor markets tight, in addition to directly benefiting workers by allowing them a better balance between work and family/life obligations.
20
References Autor, David, David Dorn, and Gordon H. Hanson. 2013. "The China Syndrome: Local Labor Market Effects of Import Competition in the United States," American Economic Review, vol. 103, no. 6, pp. 2121-2168. Baker, Dean. 2007. The United States Since 1980, Cambridge: Cambridge University Press. Baker, Dean and Jared Bernstein. 2013. Getting Back to Full Employment: A Better Bargain for Working People. Washington, DC: Center for Economic and Policy Research. [http://www.cepr.net/index.php/publications/books/getting-back-to-fullemployment-a-better-bargain-for-working-people] Bivens, Josh. 2011. Failure By Design: The Story Behind America’s Broken Economy, Washington, DC: Economic Policy Institute. Gordon, Robert. 2012. “Is Economic Growth Over: Faltering Innovation Confronts the Six Headwinds,” National Bureau of Economic Research Working Paper #18315. House Committee on Education and Labor, U.S. House of Representatives. 1961. Manpower Utilization and Training: Hearings Before the Subcommittee on Unemployment and the Impact of Automation, H. R. 7373; a Bill Relating to the Occupational Training, Development, and Use of the Manpower Resources of the Nation, and for Other Purposes. Hearings Held in Washington, D.C. June 6, 13, and 14, 1961 International Monetary Fund. 2010. “A Fair and Substantial Contribution by the Financial Sector,” Washington, DC: International Monetary Fund. [http://www.imf.org/external/np/g20/pdf/062710b.pdf] International Monetary Fund. 2014. “Global Financial Stability Report,” Washington, DC: International Monetary Fund. [http://www.imf.org/External/Pubs/FT/GFSR/2014/01/index.htm] Karabarbounis, Loukas and Brent Neiman. 2013. "The Global Decline of Labor Share," National Bureau of Economic Research Working Paper No. 19136. [http://www.nber.org/papers/w19136.pdf] Mishel, Larry, Heidi Shierholz, and John Schmitt. 2013. “Don’t Blame the Robots: Assessing the Job Polarization Explanation of Growing Wage Inequality,” Washington, DC: Economics Policy Institute. [http://www.epi.org/publication/technology-inequalitydont-blame-the-robots/] Mishel, Larry, Heidi Shierholz, and Josh Bivens. 2012. The State of Working America, 20122013. Washington, DC: Economic Policy Institute. [http://www.stateofworkingamerica.org/subjects/overview/?reader]
21
Mishel, Lawrence, John Schmitt, and Heidi Shierholz. 2014. "Wage Inequality: A Story of Policy Choices," New Labor Forum (forthcoming). [http://s2.epi.org/files/charts/wage-inequality-a-story-of-policy-choices.pdf] Schmitt, John. 2009. "Inequality as Policy: The United States Since 1979," Center for Economic and Policy Research Briefing Paper. [http://www.cepr.net/documents/publications/inequality-policy-2009-10.pdf]
22
Appendix The Reagan Redux scenario assumes that we see the same sort of upward redistribution in the next three decades that we saw in the years from 1979 to 2012. According to Piketty and Saez’s data, the top 1 percent of taxpayers saw an increase in their share of before-tax income (excluding capital gains) of 10.27 percentage points from an average of 8.07 percent in the period from 1960-1979 to 19.34 percent in 2012 (Piketty and Saez, 2013, Table A.2). Households in the 90th to 99th percent had an increase in shares from 23.93 percent in the period from 1960-1979 to 28.82 percent, the bottom 90 percent saw their share decline from 68.0 percent to 51.84 percent. The calculations in the Reagan Redux scenario assume that shares change by the same amount from 2014 to 2044, with the adjustments taking place at an even pace over this thirty year period.15 In the case of households in the bottom 90 percent, this implies that the income share will fall by an additional 16.16 percentage points over this period. In the case of the slow productivity growth this implies a reduction in real income over this period, since the decline in shares 31.2 percent decline in income shares (16.16%/51.84 percent) more than offsets the 34.8 (1.01^30) percent increase in productivity. The 90th to 99th percentile households are projected to see their income share increase by 20.4 percent (4.89%/28.82%). Households in the top one percent are projected to see their shares increase by 53.1 percent (10.27%/19.34%). The middle scenario applies these share shifts to cumulative productivity growth of 81.1 percent (1.02^30). The optimistic scenario applies the shifts to growth of 224.3 percent (1.04^30). The status quo scenario assumes that the income shares remain constant at their 2012 levels over the next thirty years. The populist scenario assumes that the shares revert back to their 1960 to 1979 average over the next three decades. This implies a reversal of the Reagan Redux scenario with the income share of the bottom of 90 percent increasing by 31.2 percent, the income share of the 90th-99th percentile declining by 20.4 percent, and the income share of the top 1 percent falling back by 53.1 percent.
15 For simplicity the calculations assume that the change in income at each 10-year interval is one-third of the income change projected for the 2044 end point.
23
