Steve Rose has a new piece on wage growth being published by the Urban Institute which was previewed in a blog post in the Wall Street Journal. It shows a considerably better picture than most of us are used to seeing. Whereas my friends at the Economic Policy Institute (EPI) show the real median wage for men has fallen by 7.4 percent between 1979 and 2011, Rose finds that real annual compensation for men has risen by 13.4 percent. EPI’s data show that the median hourly wage for women has risen by 24.2 percent. Rose finds a gain of 73.0 percent.
There are three issues that explain the differences. The main difference for women is the increase in average annual hours worked. Women are far more likely to be full-time full year workers in 2013 than they were in 1979. This is largely due to a breakdown of the barriers that excluded women from most types of better paying jobs and social norms that tended to confine women to working in the home. While the increased opportunities for women is clearly a positive development, we would expect pay to rise accordingly. People expect to be paid more for working forty hours a week than for working 30 hours a week. Therefore if we find that people having higher earnings because they are working more hours rather than getting higher hourly pay, it doesn’t really change the wage stagnation story.
The second issue is that Rose is looking at total compensation rather than just hourly pay. This includes payments that employers make for Social Security taxes, health care insurance and defined contribution pensions. This matters more for the 1980s, when there were substantial increases in Social Security and Medicare taxes than in the last two decades.
Looking at compensation in principle is reasonable if we want to know what workers are paid for their work, but it does raise some issues. The most important is that Rose’s measure only counts payments to defined contribution pensions, not payments to defined benefit pensions. This means that a switch from defined benefit pensions to defined contribution pensions would show up as an increase in compensation in Rose’s measure, even if no more money is being paid by the employer. This switch only explains a small part of the difference in wage growth, but it is certainly peculiar.
But the most serious area of dispute between Rose’s analysis and the EPI analysis is over the correct deflator to use in converting nominal wage or compensation growth to real growth. Rose using the personal consumption expenditure (PCE) deflator in the GDP accounts, while EPI uses the CPI-U-RS, a series that applies the current methodology for the CPI back through time. The CPI-U-RS shows an inflation rate that averages 0.4 percentage points more over the years 1979 to 2013. This means that if we use the PCE rather the CPI-U-RS it will increase the annual rate of wage or earnings growth by 0.4 percentage points. This increases the growth in the pay for men by more than 13 percentage points over the period EPI examined and the pay for women by almost 18 percentage points.
The main reason for the difference in the measure of inflation between the CPI-U-RS and the PCE is that the PCE incorporates the impact of substitution. The weight of items in the index is adjusted to take account of people’s switching from items that are increasing more rapidly in price to items that are increasing less rapidly in price. By contrast, the CPI-U-RS uses a fixed weight index that does not allow for substitution. While it is arguably appropriate to include the impact of substitution in a price index, there are three points to be made about the conclusion that wages have grown more rapidly than we generally believe.
The first point is that if we adjust annual wage growth upward by 0.4 percentage points in the years since 1979 by using an index with substitution we would almost certainly have to adjust wage growth in the prior period upward by roughly the same amount. While the PCE deflator does not provide as good a measure of consumer inflation for the prior period (the shift to services tends to raise the overall rate of inflation in this period) there can be no doubt that a chain-type index that incorporated the impact of substitution would show a lower rate of inflation in that period also.
If we compare the goods component of the CPI with the goods component of the PCE, the annual gap is 0.4 percentage points over the years from 1956 (the first year for which CPI data is available) to 1979. If anything, the higher and more variable inflation in this earlier period would be expected to be associated with a larger gap due to substitution. (The Bureau of Labor Statistics was also much less careful in quantifying quality improvements in goods and services over this period, meaning that current methodologies are likely to show a lower rate of inflation that the official index.)
The point is that if we want to use an index like the PCE that incorporates substitution, it will raise the annual rate of real wage growth, but it will not change the sharp slowdown from the earlier period. If we expected a certain pace of wage growth based on the quarter century after World War II, we would still be seeing every bit as sharp a slowdown in the last 35 years.
The second point has to do with the intergenerational implications of more rapid wage growth. There is a whole industry in Washington, largely financed by private equity billionaire Peter Peterson, that argues that we are wronging our children by passing on the bills from Social Security and Medicare. An implication of the faster wage growth obtained by using the PCE is that our children will be much wealthier than official data suggest. It also means that our parents were much poorer when they were growing up relative to today’s income levels.
This doesn’t fit well with the inter-generational inequality story that is so popular in Washington. Using the 0.4 percentage point increase in wage growth obtained by using the PCE over the last thirty four years, we would project that real wages would be on average roughly 12 percentage points higher thirty years from now than our current projections show.
The intergenerational equity crowd thinks it would be a horrible injustice if we were to raise Social Security and Medicare taxes by 2.0-4.0 percentage points to cover the projected increase in the costs of these programs. But if using the correct measure of inflation would show a real wage that is 12 percentage points higher than what we had been looking at in our official projections, why should we be concerned about the risk of tax increases that are between one sixth and one third as large as the amount gained by correcting the data? In other words, if the PCE is the right measure to use to adjust wages and income then the intergenerational equity crowd deserves nothing but ridicule since they don’t even have the beginning of a case.
The final issue has to do with the meaning of a cost of living index. The underlying idea is that we are supposed to have enough money to buy a basket of quality adjusted goods and services over time that will leave us equally well off through time. An explicit assumption is that the social, physical, and natural infrastructure have not changed. This is seriously problematic.
It means, for example, that in 2015 we would be able to buy the same phone and television as we did in 1979 and we would be as well off. There is no provision for the cost of buying a cell phone and paying for the service, nor the cost of the Internet. We are supposed to believe that a person would be as well off without Internet access in 2015 as they were in 1979. That doesn’t seem very plausible.
To take an even more extreme case, the cost of living index implicitly assumes that a person without access to treatment for AIDS in 2015 is as well off as a person without access to treatment for AIDS in 1979. (In fact, the onset of AIDS would lead to a reduction in the cost of living in the CPI. When the price of drugs developed to treat AIDS falls when their patents end, this will appear as a price decline. The initial cost of the drug never entered the index.)
It is not clear that there is much coherence to the idea of a cost of living independent of the social context in which we are asking the question. We know that an index that includes substitution will show a lower rate of inflation than one that doesn’t, but it is not clear what ideal concept this is getting us closer to.
In short, we can make wage growth appear stronger if we use the PCE or another index that incorporates substitution. However this doesn’t at all change the story of a slowdown in wage growth over the last thirty five years even if it can move us into positive territory.
 Andrew Biggs, of the American Enterprise Institute, calculated that a dollar contributed to a defined benefit pension plan was worth more than three dollars contributed to a defined contribution plan in a paper comparing the compensation of public sector and private sector employers.