This question is not as easy to answer as it may first appear. In working on various posts last week I came across an apparent contradiction in the official data on compensation: some series show it rising in real terms, while others show it barely able to keep up with inflation. This discrepancy was also noted by a few readers, who deserve credit for their sharp eyes.
So I thought I’d take a bit of time to sort out these conflicting data series for myself. Here’s what I found. (A warning and apology here: what follows is a relatively econ-geeky post about data details that many may find uninteresting… and I won’t be offended if you stop reading here.)
There are three major sources for time series data on earnings: “Hourly Compensation,” from the BLS’s Productivity and Costs (P&C) dataset; the Employment Cost Index (ECI), which provides compensation series broken into the two sub-categories of wage/salary earnings and benefits; and the “Average Hourly Earnings” provided in the monthly employment report as part of the Current Employment Statistics (CES). The following two charts show the behavior of these different series since 1990. All series express hourly compensation rates in real terms.
Note: all series are expressed in real (inflation-adjusted) terms using the PCE deflator.
What explains the sometimes substantial differences between these series? There are several factors that contribute to the discrepancies, but let me point out the most important ones. (For a more complete description of their differences see this paper by Joseph Meisenheimer in the May 2005 issue of the Monthly Labor Review.)
First of all, two of the series – the CES series and the “ECI: wages and salaries only” series – do not include benefits that workers receive. In the charts, those are the pink and green series. Comparing the two ECI series shows that in the past three years or so, a significant gap has opened up between workers’ take-home pay and the amount of compensation that employers are paying, including benefits. I would argue that this is directly attributable to the soaring cost of health insurance since about 2000. Even if workers’ pay has been rising in real terms, nearly all of the increases have been going to pay higher health insurance premiums.
Secondly, the different series include and exclude different types of income and different types of workers. The table below summarizes the different types of workers and income that each series excludes.
Finally, it should be noted that the ECI differs from the other series in that it comes from a survey that is intended to compare the wage rate in a particular job over time, not the wage rate of a person. (The sample is 35,000 specific jobs across the country.) In other words, the survey compares what each job in the sample pays at one point in time to what it used to pay earlier. Furthermore, in constructing the average wage rate across the economy, the ECI holds the number and types of jobs constant at the proportions in the base year (which I believe was just changed from the year 1990 to the year 2000). What this means is that the ECI will not accurately reflect how a change in the composition of jobs in the economy might affect average wages.
Each of the series thus has its own strengths and weaknesses, and there’s no right answer as to which series is best. They each tell us slightly different things, and the differences between them tell us still more. For example, the surge in the P&C measure during the period 1998-2001 probably reflects the large-scale adoption of payments through stock options. The divergence between the wage/salary series and the total compensation series reflects the growing burden of health insurance. And the recent rise of the P&C measure compared to the ECI measure may reflect higher rates of compensation growth in for-profits firms compared to non-profit firms, or large increases in the compensation of self-employed business owners, or a change in the composition of jobs in the economy that the ECI hasn’t caught up with.
A note about income inequality: to the degree that some of the excluded groups (in the table above) may have different levels of income than others, the differences between the series may also suggest something about changes in income inequality. A word of caution about that, however: if you want to find evidence of income inequality, I think there are much better measures (such as the Census Bureau’s income data) than these compensation measures. There is too much else going on in these series to be able to safely attribute anything on the charts above to changes in income inequality.
So what’s my answer to the title question of this post? Personally, if I had to choose just one series to use it would be the P&C series. In addition to being arguably the most complete series, it seems to have done the best job of matching my sense about how the economy has done over the past 20 years. When asked, I think that most people would agree that income growth was indeed much lower during 2002 and 2003 than it had been during the late 1990s; the P&C series bears that out, while the ECI series doesn’t. Meanwhile, the CES series excludes benefits, which I think are a major part of the story today.
But let me reiterate the point that I have made several times now: just because real compensation is rising, that doesn’t mean that people are better off, particularly if nearly all of the gains are just going to paying higher health insurance premiums. This data persuasively illustrates that nearly all of our real compensation gains today (and I do think we’re seeing them) are being eaten up by the monster that we call a health care system in the US. Until we address the profound inadequacies of our health care system, this trend will only get worse.