A couple of posts ago, I showed that when analyzed properly, hourly pay has risen just as much as productivity since 1947. The keys to getting the analysis right are to
• Compare mean hourly compensation (not median compensation, and not wages net of fringe benefits or household income) to productivity,
• Compare the same workers and sectors of the economy when computing compensation and productivity,
• Look at the nonfarm business sector to exclude the housing sector (where imputed rent to homeowners is counted as income) and the government sector (where indirect taxes are counted as income) so that income sources that do not reflect the value of what workers produce are excluded from productivity,
• Use net GDP to compute productivity rather than GDP, so that income taking the form of depreciation–which does not go to workers or owners but will simply affect future productivity–is excluded from productivity, and
• Use the same inflation adjustment for both compensation and GDP.
I should have added to that list that proprietors’ income (income from one’s business) should also be excluded, as it is not at all clear how to allocate that category into income from labor and income from capital. I have updated the earlier post (and chart) to take this into account. When these guidelines are followed, the results indicate that hourly compensation is almost exactly where it should be if we expect it to rise with productivity:
It is not obvious, but when the same inflation adjustment is used for compensation and GDP, the question of whether hourly compensation has tracked productivity is just another way of asking whether labor’s share of income has changed. Income to labor takes the form of compensation, and economy- or sector-wide income is just GDP (or “gross value added” for individual sectors). Since real hourly compensation and productivity are just total compensation and GDP divided by (the same) hours and adjusted by (the same) price deflator, the two will track each other only when the change in total compensation is the same as the change in GDP. If those two changes are the same, then the ratio of compensation to GDP (labor’s share of income) will be constant. If hourly compensation fails to keep up with productivity (meaning that total compensation fails to keep up with GDP), then labor’s share of income falls.
Claims that labor’s share of income has fallen are legion–in economic journalism, the think tank world, and academia. Here’s a Mark Thoma summary of research from the Federal Reserve Bank of Cleveland. Here’s Noah Smith arguing that the drop in labor’s share of income means that robots (“capital” generally) represent a threat to workers. And here is Paul Krugman, displaying a chart from the Bureau of Labor Statistics that shows labor’s share falling from 67 percent in 1947 to 58 percent in 2012:
(Oops–and here I am arguing that the decline in labor’s share is related to the end of an era when men received a “breadwinner premium” so that married women could be kept out of the workforce.)
But the “decline” in labor’s share of income reflects the growing importance of the housing and government sectors (though these are excluded in the BLS chart, which focuses on the nonfarm business sector), of depreciation, and of proprietors’ income and how it is treated. Imputed rent to homeowners (the value of the services a home provides that would otherwise have to be purchased by renting), indirect taxes such as sales taxes, and money used by firms to replace or repair existing capital (which will make future workers more productive, producing income to be divided between workers and owners)–these forms of income do not go to owners or to workers. The chart below shows what happens when you look at labor’s share of income for the nonfarm business sector, after excluding proprietors’ income and depreciation from income:*
Looking at either the productivity-versus-compensation chart or this labor share chart, it is clear that if compensation is “too low” that has only been the case since the Great Recession started. Since labor’s share falls (and productivity outpaces hourly compensation) during recessions, that fact does not indicate that we are at the start of an irrevocable slide in how workers do relative to owners. The apparent decline in labor’s share is a too-literal interpretation stemming from a failure to think harder about what it is we are interested in measuring. To be clear, “labor” includes extremely well-paid executives as well as minimum-wage workers, so the fact that labor’s piece of the pie hasn’t shrunk does not mean that inequality between workers hasn’t grown. But it does complicate unified theories of rising inequality.
* I compute labor’s share of income as nominal compensation for the nonfarm business sector from Table 6.2 of the National Income and Product Accounts from 1947 through 2000, divided by (nominal net value added for the nonfarm business sector from NIPA Table 1.9.5 minus nominal proprietor’s income from NIPA Table 6.12). From 2001 to 2012, I apply the year-to-year change in a nominal compensation index for the nonfarm business sector from the Bureau of Labor Statistics Major Sector Productivity and Costs program to the previous year’s nominal compensation. That is, for 2001, I multiply the 2000 nominal compensation by the ratio of BLS’s 2001 index to its 2000 index and work forward through 2012.