In the last post, we explored some different common hypotheses behind the observed phenomenon of a growing gap between worker productivity and worker pay.
Today we’ll look at how tough it is to explain this gap based on how hard it is to measure different parts of the graph–and, luckily, find a few places that seem a little more pinned down.
As you saw in the last post, the three graphs we showed didn’t even look quite the same. The first graph showed a more gradual change that started earlier; the second showed an abrupt change, and the third showed many variants with less change. We’ll touch briefly on some of the measuring issues without diving too deep into the various methodologies.
In short, there are two things to measure: productivity and compensation. Neither of these is actually terribly straightforward, unfortunately. Let’s look at one at a time.
This one’s relatively quick.
Productivity is a measurement of how much stuff people make, divided by the number of total hours worked.
All sources on this show a substantial increase in productivity since 1973, somewhere between 70% and 100%. Heritage shows 100%, George Mason shows 81%, and EPI shows 72%.
From diving into their methodologies, here are the differences:
Since the differences aren’t too big we won’t dive too deep here, but we’ll touch on why the third is important: as time has gone on, assets have depreciated faster. More employees are using computers to do their work, and these depreciate after a few years. When more employees used wrenches and heavy machinery, those would depreciate over decades. So depending on how you factor in depreciation, you’re going to see a big difference in how that impacts changes in productivity as we get closer to the modern age.
This gets trickier. By our account, there are 5 major methdological decisions that differ across different ways of measuring the productivity gap, all of which are important. We’ll cover here and then summarize at the bottom of this section.
The first big question is whether you look at just wages, or compensation. At first blush wages might make sense, but total compensation makes more sense because it relates more closely to the total cost of employing someone, and the total benefit they receive from being employed.
If we look at just wages, it initially appears as if wages have decreased (inflation-adjusted) since 1973. If we add benefits, the picture is not so grim.
Non-wage benefits have indeed gone up as a proportion of total compensation. In 1973, non-wage benefits were 13% of compensation, and by 2012 they became 20% (this according to calculations done by Haver Analytics). This is due in part to rising health insurance costs, but also to increases in maternity leave and paid vacation, retirement plans, stock options, etc.
Right away, we see that we’re looking at different amounts of compensation between different sources. Heritage says that, adjusting for CPI, compensation has gone up 30%. EPI says it’s gone up 12.5%. As far as I can tell, this gap is driven primarily by whether you use median or mean measurements. In fact, when EPI lays median and mean compensation on top of each other, the picture shifts substantially:
Note: Here EPI’s median compensation increase is 8.7%, rather than the previously-cited 12.5%. This is due–at least in part–to the two studies ending their studies at different times.
Here we see that mean compensation actually rose much more quickly than median compensation, using CPI as the inflation adjuster. In fact EPI’s number of 42.5% is higher than Heritage’s of 30%, and I can’t figure out why. In this graph, EPI is drawing from multiple BLS sources (here’s the appendix if you want a gander), where Heritage is just drawing from the BLS’s Current Establishment Survey. But they’re much closer this way.
George Mason University’s analysis makes two other changes that neither Heritage nor the EPI use. First, they include the self-employed in their measure of compensation. This seems to make sense, as self-employed people have to be productive (they can’t manage others since they’re working alone). Self-employed compensation has gone up much faster than those who aren’t self-employed, and their work contributes to the total productivity of the country.
George Mason also uses NIPA (from the US Government’s Bureau of Economic Analysis) rather than the BLS’s Current Establishment Survey (Heritage) or a mix (EPI). NIPA shows higher compensation growth than BLS, and the Bureau of Economic Analysis makes a case why it should be used here.
The final methodological decision is which form of inflation you want to use when adjusting wages over time. In the graph above, you see the difference that comes about when you use IPD (the Implicit Price Deflator) vs using CPI (the Consumer Price Index) when measuring inflation.
In Heritage’s chart below, they include PCE (Personal Consumption Expenditures index) to show how these different inflation measuring mechanisms make a difference. Below we’ll look at how these different inflation measurement systems work.
In short, you’ll see different ways of measuring compensation based on how you deal with:
One of the toughest nuts to crack when showing productivity and compensation together, is deciding whose productivity and pay to track.
The most common graph (that we’ve seen) of this gap comes from the Economic Policy Institute. It compares total productivity to the compensation of “production/nonsupervisory workers in the private sector.”
A big question to ask is this: does this group of production/nonsupervisory workers generally represent the right group of people?
That took some digging, but we found the source at the Department of Labor’s Bureau of Labor Statistics. They show that this group covers about 80% of total employment on private nonfarm payrolls, so it’s not looking at a limited group of employment. That suggests we’re looking at representative data.
The other big question is whether you need to use the same inflation measurement mechanism for both productivity and compensation. At first-blush the answer might be “obviously,” though I’m not entirely sure. Turns out you have to use IPD to adjust for inflation for productivity, because a fixed basket of goods is not relevant there. Therefore, is using IPD for compensation more fair? It probably depends on whether you’re trying to measure how fairly compensated someone is versus their productivity, or whether you’re trying to measure how well off they are over time (versus their productivity). IPD probably better measures apples and apples; CPI probably give a more clear view of what someone’s actually facing day to day.
Heritage and George Mason argue that IPD is the best way to measure compensation inflation relative to productivity. This is because what a company pays its workers isn’t tied to what the workers buy, it’s tied to the marginal value of the work. It’s more “apples to apples” because the company’s productivity and its compensation are inflated in the same way. Another way of looking at it: they argue that just because the price of goods that workers buy is going up, doesn’t mean that they’re getting paid relatively less (at least not on its own)–it rather means that their world has grown more expensive.
In fact, the EPI and Heritage/George Mason actually seem to be in agreement in their final analysis as to what’s driving the growing gap, and their explanations both include consumer goods prices as part of the explanation.
There seem to be two major factors in the compensation-productivity gap, and the EPI and Heritage/George Mason get to them from different angles (the former saying they are causes of the gap, the latter saying that they are new ways of measuring which demonstrate no gap). These are the prices of goods, and inequality among productive workers.
EPI has two major explanations: “terms-of-trade” and “inequality of compensation.” The second, “inequality of compensation,” shows the gap between median (low) and mean (higher) compensation. In short, it’s saying that workers aren’t getting paid all that much less, but that some workers have benefited a ton while others have lagged behind. This is possibly not too surprising as certain jobs–such as being a retail clerk, cleaner, or waitstaff–have not become more productive, where others–such as being an architect, computer programmer, or high-tech factory worker–have grown a whole lot more productive. Technology has affected these roles very differently in the past 40 years, and a growing compensation gap for production workers reflects that.
The other is “terms of trade.” EPI explains this is the different in inflation for production and the inflation of CPI. In their words: “The middle gap in our graph—the gap between the two average hourly compensation growth lines—solely reflects the divergence between consumer and producer price trends, thus illustrating the terms-of-trade gap.” Basically, stuff not built in the US that year (imports, real estate, etc) is getting more expensive faster than what’s being produced is getting more expensive.
EPI says that the remaining gap of 10% is “loss in labor’s share,” which a more Heritage-leaning economist might say is the “real” compensation-productivity gap. By EPI’s graph that did not start emerging until about 2005.
Here are their graphs:
Interestingly their methodologies differ quite a bit, but they both explain the initial productivity vs. compensation gap with a few points.
First, they use IPD rather than CPI to account for inflation. As above, they argue that it is a better way of inflating productivity and wages in an apples-to-apples way. They are therefore saying that wages are keeping up, but prices of things that US workers aren’t producing are getting much higher.
They account for the median-mean gap by simply using the mean. This is strictly-speaking more accurate (as we’re looking at mean productivity as well), but misses EPI’s nuance that compensation of production employees (note that this excludes CEOs, investors, and the other folks we tend to think of when we talk about inequality) is growing more unequal.
They also have other explanations for why compensation is actually higher than the EPI believes, but these have been covered above, and instead of discussing causal mechanisms of a gap, they argue that it doesn’t exist, so we won’t cover them again.
So what’s really going on? You decide. Good luck!
Reader Amos pointed out in a comment on our last post that there was a hypothesis behind the growing pay gap that should also get some attention: that demand for labor has simply grown slower than the supply of labor, leading to a depression in price.
Amos suggested we read Ages of Discord by Peter Turchin, which makes the case for this hypothesis with a lot of data. It’s some reading for us to do later, and we’ll be back as we learn more!
For your exploration: the New York Times shows that post-2012, wages have actually grown faster than (low) productivity + inflation.
"It was the best of times, it was the worst of times, it was the…
https://play.acast.com/s/reconsiderpodcast/ukraine-viii-bold-strategies-and-high-stakes Ukraine 8: Bold Strategies and High Stakes Updates Situation in SeverodonetskHUGE numbers of troops…
https://play.acast.com/s/reconsiderpodcast/russia-invades-ukraine-vii-possible-aftermath . May 22, 2022 First, war update: Mariupol fell after 2.5 months of truly…
What happens when you start really believing your own bullshit? What happens when you’ve created…
The Russians are hoping to encircle Ukraine's defensive lines in the Donbass. If they do,…
So much has changed over the past month -- Ukraine is on the offensive. Kyiv…
View Comments