Earlier this week, a paper titled “Why is Europe More Equal than the United States?” by Thomas Blanchet, Lucas Chancel, and Amory Gethin began circulating among political writers. The paper has been out for nearly 1.5 years and I’ve seen it float around before, though never as extensively as this time.
The shock conclusion of the paper is supposed to be that “the US redistributes a greater share of national income to low-income groups than any European country” and that therefore “predistribution, not redistribution, explains why Europe is less unequal than the United States.”
The paper reaches this conclusion by making some very odd accounting decisions, including counting old-age, disability, and unemployment benefits as pretax income and assuming that all non-cash benefits, except health care, have zero impact on inequality. According to the latter assumption, a family with $1 million of posttax income receives 100 times more benefit from free pre-k than a family with $10,000 of posttax income.
It would be easy to dismiss the paper as silly. But here I want to explain why I think papers that try to measure these kinds of things, including Gornick and Milanovic (2015), frequently wind up making these sorts of analytical decisions.
The Entire Population
For starters, I used the 2019 Current Population Survey (CPS) to sort the entire US population into 10 groups of people. I then graphed the frequency of those 10 groups at every age between 0 and 80, which you can see below.
Each individual is assigned to only one group based on which group they first qualify for in the order below:
- Full-Time Worker. Worked all 52 weeks (not counting time off) during the year for an average of 35 hours or more.
- Other Worker. Worked during the year at some point.
- Child. Below the age of 18.
- Elderly. Aged 65 and over.
- Student. Gave school as the reason they did not work during the year.
- Disabled. Has one of the disabilities tracked by the Census or gave disability/illness as the reason they did not work during the year.
- Carer. Gave family caregiving as the reason they did not work during the year.
- Unemployed. Gave inability to find a job as the reason they did not work during the year.
- Retired. Gave retirement as the reason they did not work during the year.
- Other. Everyone else.
This graph tells a fairly intuitive story about life and the population of our society.
With the exception of a handful of mostly part-time workers between the ages of 15 and 17, people do not start working until they reach adulthood. After adulthood is reached, some continue to stay out of the workforce because they are college students.
Between ages 25 and 54, referred to as the prime working ages, over 80 percent of people are working at least some amount during the year, with over 60 percent working full-time with no gaps in employment during the year. Those who are not working are mostly caregiving (e.g. stay-at-home parents) or disabled. In the early prime working years, there are more carers than disabled people. In the later prime working years, the opposite is true. A small sliver of people in this age range are not working because they were unemployed, meaning they could not find a job for the entire calendar year.
As the prime working years come to an end, you get some early retirements. In the US system, people can withdraw money from private retirement accounts at age 59.5 and can begin receiving a public old-age pension at age 62. By the time people are elderly (defined here as age 65), most people have exited the workforce entirely and only 30 percent still have a full-time job. By 70, the retirement age at which people are eligible for their maximum public old-age pension, only 25 percent of people are still working some, with only around 10 percent working full time.
With these ten population categories established, we can now look at how they are sorted across the income distribution. The following graph shows how these categories are distributed in each personal market income percentile. “Personal market income” refers to how much money an individual personally receives from non-governmental sources. This is not how much an individual’s parent, spouse, or roommate receives. It is how much the individual personally receives.
I have never seen anyone graph this before and it’s a bit jarring at first glance. But once you think about it for a second, it is fairly intuitive. None of the nonworker categories receive any personal income from working but they do receive some personal income from assets that they own and pension entitlements. But children neither work nor own assets nor have pensions. And so the bottom third of the personal market income distribution is completely dominated by children, with elderly and disabled people making up most of the remainder.
If you use the above graph as your baseline market income distribution, then the policy battle between “predistribution” and “redistribution” — i.e. the battle between wage compression and the welfare state — is not really a battle at all. Wages are non-existent in the bottom third of the society. Redistribution, or the welfare state, is the only way to reach these people.
For as long as kids, old people, and disabled people clog up the bottom of the market income distribution, the competition for most redistributive country or most equalizing country will always be won by the country that has the most generous child, old-age, and disability benefits. Thus, on this metric, the “predistributionists” who focus on wages and employment are stuck advocating for a statistically impossible alternative.
For the predistributionists to get something going in the debate, they need to use a different metric that moves a bunch of nonworkers out of the bottom of the market income distribution and a bunch of workers into the bottom of the market income distribution. Relocating these populations in this way blunts the significance of the welfare state by moving welfare benefits up the ladder and increases the significance of wage compression by putting more wage-earners at the bottom of the ladder.
Using family income, rather than personal income, serves this purpose well. Under a family income metric, rather than looking at how much each person gets from the market, you look at how much all of the members of each family gets from the market. You divide that total amount by the size of the family, while making certain adjustments to account for economies of scale of larger families, and then assign the resulting dollar amount to each member of the family. The following graph shows what the family market income distribution looks like using the categories we have been using above.
Using family income, instead of personal income, does indeed move a lot of nonworkers up in the distribution and a lot of workers down in the distribution. This is especially true of children. Under the personal income metric, children totally dominate the bottom third of the income distribution. Under the family income metric, they no longer do. This is for the obvious reason that, under the family income metric, the kids get some of their parent’s market income assigned to them.
Although this metric does a good job of moving children out of the bottom of the market income distribution, it does not do a great job of moving elderly and disabled people out of the bottom. Elderly people often live alone or only with other elderly people and receive income mostly or entirely from the public old-age pension. The same is true of adult disabled people living on SSDI or SSI.
So this family market income measure still ends up frustrating predistributionists. Under this measure, the way a country wins the redistribution or equality competition, especially when that competition is focused on low-end inequality or poverty, is to simply have generous old-age and disability benefits, and to a lesser extent, child benefits.
Dealing with the Elderly and Disabled
Since nonworkers in the form of old and disabled people continue to clog up the bottom of the market income distribution even after switching to the family income metric, the predistributionists have to resort to more extreme adjustments to get the baseline “market” distribution to a place that is more amenable to predistributionist arguments.
Blanchet et al do this by counting old-age, disability, and unemployment benefits as market income (or “pretax income”). The obvious way that this adjustment helps the predistributionist case is by significantly reducing the size of the welfare state and thus the amount of measured redistribution. The sneaky way that it helps the predistributionist case is by moving old and disabled people out of the bottom of the market income distribution and moving more workers into it. When the bottom of the distribution has more workers in it, the significance of wages and employment to inequality and poverty increases.
In the following graph, I use the same family income measure from the prior section but modify it to count Social Security and Unemployment Insurance as market income.
In this measure, elderly and disabled people continue to be over-represented at the bottom, as do children, but way less so than before. Most noticeably, the big bulge of old people that used to sit between the 1st and 15th percentile have been pushed up the distribution by their Social Security checks. Workers now make up a lot more of the bottom of the distribution.
Counting Social Security and Unemployment Insurance as market income is not the only way to pull this trick off. The other approach you see is to simply exclude elderly people from your sample altogether, as Gornick and Milanovic (2015) do. The following graph shows what happens when you do this.
Once you exclude the elderly and the people who live with them from your population, you see that workers are quite present at the bottom of the distribution. Indeed they reach one-third of the population by the 5th percentile! This of course lends itself to the predistributionist conclusion that wage compression and employment matter a lot.
What I hope to have shown above is that these debates about “redistribution” and “predistribution” really revolve around how one chooses to construct the market income baseline, which itself revolves around where you decide to place nonworkers in the market income distribution. Notice that in the analysis above, I was able to illustrate my point entirely by moving groups of people around the distribution without ever actually looking at dollar amounts. That’s because population composition is what’s really driving these results.
If you use market income measures where lots of nonworkers — mostly children, elderly, and disabled people — wind up at the bottom of the distribution, then you are going to find the welfare state to be pretty key. If you use market income measures where far fewer nonworkers wind up at the bottom of the distribution — either because you are assigning them income based on their family, assigning their welfare incomes as market incomes, or excluding them from your population altogether — then you are going to find that the welfare state is not that key.
In an egalitarian society, you want to socialize or reduce capital’s income share, compress wage differences among workers, and provide welfare benefits primarily to nonworking people. Trying to suss out which one of these is “more important” is analytically fraught in addition to being practically useless.