David R. Henderson  

Three Problems with Chetty's Study of Income Mobility

The Perverse Effects of the Pr... Do checks and balances reduce ...

Stanford University economist Raj Chetty claims that the American dream is fading. He may well be right, but his data on which he bases that judgment don't make his case. Alan Reynolds points this out well in a recent Cato blog post titled "Misconceptions in Raj Chetty's 'Fading American Dream.'"

What is the American dream? Here I pretty much agree with Chetty. The dream is that you will be wealthier than your father. The problem comes with their proxy for wealth and with their baseline.

According to Chetty and his co-authors of a December 2016 NBER study, "the fraction of children earning more than their parents fell from 92% in the 1940 birth cohort to 50% in the 1984 birth cohort." They measure income for each cohort at age 30. In other words, at age 30, only 50% of people born in 1984 had incomes higher than the income of their parents when they were age 30, whereas for the 1940 birth cohort, 92% had income, at age 30, that exceeded their parents' income at age 30.

Here's the essence of Reynolds's critique:

What it [the study] really shows is:

First: Incomes were extremely low in 1940, so it was quite easy to do better 30 years later.

Second: Doing better than your parents is not defined by your income at age 30, but by income and wealth accumulated over a lifetime (including retirement).

Third: A rising percentage of young people remain in grad school at age 30, so their current income is lower than that of their parents at that age but their future income is likely to be much higher.

Alan goes into each of those in detail.

I remember first reading about this a month or so ago and wondering about the 1940 baseline. Think about a typical earner in 1940. That was the tail end of the Great Depression. How hard would it be for someone aged 30 in 1970 to earn more than a parent at age 30 in 1940? Not very.

Comments and Sharing

CATEGORIES: Income Distribution

COMMENTS (11 to date)
Hazel Meade writes:

I'd also question whether the "American Dream" ought to be defined by whether you're making more income than your parents at all.

Modern society provides all sorts of avenues for self-actualization that aren't measured in terms of income, and which didn't exist 40 years ago. The question should be whether people today are more able to fulfill their dreams than their parents were, not whether they make more money.

I think the answer to that is likely yes. Social and economic constraints compelling people to enter less than fulfilling occupations and marriages have dramatically relaxed since the 1940s. Although the cost of living may have risen somewhat (mainly due to rising housing prices), technological advancement and urbanization have created more opportunities for specialized occupations. So more people can afford to ditch the rat race and pursue and occupation of their choice.

Nodnarb the Nasty writes:

Tyler Cowen pointed out a series of blog posts by Vincent Geloso on just this topic back in December. The first one is here. I think there are two more...

Nodnarb the Nasty writes:

Ooops, the first one is here. There are 4 posts altogether...

David R. Henderson writes:

Excellent! Thanks.

john hare writes:

I had a conversation with my son on this topic about a month ago. His income has matched or beaten mine every year since since he was thirty. If comparison is to me at 30 though, I've been clobbered. Quality of life would depend on definitions though he probably has the edge there now as well, though if comparing to me at 30 he's been far ahead. Generationaly he is on track to do far better than I have. The grandchildren seem to be on track to do better yet. Going the other direction, I have done better than my parents, and they did better than most of their parents.

I am willing to suggest that five generations in my family have each seen improvement by any realistic measure. Quality of life is definitely better if it is restricted to age 30 each time.

It is my opinion that most people see the past through rose colored glasses and choose the data to match their priors. I.e., they don't remember the magic washing machine or its' meaning.

JFA writes:

Point 1 is well taken.

Point 2: doing better at 30 could be taken as (and likely is) a signal about whether you'll do better over your entire lifetime, so on this point Reynold's critique is weak.

Point 3: also a weak objection to the study. One can offer the argument that wages aren't high enough today to induce people to not go to grad school, which shows some weakness in the economy. And if education is mostly signaling, this part of Reynold's critique should worry us, not comfort us.

AlanG writes:

Income is only one aspect of "doing better." As Hazel Meade notes there are other aspects of how to measure this. I have done better than my father income wise but not housing wise. He was able to design and build his house in San Diego. I live in a 'previously owned' home built in the mid-1950s when Bethesda MD was first expanding. By that measure I am not better off.

My two daughters both went to grad school and decided to pursue careers working with special needs children. They will never realize the same income I did (unless they all of the sudden decide to become investment bankers) but they may realize more job satisfaction.

It's a complicated issue.

Jay writes:

As I read this blog I'm watching a movie on a 75-in TV that wasn't available to the richest American until a couple years ago, as long as I can afford one of those I'm already doing better than even the most well off parents.

Ann Bookout writes:

I reached out to Raj Chetty about his response to the article. His response is as follows:

His second two points essentially claim that measuring incomes at age 30 is too early, but
we also conduct our analysis at age 40 and find similar results (see figures 3C and S12). He also mentions that changes in tax-favored
retirement plans have increased incomes later in life. But our Figure 3B, which includes taxes and transfers, shows that results are very similar when we account for these factors. We also show that increasing incomes in recent cohorts to capture fringe/retirement benefits does not appreciably affect the results (Figure S4).

In his first point, Mr. Reynolds claims that incomes in 1940 were extremely low, so comparing mobility rates from that period to today is invalid. Using incomes at age 40 (when parents' incomes are measured in 1950) again addresses this issue. But more generally, one
of the key points of the study is that there was tremendous broad-based growth in incomes from 1940 to 1970, but there has been
very little growth in incomes (except for at the very top of the distribution) from 1980 to today. The consequence of this is that it
was very easy for children born in 1940 to do better than their parents, but it's has been much more difficult for recent cohorts.
Where we differ from Mr. Reynolds is that he appears to believe this was inevitable, and therefore makes our comparison invalid, while we think the stagnating incomes over the past 30 years across much of the distribution are a source of concern.

David R. Henderson writes:

@Ann Bookout,
Thanks for taking initiative on this.
Re the age 40, that is a relevant counter to Alan Reynolds’s point.
Figure 3B is not an effective counter. Here is what Alan wrote:
The Graph from Advisor Perspectives shows cumulative changes in real median income by age groups from 1967 to 2015. Median income rose much more at ages 45–64 than it did at ages 25–34, and the growth of median income has been fastest by far for those over age 65 (thanks in large part to rapid growth of tax-favored savings plans for retirement).
Figure 3B is simply not relevant to this. By definition, it doesn’t capture this higher income later in life.

gmm writes:

This seems like good demonstration of the slow down in wage growth, and I find the arguments about the baseline being very low to be circular.
I imagine if they measured total consumption, including healthcare, the results would be significantly different.


Return to top