Duflo and Banerjee's Deficient Thinking on Incentives, Part II
By David Henderson
Last week, I wrote Part I of my critique of a recent long article in the New York Times by new Nobel Prize in economics winners Esther Duflo and Abhijit Banerjee of MIT. The op/ed is titled “Economic Incentives Don’t Always Do What We Want Them To,” New York Times, October 26, 2019. This is Part II of the critique.
The authors write:
On the flip side, when jobs vanish and the local economy collapses, we cannot count on people’s desire to seek out a better life to smooth things out. The United States population is surprisingly immobile now. Seven percent of the population used to move to another county every year in the 1950s. Fewer than 4 percent did so in 2018. The decline started in 1990 and accelerated in the mid-2000s, precisely at the time when the industries in some regions were hit by competition from Chinese imports. When jobs disappeared in the counties that were producing toys, clothing or furniture, few people looked for jobs elsewhere. Nor did they demand help to move or to retrain — they stayed put and hoped things would improve. As a result, one million jobs were lost and wages and purchasing power fell in those communities, setting off a downward spiral of blight and hopelessness. Marriage rates and fertility fell, and more children were born into poverty.
That’s bad. And it is true that many of us economists thought that people would adjust more quickly. But there is a big difference between the 1950s and now. The difference is that the higher-paying jobs are disproportionately available in cities where housing prices, adjusted for inflation and relative to real income, are much higher than in the 1950s. And that is mainly due to government restrictions on building. Harvard economist Edward Glaeser and Wharton economist Joseph Gyourko have written on this extensively. Could that matter for mobility? I think so. And, if it does matter, that’s evidence for the importance of incentives, not, as Duflo and Banerjee assert, for their unimportance.
Duflo and Banerjee go on to write:
Despite this, the faith in incentives is widely shared.
There’s that word “faith” again.
We [the authors plus Stefanie Stantcheva] asked half of them [half of 10,000 Americans surveyed] what they thought someone should do if he or she were unemployed and a job was available 200 miles away. Sixty-two percent said the person should move. Fifty percent also said that they expected at least some people to stop working if taxes went up, and 60 percent thought that Medicaid beneficiaries are discouraged from working by the lack of a work requirement. Forty-nine percent answered yes when asked whether “many people” would stop working if there were a universal basic income of $13,000 a year with no strings attached.
But here is the twist: When we asked the other half of our sample the very same questions in reference to themselves, we got very different responses. Only 52 percent said they would move for a job, and this fell to 32 percent of those who were actually unemployed. Seventy-two percent of them declared that an increase in taxes would “not at all” lead them to stop working. Thirteen percent of respondents said they would probably work less if they received Medicaid without a work requirement; 12 percent said they would stop working if there were a universal basic income. In other words, “Everyone else responds to incentives, but I don’t.”
They are suggesting a contradiction. But look more carefully at the questions and answers, one by one.
62 percent said the unemployed person should move, and 52 percent said they would move. Is 10 percentage points a huge difference? It’s substantial but not earth-shattering. They try to make it huge by telling us that only 32 percent of those actually unemployed said they would move. But then shouldn’t they compare that with the percent of unemployed people who thought the unemployed person should move? Why don’t they give that number?
Next, they found that “Fifty percent also said that they expected at least some people to stop working if taxes went up.” But “Seventy-two percent of them declared that an increase in taxes would ‘not at all’ lead them to stop working.” They seem to see a contradiction. There isn’t one. It’s quite conceivable that 50 percent would expect some people to stop working. 28 percent said an increase in taxes would cause themselves to stop working. Isn’t 28 percent some? And, by the way, this is awfully “all or nothing” thinking rather than the “on the margin” thinking that economists usually recommend. I think their data show up the problem of surveys as a measure of what people would actually do. I would be shocked if an increase in taxes caused as many as 28 percent of people to stop working. But I would be equally shocked if an increase in taxes didn’t cause some people to work less.
Next, they write, “60 percent thought that Medicaid beneficiaries are discouraged from working by the lack of a work requirement.” And then, “Thirteen percent of respondents said they would probably work less if they received Medicaid without a work requirement.” That is a big gap and at least the two questions are symmetric. But those in the first group aren’t typically Medicaid recipients. They are answering what they think Medicaid recipients would do and and many of them probably view Medicaid recipients as being not very motivated. Ask them what they would do if they were Medicaid recipients and they will input their own circumstances, and will probably have trouble thinking that they themselves would cut work. So again there’s no necessary contradiction.
Look at the final survey question. First, “Forty-nine percent answered yes when asked whether ‘many people’ would stop working if there were a universal basic income of $13,000 a year with no strings attached.” Then, “12 percent said they would stop working if there were a universal basic income.” This is a contradiction only if one thinks that 12 percent is not many. I think it is many.
This is turning out longer than expected. Later this week: Part III.