Simpson’s paradox - Wikipedia

/Simpson%27s_paradox

  • So why is it that in lower-income countries, richer people emigrate more? What does that mean about the effects of immigration?

    @MariapiaMendola and I went all-out to find answers, using survey data on 653,613 people in 99 countries.

    Start with the facts. This shows 120,000+ people in low-income countries (Malawi, Laos,…). The orange bell-curve is the distribution of income (0=average). The blue line (with confidence interval) is the probability that people at each income are actively preparing to emigrate:

    This is not actual migration, but people who report that their intent to migrate has recently culminated in a very costly action, like purchasing international travel or applying for a visa.

    We show in the data that relatively richer people, as you would expect, are better able to turn their migration wishes and plans into reality than poorer people. So the line for actual migration, by income, should be even steeper than the blue line above.

    That has two remarkable implications. First, when poorer people get more money, they often invest it in… migration.

    This has been found in specific settings around the world. @SamuelBazzi rigorously showed this happening in rural Indonesia:

    https://www.aeaweb.org/articles?id=10.1257/app.20150548

    The second implication is about what happens to immigration on the other end.

    It means that the “additional” migrants caused by rising income in the origin country are likely to possess more & more of the things that make workers earn more, like education or work ethic.

    In the cold, hard economic terms used in the literature, it means that rising incomes in developing countries mean 1) higher propensity to emigrate from the origin and 2) more “positive selection” of immigrant workers at the destination.

    https://www.cgdev.org/publication/migration-developing-countries-selection-income-elasticity-and-simpsons-parad

    What’s going on here? Is it that richer people have an easier time paying to migrate? Is it that richer people invest more in things that facilitate migration, like schooling? Is it demographic change accompanying rising incomes?

    The literature posits all of these and others.

    And why is it so difficult for people, especially many smart people in the policy world, to accept these facts?

    I want to address all these with two pictures from the paper.

    First, pool all the surveyed people in 99 countries into one graph. As people begin to earn the (price-adjusted) equivalent of thousands of dollars a year, they are more & more likely to be preparing to emigrate. For the richest people, that reverses.

    Now just split exactly the same data by education level. In green, that’s people with secondary education or more. In red, people with primary education or less.

    The dashed line (right-hand vertical axis) shows the fraction of people in the green (secondary education+) group.

    That inverse-U shape, the #EmigrationLifeCycle at the household level, is greatly diminished. The emigration propensity barely rises with income within each group.

    This is informative about the origins of the life-cycle. A lot of it comes from rising investment in education, which both motivates and facilitates emigration.

    It confirms what Dao, Docquier, @ParsonsEcon, and Peri find in cross-country data:
    https://linkinghub.elsevier.com/retrieve/pii/S030438781730113X

    It also explains a lot about why the Life Cycle is so counterintuitive. For any given kind of person—like a person with a certain level of education—emigration either doesn’t rise or actually falls with higher income.

    But the process of economic development means that people are shifting between those groups. In the last picture above, you can see people moving from the low-emigration group (low education) to the high-emigration group, as incomes rise.

    If this seems like a brain twister, you’re not alone. This counterintuitive “flip” in correlations is such a common pitfall of reasoning that it has a name: #Simpson's_Paradox (https://en.wikipedia.org/wiki/Simpson%27s_paradox)

    Notice that the “groups” in Simpson’s Paradox can be any size. They can even be individual people! That is, it could both be true that 1) any given person would be less likely to emigrate if they had more income AND 2) more people emigrate if the whole country gets richer.

    There’s no contradiction there. I have sat across the table in Brussels from a brilliant development expert who said, “But people tell me personally they’re moving to earn more money. Do you think they’re lying to me?”

    They are not lying. It’s just that the relationship conditional on individual traits can be the opposite of the relationship across all individuals, because economic development brings shifts in those traits. That’s Simpson’s Paradox.

    There’s a long-read blog that goes into more depth, and links to the papers, here:
    https://www.cgdev.org/blog/emigration-rises-along-economic-development-aid-agencies-should-face-not-fear

    https://twitter.com/m_clem/status/1295735909564981249

    –—

    voir aussi le fil de discussion: “Does Development Reduce Migration ?”
    https://seenthis.net/messages/526083

    ping @rhoumour @_kg_ @karine4 @isskein