Read on Twitter
Well I warned you I would tweet a lot this week about emigration and development!
This is the last thread.
☞ Specialists, this is for you.
I'll address some questions that came up, then step back to summarize what we know about how emigration evolves as economies develop —>
This is the last thread.
☞ Specialists, this is for you.
I'll address some questions that came up, then step back to summarize what we know about how emigration evolves as economies develop —>
I pointed out that a recent study on this same issue made a mistake in the way that it analyzed the statistics.
It inadvertently removed the effects of long-term rises in GDP per capita—the effect it sought to measure—from the analysis. I explained here: https://twitter.com/m_clem/status/1295766121044414464?s=20
It inadvertently removed the effects of long-term rises in GDP per capita—the effect it sought to measure—from the analysis. I explained here: https://twitter.com/m_clem/status/1295766121044414464?s=20
Today there was a public response, stating that the authors were "aware" of this all along, and dismissed the problem by saying that they already addressed it by averaging over five year periods. https://twitter.com/MEDAM_migration/status/1296358012156870656?s=20
What a disappointment.
First, the authors could not have been "aware" that their regressions purge any effects of sustained rises in GDP per capita.
That's the whole, entire effect they're trying to measure. If they were aware of this, they would never run such a regression.
First, the authors could not have been "aware" that their regressions purge any effects of sustained rises in GDP per capita.
That's the whole, entire effect they're trying to measure. If they were aware of this, they would never run such a regression.
Second, if one actually reads my discussion of this problem in the paper (page 42), it's clear that the problem has absolutely nothing to do with the length of the time periods. https://www.cgdev.org/publication/emigration-life-cycle-how-development-shapes-emigration-poor-countries
But math can be inaccessible. I'll illustrate it below with visuals.
But math can be inaccessible. I'll illustrate it below with visuals.
Here is the same real data on Mexico-US migration that I showed in my original blog post… https://www.cgdev.org/blog/emigration-rises-along-economic-development-aid-agencies-should-face-not-fear-it …but now I've averaged the data into 5 year periods.
This is a classic example of the #EmigrationLifeCycle. Emigration rose with development, when Mexico was poor.
This is a classic example of the #EmigrationLifeCycle. Emigration rose with development, when Mexico was poor.
We're going to exactly follow the regression method in that paper again, as I did before, but this time using averages over five-year periods. First regress emigration on population and keep the residuals, in red…
…then regress GDP on population, and keep those residuals (with everything in logarithms)…
And now regress the first set of residuals on the second set. That's arithmetically identical to what you do if you regress emigration on both GDP and population, estimating the coefficient on GDP.
That regression, with 5 year periods, looks like this:
That regression, with 5 year periods, looks like this:
The same strong, negative, and totally spurious relationship is still there!
Because the problem has nothing to do with the length of the periods. The regression, by construction, analyzes deviations from the trend, not the effect of the trend—not the effect of *development*.
Because the problem has nothing to do with the length of the periods. The regression, by construction, analyzes deviations from the trend, not the effect of the trend—not the effect of *development*.
Recall that I asked for some kind of explanation for why actual emigration from middle-income countries is over twenty times higher (!) than the analysis in that paper would predict. The response gave no answer. https://twitter.com/m_clem/status/1295766130749997056?s=20
I don't know how I can explain this any more clearly than I have. If I've been unclear, surely that's my fault. Now it's time for me to get back to real work.
But first I want to step back and summarize what we've learned.
But first I want to step back and summarize what we've learned.
We already knew that, in cross-sections of countries, relatively richer developing countries have far higher emigration (almost triple) than in the poorest countries.
The question at issue is whether countries, as they develop, tend to follow that path, or if the pattern above is an illusion created by taking a snapshot of one moment in time.
In econ language, is the relationship robust to country fixed effects?
In econ language, is the relationship robust to country fixed effects?
Here is a graph of the *raw data* for developing countries that grew over the past half century. No technical statistics. The arrow shows the path actually followed by each country.
It's on page 13 here https://www.cgdev.org/publication/emigration-life-cycle-how-development-shapes-emigration-poor-countries
It's on page 13 here https://www.cgdev.org/publication/emigration-life-cycle-how-development-shapes-emigration-poor-countries
They are obviously, unquestionably, typically following a path that resembles the cross-section pattern.
Is this just South-South migration? No, the middle figure drops that and includes only South-North.
Is this just high-emigration microstates? No, the 3rd figure drops them.
Is this just South-South migration? No, the middle figure drops that and includes only South-North.
Is this just high-emigration microstates? No, the 3rd figure drops them.
I do the technical work in my paper, forcing this result to march through a minefield of robustness checks. I include three different kinds of country effects (fixed in intercept, random in intercept, and random in slope and intercept). It is highly robust.
But you don't need all of that to understand what's going on. It's all in that picture above. The typical experience of developing countries — large and small, and for migration far and near — has been that as they get richer, more people choose to live outside them.
Policymakers interested in real-world effectiveness need to face this clear-as-sun fact about how development has happened so far.
In my blog post I discuss specific ways to do that. https://www.cgdev.org/blog/emigration-rises-along-economic-development-aid-agencies-should-face-not-fear-it
Thank you, all of you, for your engagement this week! And stay safe.
In my blog post I discuss specific ways to do that. https://www.cgdev.org/blog/emigration-rises-along-economic-development-aid-agencies-should-face-not-fear-it
Thank you, all of you, for your engagement this week! And stay safe.
