The story behind the “double world GDP” estimates
October 10, 2012 1 Comment
Post by Vipul Naik (see all posts by Vipul Naik)
[UPDATE: I have added a new blog post titled gains from migration: GDP versus surplus that covers an issue that I didn't adequately address in this blog post: that even though the contribution to GDP is largely made directly through the migrant, the "benefits" of this, in terms of the social surplus, are captured more broadly.]
Critics of immigration generally spend very little effort critiquing the strongest quantitative argument in favor of open borders: that they would result in a one-time increase in world GDP by 50-150%, or, in short, slogan form, double world GDP (the canonical literature review is this paper by Michael Clemens and the slogan “double world GDP” seems to have first been used in this blog post by Bryan Caplan). Many of these critics are citizenists, and frankly couldn’t care less about world GDP. But others have scoffed at these estimates without providing a clear-cut argument against them.
To my knowledge, the best critique of these estimates is expressed in a series of blog comments by Ghost of Christmas Past, a commenter on EconLog. The most complete versions of the critique are in this comment by Ghost of Christmas Past on Caplan’s “National Egoism” post and this comment by Ghost of Christmas Past on Caplan’s persuasion bleg post. I would like to provide a detailed point-by-point response to Ghost of Christmas Past’s critique. Prior to that, however, I need to establish more clearly just where the “double world GDP” estimates come from and what I think are the weaknesses of these estimates.
The first question: how are these world GDP estimates calculated?
One of the things that Ghost of Christmas Past glosses over in his/her critique is that the “double world GDP” estimate is not simply the estimate made by Michael Clemens. Rather, Clemens does a literature review and considers the many different estimates, showing that they are in a range of 50-150%. The back-of-the-envelope estimate by Clemens is more of a sanity check of the following sort: okay, assume that gains are half of what we observe today, how much can that give us? Then, Clemens goes over the various aspects that go into computing these estimates and the degree of uncertainty behind each of these factors. For those who don’t have the time to read Clemens’ paper, a quick list of the factors that would need to be considered, solely from the migration of labor, are: the effect on migrants, the effect on competing labor, employers, and consumers in the target country, and the effect on competing labor, employers, and consumers in the source country. Of course, when people move, it’s a migration of both labor and consumers, as well as, possibly, a migration of capital. This would further complicate the analysis.
The overall story behind the double world GDP estimate is that the first-order effect on migrants — aka the benefit to migrants — is huge, thanks to the place premium. In other words, migrants gain a lot by migrating. What they do with these gains (for instance, whether they send part of the money to their home countries as remittances, or save and invest in businesses in the country they migrated to) is not directly relevant to the GDP gain estimate.
What about the gains and losses by others, including competing labor, employers, and consumers, in the source and target countries? These gains and losses could both be important, but they are quantitatively likely to be substantially smaller than the gains to migrants. There could be considerable debate about the sign of these effects. For instance, the debate surrounding the suppression of wages of natives is about the extent to which competition in the labor market from immigrants drives down native wages. Pessimists believe that the driving-down effect from three decades of immigration to the United States has been about 4.8% for high school dropouts in the United States, and closer to zero for other categories of workers. Optimists see the effects as zero or slightly positive at current levels of immigration to the United States (more US-specific stuff at the link). Both optimists and pessimists believe that substantially more open borders would probably lead to a bigger decline in native wages, though less so in the long term than the short term. The overall magnitude and sign of the effects is up for empirical debate. Note: Even with a modest wage decline, natives may still be better off if price declines and the expansion in the range of consumption options makes up for this decline. They may also be better off as employers and land-owners. It’s only a small fraction of natives who are unambiguously likely to be worse off — those who depend primarily on wages (i.e., don’t own land), work in unskilled jobs that directly compete with unskilled immigrant labor, and don’t value the expanded range of consumption possible through immigration.
So how does this fit in with the “double world GDP” estimate arguments? The crux is that these gains and losses figure as rounding errors compared to the huge gains experienced by migrants. There is ample room within a 50-150% range to accommodate huge differences of opinion about the impact of immigration on native wages.
The story about the effects of emigration is equally complex. On the pessimistic side, there are concerns about brain drain. The labor supply-only story would suggest that emigration raises wages in the source country by reducing competition for labor. However, it also reduces the demand for labor if we remember that workers are consumers as well. Again, there could be plenty of empirical debate. The point is that in general the effects on natives in the source country still figure as a rounding error compared to the gains to migrants. Note that remittances could lead to larger gains to the source country, but we shouldn’t count remittances in the estimates because remittances are a second-order distributional effect rather than a primary wealth creation mechanism.
There is one slight exception to the case above, which is the case of zombie towns. Zombie towns are places that have very huge populations but very few job and economic opportunities. When large numbers of people emigrate from these towns, the migrants obviously gain, but it is also usually the case that the few people who do remain can gain by making better use of what’s left — perhaps even by cooperating with emigrants they know to set up inter-regional businesses.
The upshot of all this: the main gains from migration are to the migrants, and there is considerable consensus that these gains are significant. There are secondary gains and losses to others in the source and target countries of migration, and there is considerable debate on the overall signs of these. But even pessimistic effects of these don’t move the needle much on the gains from migration.
What the analyses leave out: non-economic factors
One could critique these analyses in two ways. A weak critique would be to concede the analysis, but worry about social/political/cultural issues. A stronger critique would be to argue that the conclusions of the analysis are flawed, either because the economics itself is wrong, or because the social/political/cultural effects would ultimately cause the economic analysis to fail. It’s the latter kind of strong critique that few restrictionists have offered so far, and Ghost of Christmas Past has helped buck that trend.
The first thing to keep in mind when talking about what the analyses leave out is that they leave out a lot of the potential positives from immigration. I haven’t yet gone through all the references used by Clemens, but at least Clemens’ summary does not separately consider the innovation case. If you believe that innovation is key to economic growth, then the next question to ask is whether migration would increase world per capita innovation, and whether that effect is being captured correctly in the current models. My best guess (and I admit, I would need to study this more) is that it is not captured correctly, so to this extent, the estimates probably understate the gains from open borders. (Some people may object that the innovation gains can be achieved through selective high-skilled worker migration. This is partly, but not wholly, true, because there are lots of different sorts of innovation and only some categories of these are exclusive to high-skilled workers. More at the innovation case page).
Another thing the analyses leave out is political externalities (more about political externalities at the link). My overall view on political externalities is that these could have a modest negative effect on the gain estimates, but not a huge one, thanks to status quo bias, founder effects, immigrant political apathy, and the many other points made at the link. If considering world GDP, there are also political influences that go in the other direction, specifically exit and competitive government. If, for instance, free migration leads to more competitive government, leading to increases in economic and political freedom, that in turn could lead to improvements in GDP growth rates. I tend to believe that the net effect on world GDP through the political channel would be somewhat positive. The net effect on the GDP of the target country of migration through political channels may be modestly negative, though probably not enough to overcome the economic gains.
Although these are the most salient channels in my view, it’s also true that almost any other harm of immigration can be postulated to ultimately affect the GDP estimate. There isn’t enough space to address all of these in this blog post, so I’ll just say that you should look at the harms to immigrant-receiving countries, harms to immigrant-sending countries, and global harms pages. I do hope to address the specific question of the effect of these claimed harms on the GDP estimates in future posts.
Flawed economic assumptions in the analyses
Another way to critique these analyses is to argue that the economic models used are inaccurate even from a purely economic perspective. One formulation (that Ghost of Christmas Past alludes to) is that after an initial small amount of migration that is used for low-skilled jobs, there will be a saturation in terms of low-skilled jobs.
Now, the way the macroeconomic estimates are done, they tend not to make very specific statements about the kinds of jobs that migrants will take up. These kinds of issues are abstracted away in the macroeconomic framework. But I frankly don’t see any shortage of relatively low-skilled and semi-skilled job niches in, for instance, the United States, that couldn’t be used to absorb 300 million or so immigrants. Until self-driving cars are a reality, there are ample job for personal drivers — a job that, while not entirely “unskilled” requires a relatively short training period and a fairly minimal formal education. There could be many more jobs for nannies, cooks, home attendants for elderly and disabled people, janitors, bus boys and girls at restaurants — none of which require extensive formal education (though each job has its own on-the-job skills and training). Formerly eliminated jobs, like gas station attendants to fuel your car at the gas station, or movie attendants to show you your seat in the cinema hall, could be revived. It’s not hard for me to imagine these kinds of jobs because many of them are still common in India — I encountered them all the time while growing up — though they are becoming somewhat rarer over time. And then there is cuisine: there are many “low-skilled immigrants” with ethnic cuisine skills which makes them employable in restaurants offering that cuisine, and there is ample scope for more such restaurants. Or supermarkets. Since both the demand and the supply base of labor will go up, it’s not very difficult to imagine people finding employment. And there may be many more ways of employing relatively unskilled labor in productive ways that entrepreneurs will discover once the opportunity presents itself.
A follow-up critique is that this reversal of automation and job elimination would lead to a technological slowdown, both for the target country of immigration, and for the world at large. I’ve collected the main counter-arguments to that concern here.
My conclusion: ample wiggle room
My overall conclusion is that the main first-order effect of benefits to migrants goes a long way toward making the case for a 50-150% gain in world GDP. All the other things mentioned, put together, can alter the estimate a bit in either direction. The innovation case probably means the estimate should be revised upward, and the political externalities arguments have more ambiguous effects. But a 50-150% approximate range leaves ample wiggle room for minor disagreements without affecting the core point.