Declining Migration May Be Good News

Crow Meat Can Taste Pretty Good, Let Me Tell You

I have argued many times that declining migration is a bad thing; that a society with lower mobility is a society that will suffer meaningful societal and economic consequences. However, I, like many migration commentators, have been somewhat at a loss to explain exactly why migration is falling. A new paper lays out an extremely compelling argument about why migration is falling, and suggests declining migration (whatever its long-term general effects) may be a good thing.

Regular readers know that I don’t think declining migration is a good thing. However, no paper I’ve read thus far makes a more compelling case against my view than the one I’m going to discuss, and it is so compelling that it’s given me serious pause about many of my deeply-held beliefs concerning migration. In hindsight, however, I feel a bit dumb that I didn’t realize I was going to have to devour this crow, as the essential argument the authors lay out is very intuitive. It’s also something I tried to do myself, but didn’t have the data chops to carry to fruition, so I’m ecstatic that somebody else did. Examples of my musings on this:

The core claim of the paper is simple: the geographic variation in occupation- and industry-specific wages has declined even as information about distant locations has improved, giving dramatically less reason to migrate. To do the study, the authors used microdata from the Current Population Survey on migration, occupation, and other demographic factors. In other words:

Because economic opportunity is more evenly distributed than in the past, and because workers have better information about those opportunities, migration is low.

Many readers will object to the idea that “economic opportunity is more evenly distributed than in the past,” but I think that part is trivially easy to prove. Which is why it bugs me that I didn’t think of writing this paper myself.

So let’s get into the paper!

What Causes Lower Migration Rates?

Not Changing Demographics!

I have argued that lower migration rates may be due to the higher prevalence of 2-earner households, and that theory is not fully refuted by this paper. However, virtually every other demographically-driven theory of declining migration is refuted, most easily by this graph:

It’s a bit small, but note that a counterfactual migration rate which adjusts for age, education, marital status, number of earners per household, income, occupation, and industry has a miniscule impact on gross migration. Now, having looked over the appendix on how this adjustment was done, I think these demographic factors may actually have a slightly larger impact than the authors suggest, but not enough to actually change their results. Suffice to say, I’m reasonably satisfied that they’ve shown that changing demographics can’t be blamed for falling migration, at least not through direct compositional effects, though dynamic effects are possible (i.e. having more old people in society may also cause young people to move less if many people migrate as groups lead by elderly people, etc).

So if it’s not demographics, then what? Well, another convincing chart suggests it may be something fairly fundamental to developed, English-speaking, immigrant-dense, geographically-large countries:

The top graph here should mostly be familiar to readers, as it shows the standard measures of migration. The exception is with the addition of Survey of Income and Program Participation (SIPP) data, which I don’t normally cite for various reasons. But the second graph is interesting, as it shows gross migration rates for Canada and Australia. Now, note that these scales are different: Australia’s gross migration rate is about 1/2 Canada’s, while Canada’s gross migration rate is about 1/2 the US’. In other words, gross migration in the US remains about 4 times as high as in Australia. However, some of this may be due to differences in regional definitions, and, across all three countries, we can see a very similar trend.

So, okay, we need an explanation that has some broad applicability: all sources show declines, and we see declines across similar countries. This has got to be some kind of structural explanation.

The authors have two preferred theories about why migration is declining, and each theory basically has two sub-theories:

Economic Convergence is Happening

  1. State economies have grown more similar over time in terms of occupational mix in each state.
  2. The wages available to workers in a given occupation in a given state have become more and more similar across states with time.

Information is Improving

  1. Workers have better information about wages, costs, and amenities than in the past: formerly, information about distant places was only gained by visiting or moving there. Now, long-range search is easier.
  2. People have “narrow priors” about other places; that is, they think they have a pretty specific idea about what life is like in other places.

Combine these all together and you get:

  1. You can work in almost any industry in almost any state, so less need to move to work “in your field.”
  2. Within a field, wages don’t vary as much as they used to, so you moving within your field doesn’t earn quite what it used to, so less incentive to move.
  3. You can Google distant places and look up tons of information about them, and can visit much more cheaply, so you can search for job/lifestyle without doing “experimental” migration, so less need to move to look for work or a better lifestyle, and far less migration for failed work searches or non-preferred lifestyle experiments.
  4. We have pretty specific ideas about what life is like in other places and already have good information, so it takes a lot to get us to radically alter our estimate of how much we value living somewhere else, so less incentive to move.

The result is less migration.

Let’s look at the proof for these claims.

Is Occupational Mix Becoming More Similar?


There’s a lot of math underlying these charts, but the core point here is that a higher Thiel Index score means greater variety or segregation of occupations. As you can see, up until 2010, occupations were getting less and less concentrated. Now, by this chart, you’d expect migration to fall sharply in the 1970s and 1980s, and more slowly in the 90s and 2000s, when in fact the opposite has occurred. But, hey, maybe there’s lag time. “Occupational mix” isn’t what would drive migration, but “occupational mix of hiring.” So maybe it took until the 1990s for the occupational mix of hiring to fall so much it impacted migration.

So this chart broadly supports the authors’ thesis, though there’s a timing mismatch.

Are Wage Rates Becoming More Similar?


This one is a boring table. But it summarizes some wage rates for major occupational categories. You’ll notice that the “Change” column for wages is negative for most occupations, and that the change in wage rates correlates with the degree of change in interstate migration rates. This one is a clear winner for the authors: it seems that, indeed, declining variance in wages is a real phenomenon, and is associated with declining migration at the occupational level.

The authors also test and see if workers tend to move towards places where there are higher wages for their occupation. Answer: yes!

The central estimate in these charts is generally above zero, suggesting that migrants do most likely, in general, move towards places where wages are higher. Crucially, these appear to be nominal wages. My guess is that if we estimated real wages using local-price-adjustment factors, we would find a much more positive value.

So then: workers do generally move to places where they make more money, wages have converged, and occupational mixes have converged. This is pretty much game-set-match for their theory that declining levels of regional economic diversity can be blamed for lower migration, except for some possible timing issues.

Is Experimental Migration Declining, i.e. Do Migrants Have Better Information?


The above charts show various ways of presenting various measures of iterative migration, that is, migrants who move somewhere, then rapidly move on somewhere else. Most migration researchers view this kind of iterated migration as indicating dissatisfaction with the first destination: failure to find a job, lifestyle not meeting expectations, poor social conditions, etc. Empirical research has generally borne this out, finding that repeat migration often arises from failure to find a job after migration.

So if workers have better advance information about other locations, then we can expect that there will be fewer “bad matches,” and thus fewer instances of iterative migration. And lo and behold, that’s exactly what we see: repeat migration has declined in total, and as a share of total migration. In other words, this kind of migration that proxies for information quality has fallen even faster than general migration, suggesting that information quality may have improved.

There are other possibilities! It may be that the cost of failure has risen due to increased indebtedness, higher homeownership rates, or higher healthcare costs. It may be that time out of work is more injurious than in the past. It may be that the same amount of experimental migrations occur, indicating information has not improved per se, but that modern state economies and policy supports are better at preventing the kinds of failures that trigger iterative migration. But improving information is one reasonable, plausible explanation.

These basic stylized parameters established, the authors create a very entertaining model or simulation of migration to demonstrate their key points. They include lots of elements I think are very important: individuals do not have full knowledge of their own preferences and skills, but obtain knowledge by experimentation. Awareness of lifestyle preferences may trigger job searches in localities with preferred lifestyle amenities, leading to migration that is, on its face, job-triggered, but, at its root, amenities-driven. In the case where all earning potential is identical in all places and an individual has extensive knowledge of their own preferences and skills, all migration will be amenities-driven, and small changes in amenities will drive large migrations. However, crucially, the model specifies “tight” priors, because this kind of tidal amenities-driven migration, in reality, does not happen. “Tight” priors can be simplified as meaning that if I dislike cold weather, learning that Michigan is only a desolate frozen tundra unsuitable for human habitation 180 days out of the year instead of 200 like I formerly believed won’t really nudge me at all to change my view of its habitability. I have a narrow range of preferences and prior beliefs.

This model worked up, the authors plug in some parameters similar to the U.S. economy in 1991–1997, and find it fairly closely predicts actual migration patterns. They then plug in changes to the period 2005–2011, and re-run the model, and find that their model correctly predicts the observed decline in migration.

Here’s the model vs. the data for 1991–1997, gross migration rates by age and education:

Looks fairly good to me!

And the blue lines above show 2005–11. Certainly, there are errors: the model seems to have underestimated migration for college-educated people ages 25–35 , while overestimated for people ages 30–55. But the same under/overshooting was present in 1991–97, and it looks like the change is about the same for the model and the data.

The authors find that their model explains about 40 percent of the decline in interstate migration since 1991.


I find this study very compelling. It definitely will cause me to revise downward my view of the degree of problem represented by declining migration. I have argued many times that declining migration was a pressing economic ill in need of a large-scale remedy: I may have overstated my case. I say “may,” however, because there are unanswered questions:

  1. We know why migration is declining. We don’t know what effect that will have going forward. If we migrate less, will the pace of innovation change? Will our sense of ourselves as one country change? What dynamic effects may exist from a society with less churn of people? Some may be good, but I remain of the opinion that more staying-in-place will, on net, have ill effects, even if the reasons for immobility are good.
  2. This study stops in 2011. Migration has stabilized sense then, possibly even risen. Is this because regional inequality of compensation for given occupations is rising again? Many regional economists have suggested convergence may be slowing or even reversed in the last 5–10 years; will we see that in migration data? Are we already? And if convergence is reversing, then shouldn’t we want migration to rise?
  3. Occupational-mix and wage-variation may have become more equal, but regional unemployment rates continue to show a very wide variance, as do regional labor force participation and regional employment-to-population ratios. Is there nonetheless a case that migration among non-workers may be at some sub-optimal level, even if labor force migration is proving responsive to normal convergence?
  4. We wanted an explanation we could apply to Australia and Canada, and this may be such an explanation…but is it? Do we have comparable data for Australia and Canada? If we look at Australia and Canada and find that the key parameters are different, or have gone the other way, what do we make of it? Right now, our sample size is effectively n=1.
  5. Do real or nominal incomes matter more? This paper seems to have focused on nominal incomes, but my guess would be that the predictive power of real incomes would be even higher. However, this may not be true across all times and groups. A young person in the 2010s may prefer nominal income due to student indebtedness. An older person on a fixed income in the 1990s may prefer real income. Knowing which of these matters, when, and for whom would (1) help guide policymakers and (2) help us understand what kind of effect we are actually seeing.

Consider the Humble Air Conditioning

I have written at length about why I think the “AC Invented the Sun Belt” story is bonkers. This article strengthens my hand.

Think about the post-war migrations south. If you think AC caused that, then this study is problematic, because this study suggests that at least back to 1970 (plausibly earlier!), migration around the country was characterized by people seeking higher earnings for a given occupation. All your “but the South is poorer on average!” stories in the world won’t change the fact that migrants selected into occupations where wages were good: in the South. And that appears to be nominal wages; real wages would be even higher.

Plus, this article suggests information constraints are a big part of what drives migration. Well, okay, imagine if from 1930–1950 we launched a huge information experiment where we reshuffled people into massive pop-up industries, deployed them to rando places with no previous economic draw for them, and broadly just moved millions and millions of people around. Like, ya know, for wartime mobilizations or something. And then what if we built a giant highway system for them to return to all those places where we deployed them for training, housing, or wartime industries.

The mobilization years were an information shock that fundamentally altered U.S. migration patterns. What built the south wasn’t air conditioning but the massive information shock of the mobilization years where millions of Americans discovered that, huh, living in California is pretty nice, or North Carolina, or what-have-you. Federal and state investments afterwards in roads, universities, and military-industrial complex industries in the Sun Belt then poured fuel on this informational fire, and every 5 years or so another new wave of Sun Belt metros starts to open up and become hip. We are seeing the repeated information shocks of Sun Belt convergence, not amenities-driven migration. Most people do not like hot, humid summers (I disagree of course; humidity is a blessing, not a curse):

All that to say: if you lean on the brute explanation of climate, you’re (1) ignoring the cutting edge of migration research, which is largely focused on information, the life cycle, composition of flows, etc, and (2) just plain wrong.

Check out my Podcast about the history of American migration.

If you like this post and want to see more research like it, I’d love for you to share it on Twitter or Facebook. Or, just as valuable for me, you can click the recommend button at the bottom of the page. Thanks!

Follow me on Twitter to keep up with what I’m writing and reading. Follow my Medium Collection at In a State of Migration if you want updates when I write new posts. And if you’re writing about migration too, feel free to submit a post to the collection!

I’m a native of Wilmore, Kentucky, a graduate of Transylvania University, and also the George Washington University’s Elliott School. My real job is as an economist at USDA’s Foreign Agricultural Service, where I analyze and forecast cotton market conditions. I’m married to a kickass Kentucky woman named Ruth.

My posts are not endorsed by and do not in any way represent the opinions of the United States government or any branch, department, agency, or division of it. My writing represents exclusively my own opinions. I did not receive any financial support or remuneration from any party for this research.