What Happened to Migration in 2015?
IRS Statistics of Income Edition
We just recently got new IRS SOI migration data, one of the major migration datasets I cover regularly. It was a bit delayed, but ultimately came out a bit over a week ago.
But I’ve taken my time in discussing it because I wanted to get clarification from the IRS about the data. I’ve written extensively about the structural features of IRS data several times before so, when I saw a major oddity in the data this time around, I thought, “Well, golly, that’s worth explaining. It’s probably a cool story, and maybe I can help make this data even better!” So I emailed the migration analyst at the IRS (swell dude), had some back and forth; then a phone call; then more emails.
Unfortunately, I have failed you. I do not have a satisfying explanation for the data I am about to present, nor a corrective to make it more useful.
What’s the big deal you ask? Let me show you:
Look at that drop! That’s the steepest decline in migration we’ve seen for any source in any year.
Now let me make one thing clear.
Actual migration absolutely did not decline that much.
I can tell you with near-absolute-certainty that what we’re actually seeing is a discontinuity in the data, and specifically a change in how data arrives into the IRS SOI master file. But what I can’t tell you is whether last year or this year is a more realistic depiction of migration rates for the population (IRS tax filers) in question.
Let’s walk back a bit here.
After 2010–11, the Census Bureau stopped assembling migration data from the IRS SOI data. However, there was such an outcry about this, that IRS took up the responsibility for the whole project itself, beginning with 2011–12. They then added more and more information, greatly improving the quality of the product. I’m particularly happy about the addition of 1st year/2nd year AGI values, as that was a change I suggested. Go me! That change has since been retroactively applied back to 2011–12. The IRS also made a huge improvement to their system that identifies migrant returns, incorporating a large number of returns that were previously missed. That’s what explains the jump up in the beginning of the highlighted section.
For this most recent year, the IRS SOI analysts observed that the migration value fell dramatically, and then went hunting for answers as to why. They ran the same processing system on the master file that they always do; there was no change in methodology on their part.
What appears to have happened is that something has changed in how returns are passed into the IRS’ internal database. This may mean that commercial tax preparers received new guidance on how to list addresses. It may also mean (as a note in the user guide now says) that IRS identity-theft protections have kicked in, and many of those former “migrants” were actually stolen tax returns. It could also be something else: maybe changes in peoples’ choice to use direct deposit or a mailed check? Who knows? Somehow, some way, there has been an administrative change in IRS’ front-end system, greatly reducing the number of returns passed on with address changes. Neither I nor the IRS migration analyst know exactly what this change was.
The result is that gross migration fell between 25% and 35% in every state (the decline is extremely uniform; the only exception is Texas, where outflows only fell 7%). This puts IRS gross migration far below any previously measured value. Ruh-roh.
However, the impact on net migration was far more muted. Here’s what I mean:
See? The amount of population redistributed as a share of the total population fell by far less than gross migration, and in fact remains above levels seen during the recession (although, again, there was a method change in 2012 that could impact these estimates).
Looking at the states, the absolute value of migration for states shows a wide variation. In some states, net flows got bigger. In some, they got smaller. They were indeed smaller on average, but in 8 states, net migration actually became larger (i.e. more deeply negative or more steeply positive).
In other words: the administrative change reducing measured gross migration had a much smaller impact on measured net migration. This means that the effect of that administrative change was fairly uniform.
Comparisons of gross migration from 2013–14 to 2014–15 are worthless, but net migration comparisons should hold up somewhat better.
Well okay then! Let’s look at net migration!
State-Level Net Migration Rates
The map below shows net migration rates for each state in the most recent data.
And here is last year:
The big change here is, of course, Texas. Texas went from being a big gainer state to having just a moderate inflow, mostly because of worsening conditions in 2015 in the Texan oil fields. Curiously though, North Dakota shows no similar decline.
We can directly compare these differences in one map as well.
Yeah, so, Texas’ migration rate worsened. Whereas gross outflows fell by a fairly uniform 25–35% everywhere else, they only fell 7% in Texas, even though Texas experienced a similar 25–35% decline in inflows, just like every other state.
The result of this is that Texas went from being about 6.3% or 6.4% of national outflows from 2011–12 to 2013–14 to being over 8.1% of national outflows in 2014–15. As oil field employment hasn’t improved much since spring of 2015, we can expect that the next IRS migration file will show a similar or lower level of net migration.
With all these extra outflows, other states have benefited. Many workers who moved to Texas may even be coming home.
Does Migration Boost Incomes?
I’ve discussed this topic in extraordinary length elsewhere, especially as it relates to IRS data. Here, I will just provide a few simple charts elaborating on the generally positive association between migration and income growth.
The chart above shows that, since 2012–13, interstate migrants have seen their income rise faster than non-migrants. The gap has been fairly consistent: about 4%. In 2011–12, this trend did not hold. It is possible that since migration is correlated with employment changes that periods of increasing disemployment would show much lower migrant income growth (migration due to being fired), while periods of rising employment would show the opposite (migration due to getting a new job). It’s not entirely clear. But many other studies also show the pro-mobility impacts of migration, so I won’t belabor the point more here.
Broadly speaking, this year continues last year’s trends: income gains from migration decrease with age and increase with income, though both of these largely reflect patterns like retirement, or just the natural result of IRS’ sorting methods.
Change in County Outflows
We can zoom in for more detail than the state level, however.
Of particular interest might be country inflows and outflows. Now, of course, gross flows have fallen dramatically, so if we just compare county flows this year to last year, we’ll see that nearly everywhere had falling rates, which isn’t helpful. So instead, I normalize county flows by adding to them a percentage value equal to the total estimated YoY change. In other words, I assume that there was no change in the “true” aggregate migration rate. You can quibble with this, but it’s a reasonable-enough way to look at some changes.
First, let’s see the areas with high outflows.
Outflow rates range from 1% to about 20%, with a few statistical outliers due to small populations or poor data coverage. By and large, however, you can see high outflows throughout much of the south and west, and especially North Dakota, while outflows are lower in Pennsylvania, Appalachia, the California coastal areas, and much of the northeast.
But is any of this new?
Well, it does seem like outflows rose in many deep-south counties, as well as in parts of Texas. We also certainly see higher outflows in North Dakota. Outflows also seem to be edging higher in Appalachia, California, the Northeast: all those places we said had low outflows. Meanwhile, there is an outflow decline in most of the southeastern Atlantic and Gulf coastline, and in the core rocky mountain areas.
What about inflows?
Well… same trend as outflows. Looks like some areas are just high inflow and outflow.
What about change in inflows?
Here we can see the same decline in inflows as we saw for outflows along the southern Atlantic and Gulf coasts, and especially around the DC metro area. But where we really see a decline in inflows is around Austin, Texas. Meanwhile, we actually see rising inflows in the deep south, Appalachia, California, parts of the northeast… even as inflows fall sharply across much of the great plains and west.
But now that I’ve shown you all this, I’m sure you want the net rates. So be it.
Viewed in net terms, it becomes clear that the Deep South and Mississippi River valleys are epicenters of negative migration. Travis County, Texas is experiencing some of the most steeply negative migration rates in the country, though much of that is just suburbanization. The Pacific Northwest is gaining people, while most of California is stagnant. Most Northeastern counties show moderate to slight losses, while the mid-south and Florida continue to show strength.
You know what’s next.
How is this different from 2013–14?
To start with, let’s note how little regional trend there is. There are very few large geographic areas all in the same color. Some clusters, the Austin metro area, or the Texan and North Dakotan oil fields, or northern Michigan, seem to have declined, but these are mostly small areas, not huge swathes of the country. Meanwhile, we can see that coastal North Carolina, as well as much of the western Great Plains, have seen some improvement, though, again, this is far from uniform.
In general, it’s hard to give a general narrative for these changes. Plus, we really have no way of knowing how much of an effect the little-understood mysterious administrative change may have had in altering region-specific measured migration.
Conclusion
It’s not clear what exactly caused the sharp change in IRS SOI migration data, nor whether the new or old data is a more accurate measure. The peril of using administrative data like IRS data instead of a survey like the ACS is that administrative data is usually less consistent over time due to legal, regulatory, or administrative changes, and because administrative data is often “gamed” by respondents due to its primary use for tax collection or some other government function. Nonetheless, this year’s data due show us some worsening migration in the Texas and Dakota oilfields, as well as the Austin metro area, while net migration has risen by more moderate amounts across much of the nation. IRS data continues to give suggestive evidence that migration is associated with faster income growth, although as usual this is only suggestive, not conclusive on its own.
Hopefully next year’s IRS data will clear up some of the questions as we see whether this low level persists, or if migration rates shoot back up. Time will tell.
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 graduate of the George Washington University’s Elliott School with an MA in International Trade and Investment Policy, and an economist at USDA’s Foreign Agricultural Service. I like to learn about migration, the cotton industry, airplanes, trade policy, space, Africa, and faith. 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. More’s the pity.
