Winners and Losers In the 2016 Population Estimates
Poor, Poor California.
Census released their first estimates of 2016 population this week. Lots of attention will be given to who grew at what rate from 2015 to 2016, which is fine. But what often gets forgotten is that these numbers are revised every year, often substantially so. Indeed, this year, population estimates for 2011–2015 were reduced by a significant margin for most states. In this post, I’ll walk you through those changes.
Also, Merry Christmas!
The main takeaway is simple: Census has sharply lowered their estimates of US population due to much lower estimates of recent net international migration, with the result being that population was cut most sharply in states with large immigrant populations. Examples include Hawaii, New York, California, DC, New Jersey, Nevada, and New Mexico.
A note: when I say “Vintage 2016,” I mean “the Population Estimates released in 2016, which describe the most recent population estimates for the period 2010–2016”. When I say “Vintage 2015,” I mean “the Population Estimates released in 2015, which described the estimates that were most current in 2015, for the period 2010–2015.”
Mapping Census Population Revisions
There you go! Most states saw a downward revision in their estimated population. Exceptions were in Wyoming, Vermont, and Maine. To be honest if you’d asked me to guess which states Census was underestimating for population estimates, those states would probably not have been my top picks. Meanwhile, on the flip side of the coin, Census’ deepest cuts in estimates were to California, Hawaii, New York, and New Jersey. Those states together got their populations whacked by 230,000 people. Overall, national population for 2015 was cut 550,000 people.
Here you can see all the cuts lined up by percent of Vintage 2015 estimates of 2015 population:
And here, for the curious, is a comparison time-series of estimates for total US population:
Now, I know what you’re thinking: “Lyman, those are two nearly-identical lines that, if you used a different scale, would be exactly identical! You’re making this seem big when it’s actually really small!”
Nope. This is a substantial cut. To see why, it may help to compare annual growth rates of population in Vintage 2015 vs. Vintage 2016 data.
Now, again, those differences aren’t huge… but they are pretty big for revisions. We just revised away a substantial amount of population growth and, as you saw, those revisions had very non-neutral impacts across regions. There were winners (Wyoming) and losers (California).
The last few years have featured debate between people who think big coastal cities fueld by high immigration are the population future, and others who think “Sunbelt”-style stuff is the future. These revisions suggest that previous data used to support the view of the Team Coastal Cities was probably not entirely accurate, and Team Sunbelt/Heartland was probably on to something.
But what caused these revisions? Let’s look at that.
It’s All About Migration
To start with, understand that Census made zero net revisions to births and deaths for the nation on the whole for years before 2014. A few states had very small changes but, overall, there was virtually no serious change in core natality and mortality data in the back years. The reason for this is that Census uses pretty definitive birth/death data from medical and Social Security records that doesn’t need major revisions. So the most recent two years sometimes have revisions as that data comes in, but it eventually gets pretty solidified.
Migration is a different story. Regular readers know that internal migration data comes from numerous sources and can have a wide margin of error. So internal, domestic, migration estimates continue to be revised for years. Substantial revisions to domestic net migration for 2010 still show up in the Vintage 2016 data.
But where things really get crazy is international migration. Regular readers of this blog know that (1) immigration data is highly varied, especially if you’re trying to tie down where this immigrants live and (2) emigration data is a total and complete crapshoot. So estimating net international migration is really tricky.
So! Let’s look at revisions. To start with, I’m going to do something weird: I’m going to add up the absolute value of all changes to state-level components like birth, death, domestic net migration, and international net migration, and add those up, so you can see which components make have very big changes.
As you can see, international migration revisions are significant all the way back to 2011. Revisions to births and deaths are insignificant until 2014. Revisions to domestic migration are small but meaningful back to 2011, then get moderately sized in 2014 and 2015.
Because Census doesn’t give gross flows, it’s not clear how much of changed international migration is lower immigration versus higher emigration, but I suspect it’s a mixture of both. Census uses rolling 3-year averages of ACS immigration, based on availability, which means for the 2015 estimates they were using 2012, 2013, and 2014 ACS-measured migration, which probably would reflect some lower immigration, especially from Mexico.
We can also look at the net, cumulative change to population from each component of change over time. Here we go:
The above chart shows the net revisions to the US total population estimate by population component. Because domestic migration balances out at zero no matter what the state-level changes are, its net effect on total population is always zero.
And here you can see the story of the Vintage 2016 population estimates. Net international migration is solidly half a million lower than previously believed over the period in question.
That is a big deal. How big a deal?
Well, I showed you before the map of changes in state populations in 2015 in Vintage 2016 versus Vintage 2015. Below is the same map, but its Vintage 2016 versus Vintage 2015 assuming international migration was unchanged; that is, only changes to estimates based on different births and deaths.
The story here looks very different. In this map, California is one of the very best performers, along with Maryland! California’s estimated births were lowed slightly, but the estimate for deaths was lowed dramatically: by about 13,000 deaths across the period.
Meanwhile, West Virginia shows up as a pretty abysmal performer. For total revisions, West Virginia was just middle-of-the-pack. But restricted to natality/mortality-driven revisions, West Virginia looks worse, as birth estimates were lowered and death estimates were raised.
You are likely going to see headlines talking about new population estimates. They may be reasonable stories or they may not be. They are very unlikely to take any retrospective look at what these estimates are really showing us, that is, evolving views of recent population history. Census estimates are just that: estimates! They change as we get new information! More interesting than their highly-erratic and error-prone most-recent-year estimates is the evolution and eventual stability of their back-year estimates. And this year, the back year estimates showed substantially lower population than previous estimates, almost entirely due to lower estimates of international migration. Here’s a simple chart of those estimates:
PS- Puerto Rico
Estimates for Puerto Rico are not as heavily revised partly because there’s less data for Puerto Rico, so the changes in back years were essentially non-existent. What we can see, however, is that Puerto Rico’s net migration probably stayed very low in 2016, but likely did not worsen.
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.