Tag Archives: census

Post-Recession Intra-State Variability in Metro Growth by Size

One of the changes in demographic trends after the Great Recession was a shift of growth from smaller metros and suburban counties to larger metros and urban counties. There has been a lot of ink spilled over how much of this trend is caused by different factors:

  • Structural factors, reflecting permanent changes in the economy and preferences. The story here is that the economic benefits of being in a large city relative to being in a small city have increased, and that people like cities more than they used to, both of which are causing more people to want to live in cities.
  • Cyclical factors, related only to the severe economic downturn and its concentrated impact suburban housing values. The story here is that suburban areas in large cities suffered more because their housing markets crashed harder, but when housing markets recover, it will be business as usual like it was in the 1990s and 2000s.
  • Demographic factors, related to a large cohort of twentysomething Millennials who moved to cities at a time young people always move to cities, which just happened to coincide with the Great Recession. The story here is that when Millennials get old, they will decide they want suburban houses and it will be business as usual like it was in the 1990s and 2000s.

My personal take is that it is a combination of all these things and more, but I always expected that suburban growth would ramp right back up because zoning makes it so hard to build new housing in cities. The last few years have disabused me of the idea that things would go back to business as usual. Suburban growth in the Inland Empire has not taken off like I thought it would, and cities in the Central Valley have not recovered to their previous growth rates. It seems that perhaps there have been structural changes to the US economy that increase the benefit of being in a large city. Perhaps upper class workers are also less willing to commute long distances, leading to lackluster suburban growth.

In this blog post, we’ll explore intra-state variability in growth by metro size, comparing the 1990s, 2000s, and 2010s. The goal is to see if there have been changes in relative growth rates between large, medium, and small metro areas. The reason to look at intra-state differences is that we want to ex out the impact of regional differences in growth in the US. We stand to learn more about the differences between large and small cities by comparing Tucson to Phoenix than to Chicago. We will look at a few western states (AZ, UT, and WA), a few southern states (GA and NC), a few Midwestern states (IL, OH, and MI), and California and Texas (in a league of their own, of course). I’m not going to look at Florida because of the heavy impact of retirees, who are less influenced by economic opportunity.

This analysis does not consider rural areas, which are continuing their decades-long loss of population to cities.

The West

Looking first at Arizona, no MSA has recovered to its 1990s or even its 2000s rate of growth, somewhat understandable in the case of Phoenix, since you can’t grow logarithmically forever. Tucson has always grown more slowly than Phoenix, and the growth differential is consistent. Smaller metros like Prescott, Yuma, and Lake Havasu were growing faster than Phoenix in the 90s, at the same rate in the 00s, and slower today. One note of caution is that those three areas see many retirees, so there could be factors besides economics at play. Growth in total population in MSAs in Arizona has consistently lagged Phoenix.


In Utah, the smaller metros are still growing as fast or faster than Salt Lake City, but the growth differential has shrunk since the 00s. Of particular note, the Salt Lake adjacent MSAs, Ogden and Provo, have seen their differential shrink by over 1.5%. Unlike Arizona, growth in total population in MSAs in Utah has consistently exceed Salt Lake City, though by less in the 10s than the 00s.


Finally, in Washington, we see the most striking change. In the 90s, most of the smaller metros were growing faster than Seattle, and the second biggest MSA, Spokane, was keeping pace. In the 10s, Seattle is growing faster than all except Kennewick, which has still seen its growth differential shrink by 1.5%. In the 90s and 00s, growth in total population in MSAs in Washington exceeded Seattle’s growth, but that has flipped in the 10s. Seattle is one of the few MSAs in the country that has grown faster in the 10s than in the 00s – let’s go Seattle!


The South

Looking first at Georgia, the story is a little different. Atlanta has long dominated the state, with other areas almost never growing as fast. However, unlike in the West, the general trend has been for the gap to close. This is not due to faster growth in the smaller areas; all of Georgia is growing more slowly than in the 90s and 00s. In fact, some small MSAs in Georgia have had negative growth in the 10s, but the decrease in growth in Atlanta has been larger than the decrease in growth in most other metros.


North Carolina has been dominated by Charlotte and Raleigh, the latter of which is North Carolina’s second largest MSA and consistently its fasting growing MSA. A few small MSAs exceeded Charlotte in the 00s, but the trend in the 10s has been for Charlotte’s growth differential to increase. Even Raleigh, while still growing quickly and faster than Charlotte, has dropped from growing 1.5%-2% faster to 0.5%.


The Midwest

Illinois is a different picture in several ways. No state on our list is more dominated by a single MSA, with Chicago over 25 times the size of Peoria, the second largest MSA. Chicago outpaces almost all other MSAs in the 90s, but in the 00s the small metros did as well or better than Chicago, excluding the two smallest which have consistently declined. In the 10s, Chicago again outpaces the others, except Champaign-Urbana. Chicago has grown slowly in the 10s but other than Champaign-Urbana and Bloomington, the small metros declined. The lesson, as always, is to get yourself a state research university.


Ohio is yet another case, with three MSAs of almost exactly the same size, but all going in different directions. Cleveland was the largest MSA in the state in 2000, but was growing slowly, and shrank throughout the 00s. Cincinnati, counter to the Rust Belt narrative, was growing at a moderate pace and became the largest MSA in the state in the 2010 census. Columbus was a good bit smaller in 1990, but has been growing quickly, and at current rates will become the largest MSA in Ohio by about 2023. Ohio is unusual in the modern US in that we are watching the dominant MSA in a state change in real time. Meanwhile, all of the state’s smaller MSAs have grown more slowly than Cincinnati and Columbus, and many have shrunk. Get yourself a state research university.


Michigan appears to be different than both Illinois and Ohio, with Detroit as a large, very slowly growing dominant MSA, a fast-growing mid-size MSAs (Grand Rapids), and two fast-growing small MSAs (Ann Arbor and Kalamazoo). Detroit is four times larger than Grand Rapids, and its supremacy as Michigan’s largest MSA won’t be challenged any time soon, but Michigan actually appears to be a case where mid-size Metros are outperforming large metros. Trends in Flint for the 10s are probably impacted by the terrible state-inflicted water crisis. Meanwhile, Michigan’s smallest MSAs, fewer than 200k people, are underperforming relative to Detroit in the 10s, with many actually shrinking. I’m not sure about Grand Rapids, but still, get yourself a state research university.


Don’t Mess with Texas

For Texas, I’m only doing the top 15 out of 25 MSAs, because, come on. This also cuts out the oil MSAs, Midland and Odessa, which are somehow not combined into one MSA.

All of Texas has grown fast. Let us pause for a moment and appreciate how much Texas is doing to create opportunity for people who can’t afford places like, well, California.

Despite impressive statewide growth, Dallas has remained the biggest MSA, at about 500k larger than Houston in 2017 as it was in 1990. Dallas has outperformed most of the smaller metros in the state throughout the analysis period, with only Austin consistently beating Dallas. If you can do it, for the love of everything good, get yourself a state research university!

McAllen and El Paso grew faster than Dallas in the 90s and 00s, since if you don’t have a state research university, becoming an important border town after a free trade agreement is signed is still pretty good. Houston dunked on Dallas during the oil boom of the 00s, but has barely outpaced Dallas in the 10s. Still, the overall picture in Texas is dominance of the large MSAs.


Still Golden

For California, I am going to do all 26 MSAs. And if Texas doesn’t like it, they can get their own damn blog. For the sake of readability of the graphs and general interest, I will present some analysis statewide, and then also graph things in two groups, NorCal and SoCal. If you disagree with my breakdown of north and south, I’m sorry, but my decisions are final.

On a statewide basis, the majority of MSAs smaller than LA/OC are growing more quickly. However, most of them have seen their growth differential to LA shrink, and LA actually grew faster in the 10s than the 00s (take that, haters)! The MSAs that have seen their growth differential to LA increase are mostly greater Bay Area/Silicon Valley MSAs – San Francisco, San Jose, Vallejo, Salinas, and Santa Cruz – plus Sacramento. The Bay Area’s distant suburbs like Stockton and Modesto are still growing faster than LA, but not as much as they were in the 00s. The most stunning change is the Inland Empire, which went from 2.69% growth and 2.31% greater than LA in the 00s to 1.10% growth and just 0.54% greater than LA in the 10s. Given the high price of housing in LA/OC and historic trends, you can see my surprise at relatively slow growth in the IE!


Looking at northern California, the change is more dramatic. Stockton, Sacramento, Merced, and Yuba City all grew much faster than San Francisco in the 00s. In the 10s, San Francisco has greatly outpaced the smaller MSAs in NorCal, with only Sacramento, San Jose and Stockton holding close to SF’s growth.


The pattern is clear looking at SoCal as well. Every smaller MSA grew at least 0.50% faster than LA in the 00s, with the Inland Empire and Bakersfield over 2% faster. In the 10s, only the Inland Empire has managed to grow more than 0.50% faster than LA.



I think the big takeaway here is that regions are different, states are different, and it’s hard to draw any broad conclusions. Some big MSAs are growing quickly, some are growing slowly, and some are shrinking – sometimes even within the same state. Some big MSAs are outpacing smaller metros and some are not. It does seem that smaller metros are generally growing more slowly compared to big metros than they were in the 00s. The only states where smaller metros seem to be doing better than larger metros are Utah, some small metros in Michigan, and college towns.

Now, large metros in CA have been seeing their growth slow down in the last year or two. But this has coincided with these counties hitting full employment and high housing prices keeping more people from moving in. It does not, yet, seem to be translating into faster growth in the outlying MSAs of larger cities (such as Stockton and the Inland Empire) or in the smaller MSAs like Bakersfield and Fresno. If it were easier to build housing in LA, might it grow faster than the IE? With luck, one day we’ll get to find out.

Four Southern Cities

This is way off topic, but back in March when the new Census population estimates came out, I tweeted about slow growth in Alabama and negative growth in Mississippi. I’m wondering why they aren’t part of the “new south”, generally taken to be fast growing Sunbelt cities like Charlotte and Atlanta. Why does Atlanta boom, but Jackson doesn’t? Many of the responses boiled down to “because it’s Mississippi” which is sort of circular logic.

I thought it would be interesting to compare population growth since 1860 in four southern cities: Jackson MS, Birmingham AL, Atlanta GA, and Charlotte NC. Since municipal boundaries in the US are illogical, often changing, and hard to trace, I’m using county level data, comparing the populations, based on the 2015 definitions of the metropolitan statistical area (MSA) for each city.

I also added an “Atlanta Redux” where I threw out the 9 counties (out of 29) in the Atlanta MSA that still, as of 2016, have not reached population 30,000. These counties are way out near the edge of the city and were only recently sprawled into the Atlanta MSA. Their inclusion in the regional population back to 1860 probably unfairly inflates Atlanta’s size at the outset of the Civil War, when counties 40 miles out would not have had much plausible connection to the city but still had some population. I’m including both the full MSA and the reduced version to let readers compare.


Even the redux version of Atlanta was the biggest city of the four in 1860. Not much divergence is seen between Birmingham, Atlanta, and Charlotte until after World War 2. Jackson appears to have always grown more slowly, though this fate would not have been at all obvious to an observer in 1880, when Jackson was over 50% larger than Birmingham.

Plotting the population on a logarithmic scale can help us see patterns more easily, as the runaway growth of Atlanta makes the others look flat by comparison.


Plotted this way, the post-World War 2 growth of Atlanta and Charlotte doesn’t look unusual; it just looks like a continuation of previous trends. Atlanta, especially the redux version, starts growing a little faster than Charlotte after the war.

More interesting is Birmingham, which grew very quickly for decades after the Civil War, going from about 60% of Jackson’s size in 1870 to 120% of it in 1890. Birmingham even passed Charlotte for a few decades, going from 40% of its size in 1870 to larger by 1910. An observer in 1930 might have expected Birmingham to surpass Atlanta in another 10-20 years, especially the redux Atlanta. Jackson, meanwhile, experienced a strong decade from 1870-1880, but never grew very quickly again.

After 1930, Birmingham, which had been the rising star, began to fade, and was surpassed by Charlotte by 1960. So, what happened to Jackson after 1880? And what happened to Birmingham in 1930? I’ve yet to find out.

Another way to look at this data is to look at the amount of growth in the central county (where the main city is located) and compare that to the growth of suburbs. Here’s the breakdown for these four cities, with the central county on the bottom of the graph.


It is interesting that in Birmingham and Jackson, the central county stopped growing when the MSA as a whole started growing more slowly. However, it’s important to not try to read too much into this – did the MSA grow slowly because the central county was being neglected, or did the central county stop growing because there wasn’t much growth to be accommodated anyway?

One unappealing answer is that there is simply a great deal of luck in these outcomes. Obviously it helps for the city to be well run, but there are many factors making a complex result, and sometimes you just get bad luck. This rudimentary look is obviously not enough to tell us much about why these cities grew at different rates, but if anyone has any ideas, I’d love to hear them.

Census 2016: The Sunbelt Strikes Back

Just some short thoughts for now, perhaps more soon if time permits.

The Census Bureau released its 2016 county population estimates late last week, and the numbers show a reversion to the trends of the 1990s and early 2000s: Sunbelt suburbs are growing faster than legacy cities. Blame land use policies if you like (I sure will) but those are the facts.

If cities are not growing as fast as once thought, though, it hasn’t been to the benefit of rural areas. Small counties were disproportionately represented among counties that lost population. The table below shows the number of counties gaining or losing population by county size.


Of the 37 counties that had a 3% or greater population loss, 27 had population under 10,000 and 36 had population under 50,000. The only sizeable county to lose 3% or greater was San Juan County, NM, which lost -3.05% of its 2015 population of 118,701 and has lost over 11% of its population since 2010. It is the only county of over 100,000 people to lose anywhere near that much; the next closest was Cambria County, PA, which lost 6%.

Further, of the 456 counties that lost over 1% of their population, 424 had population of 50,000 or less. Only 3 had population over 200,000 (St Louis city, Baltimore city, & Muscogee County GA). Of the 45 counties with population of 1,000,000 or more, only 6 lost population: Cuyahoga OH (Cleveland), Wayne MI (Detroit), Cook IL (Chicago), Suffolk NY (Long Island), Allegheny PA (Pittsburgh), and St Louis MO (St Louis suburbs). Six had population growth over 2% (Mecklenberg NC, Travix TX, Hillsborough FL, Clark NV, Orange FL, and Wake NC).

In short, the census estimates reflect long-standing trends of rapid growth in the Sunbelt and stagnation in the Rust Belt.

It’s a little unfair to compare small counties and very large counties, because it’s easier to have a large percentage swing in a small county. For example, Los Angeles County has over 10 million people, and there are probably some communities of 10,000 therein that shrank. The graph below avoids this problem by showing the distribution of counties gaining and losing population for counties up to 500,000 people. As we can see, the distribution skews towards the negative side for counties with population under 50,000 and toward the positive side for counties with population above 50,000.


For counties with over 1,000,000 people, the fastest growing counties are all in the Sunbelt, dominated by places like NC, FL, TX, and NV. Perhaps the most impressive county was Maricopa AZ (Phoenix) which grew by almost 2% and over 80,000 people. The fastest growing was Wake NC with 2.43%.

In California, the fastest growing big county (population over 250,000) was Placer, in the Sacramento suburbs, with 1.64%. Unlike many counties, growth in Riverside County actually picked up last year, with 35,000 people or 1.48%. Riverside was also the only big county in California to have a large positive contribution from domestic migration. The negative domestic migration from California counties is shocking: more domestic migrants (75,168) left LA County than the two largest domestic migrant gaining counties combined – not coincidentally, Maricopa County with 43,189 and Clark County NV with 27,735.

There will surely be many takes on the meaning of this census data, but the simple take here is that this is just a triumph of people doing what they’re allowed to do. A lot of economic growth in the US is in the Sunbelt, so people are moving there, and they’re moving to suburbs because that’s what you’re allowed to build. As for why growth is in the Sunbelt, that’s a more complicated question. The structural reasons for the decline of rural counties and the continued stagnation of Rust Belt cities are beyond the means of this blog.

However, given California’s high and still-rising cost of housing, our story isn’t very complicated: the economy is pretty good and lots of people want to live here, but we don’t build enough housing to let them do it. High housing costs are hurting California, and the sooner we do something about it, the better.