Are Population Statistics Really Misleading About Middle America Cities? The Case of Indianapolis.

Introduction.

A few weeks ago, urbanist Conor Sen wrote a Tumblr post on "How Population Statistics Mislead About Middle America Cities".

He wrote:

"On my road trip I saw this over and over again — most of the neighborhoods we visited were in the walkable/downtown parts of the city. [...] According to Census data Indianapolis is adding a few thousand people a year to its population base of 820,000. I don’t have a good way of analyzing the data, but if Indianapolis is losing a couple thousand people every year on the outskirts and gaining several thousand in that inner core, that’s a huge deal for the downtown, for density, and for the vitality of the city as a whole... simply looking at aggregate population data for [cities like Indianapolis] would have you miss this."

So I thought: why not check the tract-by-tract Census estimates, and see if aggregate Indianapolis population statistics really are missing this "outskirts losing people, downtown core gaining people" trend?

(Sen mentions other cities--Pittsburgh, Cleveland, Columbus--and I haven't finished mapping those yet, so the conclusions here might or might not apply to them.)

Mapping population change in Indianapolis.

Here is a map of Census tracts in Indianapolis (coterminous with Marion County, IN) colored by percentage change in total population from 2000 to 2012 (using Census data and American Community Survey 5-year estimates, respectively). I used my usual "combine tracts into clusters" approach, and as usual, red means increase and blue means decrease.

You can click any tract to see a little psuedo-scatterplot of the 2000, 2009, and 2012 population estimates, with the latter two being ACS 5-year estimates. The labels specify when 2010 tracts are aggregated together, although right now there are still individual 2010 boundaries on the map even for the aggregated tracts. Click the same tract again to remove the label.

As you can see, according to these numbers, only one "downtown Indianapolis" tract had significant population growth from 2000 to 2012.

This particular tract is Tract 3542. The population increased from 3,699 in 2000 to an estimate of 5,650 in the ACS 5-year results.

A couple of other urban core tracts had small gains, but most of central Indianapolis is actually estimated to have lost population from 2000 to 2012. The tract immediately above Tract 3542, Tract 3909, went from 2,544 people to 1,891 (apparently mostly because of African-American population loss, incidentally--see below).

More generally, the population of Center Township (which covers central Indianapolis/Marion County) is estimated to have declined from 167,055 (2000 Census) to 145,097 (2012 ACS 1-year estimates).

Meanwhile, many tracts on the outskirts of the county--although not necessarily on the outskirts of the city as Sen defined it, I'll admit--soared in population. Tract 3801--the triangular Southwest tract--more than doubled from 7,054 to 15,864.

And these aren't just large percentage increases from small absolute increases. Tract 3904.02 (one of the skinny rectangular tracts in the Southeast) had 6,582 people in 2000 and 12,873 estimated by 2012.

What about recent trends?

Still: 2000 to 2008-2012 is a pretty broad timespan, and looking at change over that whole period might be masking more recent trends, especially since migration patterns might have changed since the recession.

I added an option coloring the map by comparing 2005-2009 estimates to 2008-2012 estimates. Obviously, there's a lot of noise and so on in overlapping five-year averages, but I figured any really strong trends would presumably show through that.

The resulting map isn't exactly the same as the first map, but I wouldn't say the trends are totally different either.

We still have only one downtown Census tract with significant population growth, albeit a different one. This time it's Tract 3562, the population of which apparently dropped from 3,144 to 1,844 between 2000 and 2005-2009 before leaping to 2,721 for the 2008-2012 estimates. (In other words, the population dropped over the 2000-2012 timespan, but there's evidence of a rapid recent increase, apparently mostly non-white, interestingly.)

Tract 3562, incidentally, seems to contain the corporate headquarters of major Indianapolis employer Eli Lilly.

On the other hand, the only downtown tract to show significant growth over the 2000-2012 timespan, Tract 3542, apparently had most of its growth early on. The 2009 five-year averages give an estimated population of 5,654, versus an estimate of 5,650 in the 2012 five-year averages.

It's certainly possible, of course, that Census data is missing people who recently moved to downtown Indianapolis, but the same applies to people moving to the periphery. Either way, there doesn't seem to be evidence that the city "is losing a couple thousand people every year on the outskirts and gaining several thousand in that inner core". Broadly speaking, these population estimates seem to point in the other direction.

Different shifts across racial/ethnic lines.

While it doesn't directly relate to Sen's claim, there are some interesting splits in population shift patterns across Indianapolis' racial and ethnic lines. I added options to look at the percentage change in the non-Hispanic white and (non-Hispanic) African-American populations.

You have to be careful with percentage shifts when some of the base amounts are so small (one of the more trivial negative effects of segregation), as well as with margins of error (so don't read much into small shifts or shifts in small groups). Also, some tracts actually have missing values for racial/ethnic groups, which I coded as "0".

Quite a few central and north-of-central Indianapolis Census tracts had large percentage and/or absolute increases in their non-Hispanic white populations, as did a few of the high-development areas in the southern outskirts of the county.

Meanwhile, a kind of V-shaped area had notable declines in non-Hispanic white population. Consider northwest Indianapolis, also known as Pike Township. From 2000 to 2012 (and we can use ACS 1-year estimates for a geography this size) the non-Hispanic white population dropped from 40,167 to 27,306 while the African-American population increased from 23,002 to 36,531.

It's surely not a coincidence that the area's city council district elected its first Democrat, José Evans (who is African-American and Hispanic), "in history". (Although Evans apparently later switched parties to Republican. He should probably take that "first Democrat elected in history" part out of his biography.)

Similarly, the African-American population dropped in some of those same central and north-of-central Indianapolis neighborhoods, but there were huge percentage and/or absolute increases in most outlying areas.

Just one example: The African-American population of Tract 3614 increased by more than a factor of 10, rising from 195 in 2000 to 1,983 in 2012.

You can also look at the shift in Indianapolis' Hispanic population, which again was particularly strong in the periphery, either in absolute terms or in percentage terms.

Conclusion/A possible explanation?

To speculate a bit about a possible explanation: Sen took a trip to Indianapolis, and wrote that most of the neighborhoods he visited were in the few "walkable/downtown parts of the city". But population doesn't measure where people visit--it measures where people live.

The outskirts of Marion County might not have much to interest a tourist, unless you want to visit Dick Lugar's family farm. But what they have is a lot of room for growth--recently rural space (like the Lugar family farm, which some people were surprised really was in Indianapolis proper) where new developments can be built.

For example, according to Wikipedia (I feel silly citing it, but I may as well source my impressions), Decatur Township, in Southwest Indianapolis, was "[l]ong one of the most rural sections of the county, [but] has seen many new residential and commercial developments. AmeriPlex, one of the largest industrial parks in Indiana, is in Decatur Township [as is the new airport terminal]".

Similarly, Franklin Township, covering the even-higher-growth region of Southeast Indianapolis, was "one of the last areas in Marion County to see heavy suburban land development from the outward expansion of Indianapolis. This is primarily because of a history of family farms in the area, but is also due to various civic efforts to maintain a rural atmosphere in the area over time."

Again, these kinds of neighborhoods (which are technically part of the city of Indianapolis despite being townships) probably won't be tourism or nightlife hotspots. But they are residential hotspots, and seem to be where Indianapolis is most notably gaining residential population.

Data sources and notes.

As I said, I used numbers from the 2000 Census, from the 2005-2009 American Community Survey estimates, and from the 2008-2012 American Community Survey estimates.

All of the data sources, and the script that grouped things into clusters and generated the GeoJSON and so on, can be found here.

As is my usual approach, I took the 2000-2012 Census tract relationship files, threw out polygons with no 2010 population, and then used a "union-find" algorithm to group tracts into comparable "clusters".

A quick note: For the 2000 Census data, I used "VD05,Not Hispanic or Latino: - Population of one race: - White alone" and "VD06,Not Hispanic or Latino: - Population of one race: - Black or African American alone".

For the 2009 and 2012 ACS data, I used "HC01_EST_VC76,Number; Estimate; HISPANIC OR LATINO AND RACE - Total population - Not Hispanic or Latino - White alone" and "HC01_EST_VC77,Number; Estimate; HISPANIC OR LATINO AND RACE - Total population - Not Hispanic or Latino - Black or African American alone". I think those are directly comparable, anyway.