Sunday, October 3rd, 2010
Megaregional Migration
Update: The self-service tool that has easy query access to the IRS migration data, including at the metropolitan area level, is available at www.telestrian.com. You can also read the launch announcement with more info on this tool and what it can do for you.
It’s almost become a truism that human capital is of paramount importance in the modern economy. It’s also becoming evident that networks play an increasingly powerful role as a determinant of urban success. It doesn’t seem like such a leap of logic to hypothesize that a city’s human capital network might be the most important one it has. So you would think cities would consider it mandatory to have a good understanding of their human capital flows. Without that, you are flying blind.
Alas, few cities seem to have even the most rudimentary ideas about this. Why is that? For one thing, it’s not easy to determine. The Census Bureau publishes net migration figures, but that doesn’t tell you anything about where people are coming from or going to, or anything about them. The Internal Revenue Service publishes very useful information on place to place migration, including data about income flows. You’ve probably see this before, such as via the Forbes map that made a splash a few months back.
Unfortunately, the Forbes analysis and most others like it are basically useless. It’s excellent eye candy, but it has key limitations. For one thing, the Forbes map only looks at one year. More importantly, the IRS reports this data in terms of county to county flows, but this isn’t really what we want. What we really would like to understand is metro to metro flows and such. But the data is so painful to work with that you rarely see this done.
However, I cracked the code on this problem. I hope to put out a self-service tool that will revolutionize analysis of this data. If you are interesting in being notified when I launch it, or if you are interested in commissioning custom analysis in the meantime, please shoot me an email.
In the meantime, I will be sharing some interesting findings that I’m already generating with this, though I’ve only scratched the surface myself.
Human Capital Circulation and Gross Migration
Migration numbers are typically reported in terms of net migration: where you are gaining people from and where are you losing them to. That’s important analysis to be sure, but is only one piece of the puzzle. For example, net values don’t tell you the absolute magnitude of the flows between cities. With City A you could have an inflow of 10 and an outflow of 5 for a net inflow of 5 people, while with City B you have an inflow of 10,003 and an outflow of 10,000 for a net inflow of 3. The net number alone obscures important information.
I think you need to look not just at net flow, but also the individual inflows and outflows. And also at something that, with a hat tip to an old University of Louisville study, I call gross migration. That is, the inflow plus the outflow. This I believe is the best measure of human capital circulation between cities and the relative strength of their human capital network relationship.
Let’s illustrate with some quick data from Pittsburgh. Here’s a graph of the top ten net inflow sources for the Pittsburgh MSA. The data is for release dates 2000-2008, which measures people who moved between 1999 and 2008:
| Row | MSA | Total |
| 1 | Johnstown, PA | 1,305 |
| 2 | Erie, PA | 776 |
| 3 | Wheeling, WV-OH | 518 |
| 4 | Altoona, PA | 379 |
| 5 | Scranton–Wilkes-Barre, PA | 263 |
| 6 | Reading, PA | 187 |
| 7 | Buffalo-Niagara Falls, NY | 181 |
| 8 | Williamsport, PA | 175 |
| 9 | Provo-Orem, UT | 167 |
| 10 | Rochester, NY | 147 |
Here’s the same data, though this time showing the top ten net outflow cities:
| Row | MSA | Total |
| 1 | Washington-Arlington-Alexandria, DC-VA-MD-WV | -4,990 |
| 2 | Tampa-St. Petersburg-Clearwater, FL | -3,161 |
| 3 | Charlotte-Gastonia-Rock Hill, NC-SC | -3,132 |
| 4 | Phoenix-Mesa-Glendale, AZ | -2,514 |
| 5 | Atlanta-Sandy Springs-Marietta, GA | -2,262 |
| 6 | Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | -2,068 |
| 7 | Orlando-Kissimmee-Sanford, FL | -1,830 |
| 8 | Raleigh-Cary, NC | -1,640 |
| 9 | Baltimore-Towson, MD | -1,624 |
| 10 | Miami-Fort Lauderdale-Pompano Beach, FL | -1,531 |
Looking at this, what you see is a city that takes in people from regional metros and exports them largely to the Sun Belt. It’s the story you would expect of a Rust Belt regional center. But look at the 2000-2008 gross migration and we see another part of the narrative:
| Row | MSA | Total |
| 1 | Washington-Arlington-Alexandria, DC-VA-MD-WV | 22,190 |
| 2 | Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 20,864 |
| 3 | New York-Northern New Jersey-Long Island, NY-NJ-PA | 18,419 |
| 4 | Youngstown-Warren-Boardman, OH-PA | 14,758 |
| 5 | Cleveland-Elyria-Mentor, OH | 10,002 |
| 6 | Erie, PA | 9,722 |
| 7 | Miami-Fort Lauderdale-Pompano Beach, FL | 9,181 |
| 8 | Tampa-St. Petersburg-Clearwater, FL | 8,761 |
| 9 | Columbus, OH | 8,677 |
| 10 | Chicago-Joliet-Naperville, IL-IN-WI | 8,514 |
Now that’s interesting, isn’t it? Note the very large circulation numbers with New York, Youngstown and Cleveland that were completely missing from the net view. What we see from this view is a strong alignment with major East Coast metros as well as a secondary flow through the emerging Tech Belt between Cleveland and Pittsburgh.
I hope this gives you a flavor for the reason you should look at numbers other than just net.
Midwest Megaregional Migration
I ran the numbers for gross migration between the twelve large greater Midwest metros I tend to focus on here, and it’s fascinating stuff. Since Chicago is the capital of the region, I’ll share today migration between Chicago and these others cities. Here is a graph of total gross migration for the 200-2008 period. It’s MSA-MSA data – I just shortened the names for space reasons:
Milwaukee is number one, as you might expect. I find it interesting that there is such a gap between the top five and the rest. At one level it is not surprising, as migration follows a sort of gravity model pattern related to size and distance. These cities are either closer and/or bigger. On the other hand, Cleveland isn’t that much smaller or further away than, say, St. Louis. All of the Ohio cities are comparatively low. Indeed, Ohio is only the ninth highest state for migration with Chicago overall.
From this you could probably trace the outline of Chicago’s real sphere of influence from a human capital network perspective. Those top five cities clearly have a tighter relationship with Chicago than the rest. Pittsburgh and Louisville unsurprisingly bring up the rear. They are dubious as Midwest cities to begin with, and this just helps illustrate why.
Incidentally, the low standing of Pittsburgh on Chicago’s migration list, combined with its own East Coast orientation shows perfectly why Penn State has no business being in the Big Ten (IMHO). Also, while there might be a megaregion called Chi-Pitts, it’s not clear to me if Pittsburgh is really in it. Perhaps the are actually a couple of regions in the area from a human capital perspective, one centered on Chicago and another from the remains of the old metals region. This is an area for further research.
By the way, here’s the year by year on the gross migration data:
Migration has held fairly steady, with a sort of mid-decade dip. The exception is Indianapolis, which has been on a fairly steady upward streak. Here’s the detail on Chicago-Indy:
I’d also note that Indy’s ranking is, based on my gut feel, higher than it probably should be, and particularly on a run-rate basis. It’s more or less equal to Detroit and Minneapolis-St. Paul, both much larger cities, and a bit ahead of St. Louis, which is about a million people larger. All of those cities are further away, but it’s interesting nevertheless. And there appears to be a deepening of the human capital relationship. Possibly Indy simply missed the early decade spike some places had and that skews the chart, but it’s certainly something worth noting.
If you are interested in seeing the raw Chicago data, you can download it here in Excel format.
Eye Candy
In case you’re wondering, I can make eye candy too. Here’s a bonus map of net migration for Cook County, showing all the counties with which it had measured migration from 2000 to 2008. Net inflows are in blue and net outflows are in red – and no, there isn’t a political message intended! While this doesn’t indicate magnitude, it does in a sense reinforce the migration geography I noted above.
Share on Facebook
Email This
Tweet ThisTelestrian Data Terminal
A production of the Urbanophile, Telestrian is the fastest, easiest, and best way to access public data about cities and regions, with totally unique features like the ability to create thematic maps with no technical knowledge and easy to use place to place migration data. It's a great way to support the Urbanophile, but more importantly it can save you tons of time and deliver huge value and capabilities to you and your organization.
About the Urbanophile
Aaron M. Renn is an opinion-leading urban analyst, consultant, speaker, and writer on a mission to help America’s cities thrive and find sustainable success in the 21st century.
Contact
Please email before connecting with me on LinkedIn if we don't already know each other.
Urbanophile in the News
The Wall Street Journal: New York City Explores Private Deal for Meters
The Washington Post: Will the exurbs ever make a comeback?
The New York Times: Are Some Buildings Too Ugly to Survive?
The Washington Post: Is DC Becoming America's True Second City?
Richard Florida/Atlantic Cities: Why Some Cities Lose When Others Win
Twitter Feed
RT @Diggingpitt: It's not just that debt matters. Return on capital matters, or one never gets the money back.
Polis: A Tale of Two Technology Cities - http://t.co/xKWv6FOA
WSJ: Competing Agendas Push Back East Side Access Project - http://t.co/rSo3DKFC #nyc
Latest blog post: New York City Considers Parking Meter Privatization - http://t.co/P4bDmyr7
Spike Japan: The End of the Line? - http://t.co/G3Q8mskI
National Blogroll
- A Daily Dose of Architecture
- Atlantic Cities
- BLDGBLOG
- CEO's for Cities
- City Ledes
- Cogito Urbanus
- EconoMetro
- Economics of Place
- Everybody Walk
- GOOD
- Human Transit
- Kaid Benfield
- Mammoth
- Market Urbanism
- MetroTrends
- New Geography
- Next American City
- NYU Rudin Center Blog
- Pedestrian Observation
- Places: Design Observer
- Planetizen
- Rebuilding Place in the Urban Space
- Shareable
- Steven Can Plan
- Streetsblog
- The Architect's Newspaper
- The Avenue / Brookings
- The Corner Side Yard
- The Heidelberger Papers
- The Overhead Wire
- The Transport Politic
- Urban Omnibus
- urbanOut
- Where
World Blogroll
Midwest Blogroll
- ArchitectureChicago PLUS
- Bill Testa Midwest Economy
- BlogKC
- Brewed Fresh Daily (Cleveland)
- Broken Sidewalk (Louisville)
- Buffalo Rising
- Burgh Diaspora (Pittsburgh)
- Cityscapes / Blair Kamin
- Columbus Underground
- Detroit Blog
- DiggingPitt
- Global Midwest
- Grid Chicago
- I Will Shout Youngstown
- Milwaukee Talkie
- nextStL
- Property Lines (Indy)
- Rust Wire
- Twin City Sidewalks
- Urban Indy
- Urban Milwaukee
- UrbanCincy
- VanishingSTL
Awesome post…I didn’t know data analysis was your thing!
I’m surprised the gross/net distinction hasn’t really been analyzed before. I certainly understand how my analyses as a supply chain professional can be failures if I fail to understand the distinction.
For example, with forecasting, measuring performance is vital. But measuring mean percentage error is not the same thing as mean absolute percentage error. Mean percentage error will never tell you how accurate you are because your overestimates are cancelled out by your underestimates. But this metric does tell you something very important: bias. In other words, whether you are more likely to overestimate or underestimate true demand. Mean absolute percentage error, which is similar to “gross” tells you your actual forecasting error.
Unfortunately, too many recent graduates enter the field without knowing the distinction, and as a result, make incorrect decisions by misinterpreting what their data really means.
Personally, I think all data analysis education should be accompanied by grammar and semantics education.
Thanks, Danny.
What Danny said. There’s room for both kinds of data here. We can learn about Chicago and Pittsburgh by seeing both who they exchange the most people with, and where the net migration flows are.
Very interesting, and much-needed.
Incidentally, there definitely is a distinct top-level region in between the Hortheast Coast and the Chicago-centric Great Lakes region. Hypothesizing such a region solves a lot of things that are otherwise puzzles–this being one of them.
Glad you found the county shapefile you were looking for. Looks great.
Provoking stuff, way more so than simple net numbers. I look forward to reading more about it as your refine your methods, and find applicable scenarios to utilize it.
Aaron, in regards to your comments on why St. Louis has such a stronger connection with Chicago than Cleveland, despite their being similar size and distance away, I think it’s because there’s so much more between Cleveland and Chicago. There are numerous other cities. If you think of the social network as planetary “gravity”, with Chicago being the Midwest’s sun, I bet you would see that Cleveland’s “Social Gravity” pull is much stronger from nearby cities like Pittsburgh, Columbus, and Detroit than it is from Chicago.
On the other hand, there’s not much between St. Louis and Chicago to distort their Social Gravity, so there’s a much stronger connection.
I will be interested to see if your data site supports this theory when it’s fully up and running.
Being a Chicago native who did two years at tOSU, I did notice that the Cleveland and Columbus people tended to look more favorably at New York than at Chicago, particularly in music. Also the expectation among graduates was that they would move to FL, TX, AZ, GA, CA or back east. Chicago and St. Louis share pull from U of I Champaign. I think that the pull from Detroit and more recently Indiananpolis to Chicago is of such volume that Ohio people scatter elsewere.
That Washington to Pittsburgh thing is not surprising, they are actually not far from each other driving wise. I would say that Western PA probably feeds the greater DC areas and even Charlotte and Raleigh the same way Detroit feeds Chicago.
I have noticed that in the past few years the Governemnt has started to clump cities together in statistics. Like putting Philadelphia,Camden and Wilmington togther.Or by putting Baltimore with Towson ,MD.
This sort of thing tends to distort statistics somewhat.Camden ,Philadelphia and Wilmington are in three different states.And i dont think that they have a lot in common.
As for Towson,its a wealthy suburb of Baltimore.And not the biggest suburb either , i would imagine.Lumping it in with Baltimore tends to distort things.For instance ,without including Towson , im sure that the outflow rate would probably be higher.
There are many reasons why people leave Baltimore City.Crime,high property taxes and blighted neighborhoods.None of those things are present in Towson.So lumping the two togther doesnt help much.Especially since some of the outflow from Baltimore City goes to Towson and other nearby suburbs.
Pete, clumping together those cities together accounts for economic and social interactions that are now routine; for example, commute and labor sheds.
Political boundaries are too glacial to adjust to what statistics are covering.
I analyzed the migration between Allegheny County and the four core counties of the Chicago MSA. Allegheny County had a net migration gain from Lake and Will Counties, and a net loss to Cook and DuPage Counties. However, Allegheny County had an average inmigrant income advantage over Cook, Lake and Will Counties, and a disadvantage only to DuPage County.
Basically, does the increase in average income benefit Pittsburgh in any way, or is the aggregate wealth transfer the only thing that matters? Allegheny County lost ~$2,000,000 of aggregate wealth to Chicagoland in spite of a net gain in residents with higher average incomes.
I think you need to look at all these types of numbers to understand the flows. I’m not certain how many conclusions you can draw about whether something is “good” or “bad” and what is driving these, but you can at least have facts on the table.
Aggregate income flows are heavily correlated with tax return migration. One thing to look for is whether you are measuring the flow based on a household proxy (tax returns filed) or a people proxy (exemptions).
By the way, 2000 to 2008, Pittsburgh lost about $50M in income to Chicago on an MSA basis. That’s around $5.5M/yr on an average basis. To put this in perspective (and to mix an incompatible source, btw), Pittsburgh’s total personal income in 2008 was around $100 billion.
Pittsburgh’s outbound household size to Chicago is 1.49 and its inflow size is 1.69. This is possibly consistent with younger people moving to Chicago and older families moving to Pittsburgh.
Hmm. Seems from that like a high-speed train between Chicago and Minneapolis, or one between Chicago and Indianapolis, would be quite successful.