« Mega-Regions and Urban Policy | Main | Mega Debate »
Paul Krugman is a mega skeptic.
It’s not at all clear to me that world competition is between mega-regions.
I’d say that there are two things that arguably define an economic unit for the purposes of economic geography. One is labor mobility: a region over which there’s high mobility of labor will be a region in which everyone with the same set of skills is paid more or less the same real wage (which may differ in money terms because of differences in the cost of living etc.). By that definition, the United States as a whole is the relevant unit: workers are as mobile between Chicago and Boston as they are between Baltimore and Boston.
The other definition is the reach of spillovers — positive externalities, for the econowonks. That’s probably much more localized: there’s a reason investment bankers cluster in expensive Wall Street or City of London locations. But again, it’s hard to see that this makes the Northeast Corridor, as opposed to individual metro areas within the corridor, a relevant unit.
I'll have more to say later (it's a busy weekend), but this paper has more on why the mega-region, alongside the cluster and the nation-state, is an increasingly relevant and important economic unit and why it needs to be considered as such from a policy-making perspective. But let me just say that when 40 of these megas which account for less than a fifth of world population account for roughly two-thirds of economic activity and 85 percent of global innovation, something is going on. Another piece of the explanation (micro-foundations if you will) is outlined in this paper with George Mason's Rob Axtell, which uses adaptive agent models to suggest that mega-regions are in fact emergent economic units, emerging that is from the evolution of localized clusters and city-regions. At the end of the day, mega-regions have large geographically defined markets, and people are more mobile across mega-regions than across nations (Krugman even calls his own mega, Acelaland, after the fast train). My hunch is that people are also much more likely to relocate within megas than across them, say from NY to Boston or to DC, and in fact recent conversations with many journalists, non-fiction writers and editors suggest a shift from NY to DC, Krugman's Times colleague, David Leonhardt being a case in point, though more research needs to be done on this issue of mega-mobility.
II'm in good company: Krugman's also also skeptical of the new paper on place-making policy by Ed Glaeser and Joshua Gotltlieb (unfortunately gated by Brookings). It's always great to have Krugman applying his ever facile intellect to these issues.
Your thoughts?
Krugman is, as always, admirably honest and clear about his assumptions -- and I think almost certainly wrong about the assumption about workers being as mobile between Chicago and Boston as they are between Baltimore and Boston. That is the crux of it -- are megas recognizable as labor markets? I think they are.
This is eminently testable -- the data are out there to construct the relevant migration matrix -- but I have not been able to lay hands on an analysis where somebody had done this and drawn any relevant conclusions. Note that this would have to be gross in and out migration (not net migration) and would probably have to be done at the county level to be really useful here. That translates to a really big matrix (3143 counties * 3143 counties = almost 10,000,000 elements!) Some simplification would be possible -- but tricky.
About the closest thing I can find is a piece by Manson and Groop (Professional Geographer, 2002, 52-3) which has in and out migration maps for LA, NYC an Chicago. NYC in particular exchanges a whole lot of population with other counties in the BosWash coridor. It also exchanges a lot of population with other big metros -- but the slant toward the rest of the mega is pretty strong. Chi and LA are similar.
Its an important issue. As a relatively recent migrant from NYC to DC who maintains a strong connection with NYC -- stronger than I could maintain from Chicago -- I am confident that there is a real labor market effect here (as well as some longer range spillovers) but if anybody has the numbers, I would like to see them.
Posted by: Tim Gulden | April 12, 2008 at 09:56 PM
You and Krugman work in different worlds. His main concern is the working class, and for someone doing assembly work the "real wage" and standard of living is probably similar between Chicago, Boston & Baltimore. In this view, the major reason one would move would be to find a job. You're looking at what makes some regions successful and how individual choices affect this. I happen to think your work is more important, but that doesn't make his wrong in its context.
Krugman raises a good point about the individual cities in a mega being more important than the whole. The hot spots of Boston, NYC and DC make the Northeast mega successful, otherwise it's just an accumulation of people, like Calcutta. In your Nor-Cal mega, California's central valley doesn't contribute a lot of value or creativity, although it's undoubtedly economically tied to the SF Bay area. Similarly in Cascadia, the stretch between Vancouver and Olympia, Washington is pretty dormant compared to Seattle and Portland. But both Nor-Cal and Cascadia are definitely economic units of some sort. There's probably a lot more work on this topic of differing components of megas.
I notice also that Krugman doesn't live in Cleveland.
Posted by: Michael Wells | April 13, 2008 at 01:54 AM
Krugman is brilliant, and understands the national economy, but seems to have a blind spot to the fact that the US is a series of very diverse regions.
He only recently decided that perhaps increased trade was hurting US wages (he certainly doesn't live in Cleveland).
This is a common blind spot of macro economists, they see the world in averages of very large populations.
Posted by: save_the_rustbelt | April 13, 2008 at 07:27 AM
A fifth of the population accounting for two-thirds of economic activity doesn't sound very significant to me--it doesn't even manage to be 80-20.
Posted by: matt wilbert | April 13, 2008 at 08:55 AM
Tim,
Why can't you just define your megas in terms of counties, then sum all of the flows from in counties (those in the mega) and all of the flows from out counties (those in the sets of other megas) and track from which megas they came. Essentially construct a flow matrix at the mega level using county-level data. You could consider all non-mega counties as well, but could leave them aside at first crack. In any event, the computational effort associated with your matrix would shrink exponentially while you seemingly would still have the relevant information for answering your question.
However, I think the real issue that Krugman was getting at was the issue of "technical" or "industrial" proximity in addition to spatial proximity. One I think could argue (and there exist several interesting new papers on technical proximity) that technical proximity may be as important to linkages amoung metros as spatial proximity. For instance is the arts industries in NYC more integrated with LA's than DC's? Is IT in DC more linked with the valley than Boston?
Do you think that migration data could answer these types of questions?
Posted by: Ryan Sutter | April 13, 2008 at 02:06 PM
Tim, this comment is in no way a raised hand to volunteer for constructing that 3143x3143 matrix (although I have the data if there's a SAS fan out there willing to do the legwork), but you are right about the NYC-DC connection. Putting aside the rest of the DC metro area counties, New York County is at the top of the list for inmigration to DC and outmigration from DC for 2005 and 2006, based on household counts from the IRS. Interestingly, Chicago is second in 2005 and third in 2006, behind Boston.
Posted by: Brian J. Kelsey | April 14, 2008 at 01:07 PM
Brian - Nice. I have long said DC's recent growth is tied to NY. It is close by and affordable alternative. Your data sheds light on that.
Posted by: RF | April 14, 2008 at 02:11 PM
> people are more mobile across mega-regions than across nations
Absolutely. Not only do people try to live within a half-day to day's drive of their extended families (esp. parents and siblings), business people prefer doing business with people within easy and cheap day-trip range (always far easier than having to overnight). Like, for example, 60 flights a day on Southwest between Dallas and Houston - plus the Continental and American flights.
Of course, both these reasons are also why Houston is more a part of the Texas Triangle mega-region than the Gulf Coast one...
Posted by: Tory | April 14, 2008 at 08:15 PM
Tory - I mention the Texas Triangle in Who's Your City. But the satellite imagery simply do not (yet) support a contiguously integrated Texas Triangle. As much as I might like to make certain parts of the world "fit" into a mega-region or "megalopolitan" definition as you suggest - Sydney-Melbourne-Brisbane or the metros of Scandanavia (areas which share travel patterns and are otherwise closely linked), the light emission data do not support. In this research we stand behind a simple yet elegant methodology for identifying (mega)regional geographic units across the globe from nighttime satellite images.
Posted by: RF | April 15, 2008 at 07:59 AM
If I understand correctly, the mega-region concept is about economic integration - less than within a metro, but more than across a nation (or continent). Light emission data is a fine starting point for identifying such mega-regions, but then it seems you have to take the next step and adjust the definitions based on common sense when the lights miss something. Lights are also obviously biased towards coastlines, as well as towards productive farmland between two cities (vs. deserts, mountains, forests, etc.). Flights (or trains) per day between two cities might be one good metric to look at when there's a question. If they're at shuttle-level frequencies, that's a pretty good indicator. The best way of correctly identifying mega-regions is through a mix of methodologies, not a rigid adherence to one.
Posted by: Tory | April 16, 2008 at 05:14 PM
Tory -
I'll make a deal with you. You collect these data and we will surely add them. No one, not a single person, has ever identified subnational economic units. There are only two groups that are trying ours and Bill Nordhaus' at Yale. We have done plenty of calibrations and adjustments when the light data don't meet common sense. But your common sense knows a great deal about Texas. Now what about India, or China or Latin America. In each of these places, there is somebody with deep domain knowledge. But that's a heck of a lot of common senses to collect. And even if we could trying to calibrate our method like that, based on idiosyncratic information, would render it, dare I say, a whole lot less systematic. We have take great care to adjust the data for all the light-sensitivities you mention, about coasts etc.
These things are easy to suggest - mix of methodologies etc. - but quite a bit more difficult to actually execute, that is to get systematic and comparable data. We have looked lots of places for flight data. How would one even begin to collect global level data on trains. So, I'm with you guy. If you can find these data we'll have a look.
No one here is "rigidly" adhering to anything. Quite the contrary, we are out there building new data sets on sub-national geographies worldwide using everything from light data to personality surveys.
Posted by: RF | April 19, 2008 at 04:53 PM
I'm glad the focus has changed to up to 40 megacities, instead of 20, as it was in 2006 creativeclass.com/rfcgdb/articles/The_New_Megalopolis.pdf
The only thing I would not look so much into patents, as it's so different between continents, besides many agree that innovation and patents has kind of a questionable correlation.
Posted by: carlos | April 24, 2008 at 05:34 AM
I sympathize with the complexity, and getting it perfect is an unreasonable standard, but that doesn't mean it can't get better and better. You put out a first version, get feedback (like from some of those "common sense" people you come across), make tweaks, release v2, and on and on. "Systematic" is not the goal of science - accuracy and truth are.
BTW, this page will give you flights between any two airports to test for shuttle-level frequencies. It says US, but seems to work for international airports too. It gives all possible routings, inc. through hubs, but always lists the nonstops first. Easy to count when you're curious about the level of economic linkage between two cities (although watch out for code shares on the same flight, but they're pretty obvious).
http://www.smartertravel.com/airfare/schedules/index.php
Here would be my suggestion: take the mega-regions you have now, and count the flights between the major airports internally that make up each region. That should give you a feel for a good standard, although that standard may need to be different for NA/US vs. Europe vs. Asia vs. South America to account for trains (need fewer flights) and regional economic differences (wealthier populations will obviously fly more). May also want to calibrate on a per population basis, because obviously two large metros will have more flights between them than two smaller ones. Something like "nonstop flights per 100,000 combined population." Then compare vs. that standard when there are questions about which metros fall in which mega-regions. Clearly a high number of nonstops per 100K population would indicate tight economic integration. I think it might even be a superior indicator than a continuous light path.
Now, if you want to add a very cool map to your site, take a world map and connect city-pairs with lines whose thickness is based on the number of daily flights between them (only for the larger cities, of course, and only above a certain min number of flights to avoid over-cluttering the map). I think you'll see some very clear mega-regions, as well as identify more and less integrated ones (and the less integrated ones clearly have potential to take it up a notch and do better, like, if I had to guess, Toronto-Buffalo).
Posted by: Tory | May 14, 2008 at 08:39 PM
On further thought, the numbers would probably come out better with "nonstop flights per million combined metro population," rather than per 100K. And of course this only applies to cities far enough apart within a mega-region to justify flying instead of driving - probably at least 150+ miles.
Posted by: Tory | May 14, 2008 at 11:06 PM