Austin Contrarian

Punching above its class

The Texas Transportation Institute has issued its annual report on congestion in American cities.  As usual, Austin doesn't do well, punching well above its class.  Tthe average Austin traveler experienced an annual delay of 39 hours in 2007.    That's up from 32 hours in 1997, although no worse than 2006.   

Ryan Avent puts the Austin figures in perspective:

The average traveler in the New York metro area faces 44 hours wasted per year, for instance, while the average traveler in Los Angeles loses 70 hours per year to congestion, even though New York’s metropolitan population is much, much larger than LA’s.  More interesting still, Austin and Raleigh aren’t that far behind New York with 39 and 34 hours wasted annually, despite the fact that both metro areas have less than two million people while greater New York is home to 20 million people.

TTI also calculates the percent of peak period travel that is congested.   Austin again fares poorly -- its roads are congested for 70% of the peak travel period: 

By comparison, Dallas's roads are congested for only 66% of the peak travel period.  Houston is marginally worse at 73%.   Austin's peers include New York (69%), Detroit (71%) and Baltimore (69%).  

One has to wonder what damage Austin's congestion has wrought to downtown.   Businesses want to locate where their employees can get to them, and high peak-period congestion means that commuters must fight traffic to get to work most of the time.   I suspect Austin's downtown suffers more heavily from peak-period congestion than any other employment center in the area since MoPac and I-35 -- downtown's principal commuter routes  -- are the two most congested roads in the area.     Eventually we must recognize that our congestion will stunt downtown's growth, if it hasn't already.   We wil lose the economies of scale and agglomeration benefits that a vibrant downtown would  yield.

You know, of course, that there is only one solution.

Skilled cities II

While I'm in this nerdy mood, let me point to this paper by Jaison Abel, Ishita Dey and Todd Gabe:

We estimate a model of urban productivity in which the agglomeration effect of density is enhanced by a metropolitan area’s stock of human capital. Using new measures of output per worker for U.S. metropolitan areas along with two measures of density that account for different aspects of the spatial distribution of population, we find that a doubling of density increases productivity by 10 to 20 percent. Consistent with theories of learning and knowledge spillovers in cities, we demonstrate that the elasticity of average labor productivity with respect to density increases with human capital. Metropolitan areas with a human capital stock that is one standard deviation below the mean level realize around half of the average productivity gain, while doubling density in metropolitan areas with a human capital stock that is one standard deviation above the mean level yields productivity benefits that are about 1.5 times larger than average.

This is a little technical, so let me translate:   Cities tend to become more productive as they grow denser.   On average, a city's workforce becomes 10% more productive when the city doubles in density.   But that average obscures the importance of skills.   Less skilled cities benefit a lot less than skilled cities from densification.   In fact, skilled cities, on average, enjoy three times the productivity gain from denser growth than less skilled cities.   This is yet more evidence of the increasing returns and agglomeration benefits from density.

This is a nice complement to the Glaeser and Resseger paper.  Glaeser and Resseger found that workers in skilled cities become more productive as the city grows, while workers in unskilled cities do not.   Abel et al find that workers in skilled cities become more productive as the city grows denser; workers in unskilled cities, less so.

The authors also test their conclusions using a variant of weighted density. They find even greater productivity gains (20% on average) when a city doubles its weighted density.

The authors use a coarse form of weighted density.  They weight urbanized area density by county subdivisions.  But how you chop up a city matters when calculating weighted density.  In order to calculate weighted density, you first divide the city into a bunch of smaller regions.  You then assign each region's density a weight equal to its share of the populations.  In general, weighted density increases as you chop up the city into smaller regions.

I used census tracts for my weighted densities.  There are many more census tracts than county subdivisions.  I thus got a lot more stratification than they did -- e.g., their top weighted density was 19,000 ppsm, while mine was 33,000.  How you divide a city for calculating weighted density is somewhat arbitrary, but I think using census tracts makes more sense than county subdivisions, which are more or less arbitrary.  I suspect the authors would have found even greater returns to density had they weighted density by census tracts rather than by county subdivisions.  

H/t Richard Florida.

City growth

The Census Bureau has released its July 1, 2008 city population estimates.  Note that Texas has three of the largest eight (Houston, San Antonio and Dallas) and six of the largest 21 (add Austin, Fort Worth and El Paso).  Houston now trails Chicago by just 600,000, a gap that Houston could close within 20 years if the last decade's growth rates hold.  This is a surprising statistic as long as one ignores metropolitan populations, which are what really count when calculating a city's heft.  Chicago's metro population is still much larger than Dallas and Houston's, which are probably now the fourth and fifith largest metro areas.  And, in general, Texas cities have large boundaries and aggressively annex surrounding areas, which partly accounts for their rapid growth.

What's going on with Fort Worth? It had the fastest growth of the top 25, both between 2007 and 2008 (3.6%) and between 2000 and 2008 (29%).  No other city in the top 25 was close; Charlotte and Phoenix were the closest at 21% and 18%, respectively.  Fort Worth has grown twice as fast as Austin since 2000.  I haven't heard much about Fort Worth's explosive growth.  Strange.

One surprise on the other end:  Philadelphia contracted at a faster rate than Detroit between 2000 and 2008.

The most noteworthy fact here may be that, as the WSJ points out, central cities did very well between July 1, 2007 and July 1, 2008.  

The central-city population in U.S. metropolitan areas with more than one million people (excluding New Orleans) grew at an annual rate of 0.97% between July 2007 and July 2008, compared with a growth rate of 0.90% in 2006-2007, and growth rates around 0.5% in the years between 2002 and 2005.  By contrast, U.S. suburbs in metro areas greater than 1 million people grew at a 1.11% annual rate in 2007-2008, down from growth rates between 1.29% and 1.48% between 2002 and 2005.  Central cities are closing the gap.

The WSJ attributes the shift in growth rates to the recession.  We have a short memory.  2007 was the year of the historic spike in gas prices.  Central cities -- especially older, monocentric cities like Chicago -- began to look a lot more attractive to commuters.  I imagine this played a much bigger role than the nascent recession.   

Nothing to see here

The Census Bureau has released a new report (pdf) comparing the growth of Metropolitan Statistical Area central and outlying counties between 2000 and 2007.

Wendell Cox says the report shows "that the nation continues to suburbanize, despite the consistent media 'spin' that people are leaving the suburbs to move to core cities."

[T]he conclusion of the new report is clear. The nation’s most remote suburbs – its exurbs – are growing much faster than the central counties. Whether this trend will now reverse, of course, is up to debate. Perhaps demographic changes and higher energy costs will slow expansion on the outer fringes. More likely, the current recession may well lead to less exurban growth, but history suggests this may prove only a short-lived trend.

My own take is that this report offers little meaningful information other than growth and migration rates for the MSAs as a whole.  Breaking out the numbers for central counties and outlying counties tells us little because the Bureau defines "central counties" expansively as those containing all or a substantial portion of the MSA's urbanized area.   Since a city's urbanized area can and often does stretch across several counties, the Bureau's definition sweeps up counties that we typically consider suburban.  For example, Travis and Williamson are "central counties" in the Austin-Round Rock MSA, even though Williamson is the quintessential suburban county.   (Hays is considered an outlying county, but will almost certainly be a central county in the next census due to the rapid spread of Austin's urbanized area.)

Because of the broad definition of "central" counties, almost all MSA population is "concentrated" in these counties.  In MSAs with more than 5 million, for example, 97.4% of the population lived in central counties (2007 estimates).  For all MSAs, the number is 91.8%.

Naturally, because central counties began with a large baseline population, even large absolute increases in population yielded smaller percentage increases than small absolute increases in the less populous exurbs.   For example, in the Midwestern MSAs, the outlying counties grew at nearly twice the rate of the central counties.  But the central counties added 5 times as many people.   The New York-Northern New Jersey-Long Island MSA is an even more extreme example.  Its central counties grew by 2.6% between 2000 and 2007 while its outlying counties grew at an explosive 26.6% rate.   But the central counties added 480,000 people to just 12,000 for the outlying counties, 40 times as many.

For the United States as a whole, MSA central counties added nearly 17 million people compared to 2.4 million for outlying counites.  Central counties in the northeastern region (which includes New York) added nearly 1 million while the outlying counties added fewer than 24,000.  The South Region's outyling regions added the most by far -- nearly 1.8 million -- but even here the outlying growth was dwarfed by the 7.7 million added by the central counties.   

If the Bureau's goal was to tell us something important about the relative growth of exurbs and central cities, it failed.   There's just not much to the data because of the overly-expansive definition of central counties.   As Wendell points out, this report might very well understate suburban growth (in fact, this is almost certainly true), but this report doesn't shed much light on the question.  Wendell himself recognized this limitation on the data, but unreasonably, I think, attempted to draw an unwarranted conclusions from the report anyway.

International migration and city growth

Wendell Cox recently posted a piece on Newgeography slicing and dicing the Census Bureau's 2008 estimates of county populations and domestic migration.  He tells us that:

• people continue to move to the suburbs of large metropolitan areas, and away from core areas;
• in 2008, large metropolitan areas' core counties experienced a net domestic loss of 314,000, and an annual average loss of 580,000 between 2001 and 2008; 
• net domestic migration gains were down to 192,000 in suburban counties from an average gain of 246,000 over the decade.

The story, as Wendell sees it, is that slowing domestic migration to the suburbs does not signify a flight back to the cities.   Rather, it is a return to normal migration patterns -- spiking home prices between 2001 and 2007 created a domestic migration bubble as homeowners cashed out and moved to cheaper areas.

But there is another interesting phenomenon lurking in the the Census data.  Even after accounting for domestic migration, demand for central counties has risen.  And that rise often has been driven by international migration.

It's clear that demand for central counties has been rising if we look at changes in total population rather than just domestic migration outflows.  In 36 of the 48 large metropolitan areas analyzed by Wendell, the core counties added people between 2001 and 2008.  In four of the other twelve metropolitan areas, both core and suburban counties lost residents; it is hardly surprising that the core county of a declining metropolitan area will lose population.  Only eight metropolitan areas fit the "flight to the suburbs" model -- an increase in suburban population and a decrease in  core county population.

Some cities can thank domestic migration for the rise in their core-county populations.  Austin, for example.  It added 41,000 from domestic migration between 2001 and 2008, nearly 11,000 from 2007 to 2008 alone.

But the growth of other core counties has been driven by international immigration.  For example, Los Angeles County lost a staggering 1,007,000 residents due to domestic migration between 2001 and 2008.  Its population nevertheless grew by a total of 318,000.  It more than replaced its departing residents with healthy natural growth and international immigration.  The Census Bureau estimates that between July 1, 2007 and July 1, 2008, Los Angeles County lost 103,000 residents to domestic migration -- but added  87,000 from natural growth and 70,000 from international migration, a net population increase of 54,000.

The story is the same in booming Dallas.  Between July 1, 1007 and July 1, 2008, Dallas County lost 19,000 residents to domestic migration, but added 20,000 from international immigration.  Its international migrants replaced its departing residents one for one.    

One explanation of this phenomenon is that international immigrants are outbidding existing residents.  The supply of housing in central cities is relatively inelastic thanks to tighter restrictions on new construction, NIMBYism, and the generally higher cost of dense, infill construction.  A central city can add only so many people in a single year.  A large inflow inevitably triggers a rise in prices and a corresponding outflow.  

An alternate explanation is that demand is in fact dropping for core counties, causing a decline in housing costs, which in turn is attracting less affluent international migrants.   Home prices of course have dropped sharply nearly everywhere over the last year or two.  But that was not the case for the preceding seven years, and the international migrant inflows did not start last year.  Regardless, declining demand can't be squared with rising population.  Each year, more people, not fewer, want to live in these central counties.

I suspect that core counties are relatively more attractive for international immigrants than domestic residents.  Immigrants are attracted to clusters of immigrants of the same nationality; these clusters naturally develop in core counties first, and tend to be self-sustaining.  Perhaps immigrants are more comfortable with a denser, central-city environment.  And perhaps domestic residents have a better grasp of the opportunities that exist outside the core county.

Whatever the reason, international immigration is driving demand for many the  core counties in many metropolitan areas.  The recession will cause international immigration to plummet.  If international immigrants have been outbidding domestic residents, then domestic migration will drop sharply.  On the other hand, if domestic migration has been prompted by a distaste for central cities, the outflows will continue.  We'll see.

Demand for suburbs vs demand for central cities

Americans overwhelmingly prefer suburbs.   We know this because most Americans  choose to settle in suburbs.

No, wait.  Demand for suburbs and exurbs is collapsing; they are on their way to becoming the next slums.

Both memes are wrong, I think.  The "suburbs are dying" meme is wrong because most metropolitan growth will continue to occur in the suburbs -- at least in rapidly growing metropolitan areas -- regardless of shifts in preferences.  The first is wrong because rapid suburban growth doesn't tell us anything about the relative demand for suburbs and central cities.  

First, rapid growth in a metropolitan area is synonymous with rapid growth in the suburbs.  And it will always be that way.  (The form of suburban development is a separate issue.)  It is harder to add new housing in a built-out environment.  Denser construction is more expensive and takes a lot longer to get built.  

Take Austin.  Austin's urban core had about 170,000 residents in 2000.  Austin's metropolitan area added about 390,000 residents between 2000 and 2008.  Even if Austin's urban core had doubled in population between 2000 and 2008, most growth had to and did occur at the metropolitan area's fringes.  Of course, Austin's urban core did not double in population.  It is not possible to double the density of a built-out urban core in just eight years.  I'd be surprised if Austin's urban core grew by 20% between 2000 and 2008.  So its suburbs necessarily have boomed, even though Austin's core is healthy and experiencing steady growth.

Ditto with cities like Dallas, Houston and Atlanta, which each added one million residents between 2000 and 2008.  It would have been impossible for their central cities to add more than a small fraction of the newcomers.

But because most growth must occur in the suburbs, this growth does not alone tell us whether demand for the suburbs is growing faster than demand for the central city.  

We know that demand is higher for the suburbs when the central city is losing population while the suburbs are adding population.  But in thriving cities, both suburbs and central cities are adding population.  The central city is simply adding population at a slower rate.

To figure out relative demand, we also have to consider prices.  Because the housing supply in the central city is relatively inelastic, higher demand mainly shows up as higher prices.  Because the housing supply in the suburbs is relatively elastic, higher demand mainly shows up as more subdivisions.  To determine whether demand is rising more rapidly for one than the other, we have to weigh both the rate of new construction and the change in prices.  And, to complicate things, we have to factor in shocks like rising gas prices and recessions; these have different impacts on the relative demand.  

So determining whether demand for one is growing faster than demand for another means disentangling a complicated knot.  That is a hard thing to do.  A lot harder than focusing on just rising supply or rising prices.

U-haul stat o' the day

I don't normally do the Austin booster thing, but courtesy of Mark Schill at Newgeography:

When comparing California with Texas, U-Haul says it all. To rent a 26-foot truck oneway from San Francisco to Austin, the charge is $3,236, and yet the one-way charge for that same truck from Austin to San Francisco is just $399. Clearly what is happening is that far more people want to move from San Francisco to Austin than vice versa, so U-Haul has to pay its own employees to drive the empty trucks back from Texas.

Schill also lists some net domestic migration numbers.  In 2008, Austin-Round Rock had the highest net domestic migration rate among all U.S. metropolitan areas with more than 1.5 million people.  Austin-RR had 22 net domestic migrants per 1,000 residents.  Comps:  Houston, 6.6; Dallas, 7.0; Washington DC, -3.4; Portland, 8.3; Denver, 7.3; Detroit, -13.9.

People sometimes take net domestic migration flows as evidence that one place is better than another.  This is wrong.  Flows instead show which cities are overvalued and which are undervalued.  That's why I don't put any stock in "best cities" lists.  You can figure out the "best cities" (whether for quality of life or jobs) by looking at property values.  You can figure out the "best bargains" by looking at  domestic migration flows.

PS.  Sorry to pick on Detroit again -- no schadenfreude here -- but these numbers show just how grim things are there.  Despite a low median home sales price -- $91,000 in April 2008 -- the Detroit metro area experienced a net outflow of 13.9 residents per 1,000 residents in 2008.  The metro area was grossly overpriced despite being very cheap.  Well overpriced:  the median sales price had dropped to a little under $45,000 by last April.  (Things are much worse in Detroit proper.)  

Central-core job density and metropolitan growth

As I mentioned last time, I have a couple of problems with Elizabeth Kneebone's study of "job sprawl."  The main one is that there is less to the study than might appear at first glance.

But there is a very interesting phenomenon lurking in her data that deserves some discussion

We should expect large metropolitan areas to have more jobs in the outer rings than small metropolitan areas.  That's just a function of size.  While scanning her data, though, I noticed that large metropolitan areas invariably have more jobs in their central cores as well.  Because Kneebone uses the same land area for all central cores -- the area within a 3-mile radius of the CBD --  this means that large metropolitan areas have a higher job density in their central cores than small metropolitan areas.

I decided to test this a little more rigorously.  I regressed central-core job densities on the number of metropolitan area jobs within 35 miles of the CBD.  (Actually, I used logs, but I don't want to get bogged down in the technicalities.)   I expected there to be an association between the two, but I was surprised by the tightness of the fit:

The correlation is .89 and the R-squared is .79.  This means, roughly speaking, that 80% of the variation in central-core job density can be explained by the number of jobs in the metropolitan area.  Doubling the number of metropolitan area jobs tends to increase the central-core job density by about 68%. (The slope of the trend line is 0.75.)

Think about this a second. This data set includes small metropolitan areas and large metropolitan areas. Young cities and old cities. Sunbelt cities and Northeastern cities.  Western cities and Midwestern cities. Cities with manufacturing economies and cities with information economies. Transit-oriented cities and autocentric cities. Dense cities and sparse cities. And yet central-core job densities grow at roughly the same rate across all types of cities; city form or structure doesn't matter much.*

Why?  One explanation might be that as cities add jobs, they invariably need more lawyers, insurance companies, banks and other support jobs.  This is true regardless of the type of economy or the distribution of jobs within the metropolitan area.  Perhaps this "invariant" (too strong?) shows that these service jobs benefit from a central location regardless of how the city grows outside the central core.  In other words, this might be evidence that certain professions believe there are benefits to agglomeration.  Just a theory. 

Or perhaps I'm making too big a deal about this.  In any event, I think this is the interesting story and one I'd like to see Kneebone or some other smart person address.

*Detroit and New York City are outliers here.  Detroit is an outlier because its central city is emptying out at an alarming rate.  New York is always an outlier because of Manhattan's extreme density.  Los Angeles is somewhat of an outlier but I think that's a function of Kneebone's narrow definition of CBD's; Los Angeles has some fairly old, dense concentrations of jobs (Long Beach, for example), that she excludes from the central-city core.

Update:  Spreadsheet on Google Docs.

Austin's "job sprawl"

This paper on "job sprawl" from Brookings has been getting lots of buzz.  The author, Elizabeth Kneebone, used Census data to measure the shift of jobs from central cities to suburbs for the 98 largest metropolitan areas.

She concludes that employment steadily shifted to the suburbs between 1998 and 2006:

Employment steadily decentralized between 1998 and 2006: 95 out of 98 metro areas saw a decrease in the share of jobs located within three miles of downtown. The number of jobs in the top 98 metro areas increased overall during this time period, but the outer-most parts of these metro areas saw employment increase by 17 percent, compared to a gain of less than one percent in the urban core. Southern metro areas were particularly emblematic of the outward shift of job share with a 2.6 percentage-point decline in urban core job share and a 4.8 point gain in the outermost ring, outpacing the 98 metro average (a 2.1 point decline and a 2.6 point gain, respectively).

I've got a few problems were her analysis (and her term "job sprawl"), but I thought the Austin statistics were surprising. 

In 2006, the 45 largest metropolitan areas  -- those with more than 500,000 jobs -- had an average of 19.6% of their jobs within 3 miles of their CBDs.  Austin was well above average:  24.4% of its jobs lay within the "urban core."  By contrast, Dallas had 10.6% of its jobs within the urban core, Houston had 11.6% and San Antonio had 14.7%.  (Although the San Antonio metropolitan area has 14% more jobs than the Austin metropolitan area, Austin has 45% more jobs in its urban core.)

That isn't really the surprising part, though; Houston and Dallas, at least, are gigantic, sprawling metropolitan areas with several major employment centers each.  Austin has three major major employment centers (UT, the State Capitol Complex and downtown) in its core.

What was surprising was the change in urban-core employment.  Although the share of Austin's urban-core employment declined from 27.8% to 24.4% between 1998 and 2006, it actually added jobs to its urban core.  A lot of jobs.  Austin's urban-core added 16,400 jobs between 1998 and 2006, a 12.6% increase.  By contrast, Dallas lost 17,683 urban-core jobs (a 6.5% decrease); Houston's urban core lost 19,356 jobs (a 7.8% decrease); and San Antonio's urban core lost 2,655 jobs (a 2.6%) decrease.  Thus, of the four large Texas metropolitan areas, Austin was the only one to grow its urban-core job base.

I frequently use Columbus, Ohio as a benchmark for Austin.  They both have large universities in their urban cores, and both are state capitals.  But Columbus looks a lot more like Dallas or Houston than Austin:  Between 1998 and 2006, it lost 19,950 urban-core jobs, an 11.8% decrease.  (And its urban-core's share of all jobs dropped from 23.2% to 19.3%).  

This is surprising for a couple of reasons.  First, Austin's central city-to-suburb transportation network is primitive, bacically just hypercongested MoPac and I-35.  Houston and Dallas have much more extensive networks, and Dallas has a rail network.  Since transportation costs act like a tariff on central-city employment, I expected to see most growth take place in Austin's 'burbs.  Instead, it was the only one of the big four to add jobs to the central city.

Second, I'm surprised to see Austin add so many jobs to its urban core given its relative lack of density.  Both Houston and Dallas are denser than Austin using the weighted density metric.  While they both have several major job centers scattered throughout the metropolitan area, I did not expect the least dense city add the most central-city jobs.  

Simpson's paradox

Counterintuitive fact o' the day (and one I'm a little embarrassed I just learned).

Let's say you run an experiment on a sample population.  You do this by dividing the population into two subgroups and running the experiment on each separately.  The experiment succeeds for each subgroup.

If the experiment succeeded for each sub-sample, then it necessarily succeeded for the combined sample, right?

Nope.  Sometimes an experiment's apparent success on two samples disappears  when the samples are combined.  This is Simpson's paradox.

Example:  A pharmaceutical company wants to determine whether a new drug performs better than an old drug.  The manufacturer tests New Drug on 300 patients in Chicago and 300 patients in Cleveland.  In Chicago, 240 of the patients get New Drug and the other 60 get Old Drug.  Ninety of the former (37.5%) and 20 of the latter (33.3%) recover.  Thus, New Drug performs better than Old Drug in the Chicago trial.

In Cleveland, 60 of the patients get New Drug and 240 get Old Drug.  Thirty of the former (50%) and 110 of the latter (45.8%) recover.  Hence New Drug performs better than Old Drug in the Cleveland trial, too.

But New Drug's apparent superiority disappears once the samples are combined -- 120 of the 300 that took New Drug recovered (40%) while 130 of the 300 that took Old Drug recovered (43.3%).  Thus Old Drug did better than New Drug in the aggregate sample even though New Drug did better than Old Drug in each of the two sub-samples.*

Weird.

(*Example lifted from Simon & Blume, Mathematics for Economists)