Density: Too Dangerous to Mention

I didn’t respond to Randal O’Toole’s first take on The Gated City because, quite frankly, it didn’t appear that he’d read it. Now he insists that he has, and yet…well let’s just go through his points. First, on whether density facilitates productivity growth:

Avent cites research showing that denser American cities have higher labor productivity. For example, a 1996 study uses 1988 data to show a correlation between the urban densities and the wages and salaries earned in the various states such doubling job density raises incomes by 6 percent.

Six percent isn’t very much, and in order to reach even this conclusion, the authors of the study left out states that had lots of mining industries, which depend more on mineral density than population density. If they hadn’t left those states out, Alaska would have been the most-productive state. Even if some industries do benefit from density, not all do.

Avent cites several other studies, but does not mention this one, which compares density with technological innovation, a field that–unlike mining–strongly depends on the personal contacts that supposedly are gained from density. The study did find a correlation between density and innovation, but it also found an optimal density and an optimal city size. In other words, at some point, more density is no longer better. In particular, the optimal urban area size is about 750,000 people and the optimal density is about 2,000 jobs per square mile–roughly Baltimore or Philadelphia. Baltimore and Philadelphia are far from the nation’s densest urban areas, and in fact Houston is right between them.

O’Toole seems to accept the relationship; indeed, he directs us to a paper in which the authors conclude:

These findings confirm the widely held view that the nation’s densest locations play an important role in creating the flow of ideas that generate innovation and growth.

His principal criticism seems to be that in the paper in question, the authors estimate an optimal density which is not, in his view, that dense. That’s not much of a complaint, in my view. I’ve been careful not to specify an optimal density; in fact, I suspect that such a density is likely to vary significantly across industries, institutions, and with transportation and communication technologies. The empirical evidence supporting a positive relationship between density and productivity is strong. There is comparatively little work on optimal densities. At any rate, the policy implications don’t seem to tilt in O’Toole’s supposed ideological direction. I can’t imagine why a libertarian would want to go around telling willing city residents that they’re living at too high a density. And if productivity falls above certain densities, the market should quickly adjust; we’d see little wage growth in cities near the threshold, and consequently little growth in housing costs. Of course that’s not what we observe.

O’Toole goes on:

Another problem with Avent’s first claim is that most of the studies he cites compare the relationship between density and productivity at a single point in time. But the whole point of The Death of Distance is that this relationship is getting weaker all the time. This study, for example, found that the correlation between density and productivity declined from 41 percent in 1940 to 7 percent in 1980. It seems likely that, by the time any city can do anything about its density, the benefit of being denser will be negligible.

Now, this is really remarkable. I argue — I think quite clearly — that the relationship between density and productivity appears to be growing stronger, precisely because of technological developments. I cite research indicating that since the 1980s R&D spillovers appear to have become more important, that in recent decades the return to talent has increased in dense cities, and that in recent decades density explains a surprisingly large share of real wage growth. I point to Ed Glaeser’s work, which indicates that the “Death of Distance” phenomenon that gutted urban industry appears to have strengthened knowledge agglomerations and increased the return to ideas. Cato’s chief urban scholar seems to have no knowledge of these strands of the literature, nor does he appear to have spent all that much time with my book.

O’Toole then argues that I’m incorrect to blame NIMBYs for high housing costs. Instead, local zoning rules are to blame. Again, he seems strangely ignorant of the arguments I actually made. I write that residents of dense cities oppose new development and use a variety of mechanisms — tight zoning rules, historical preservation, and political pressure among them — to achieve their ends. I’m not sure how O’Toole thinks most city rules get on the books; presumably, local interests have had something to do with it.

Finally, there’s this:

Perhaps the greatest flaw is in the third step of Avent’s reasoning. “I don’t wish to tell anyone where to live, and I certainly don’t want the government having its way in the matter,” says Avent. But if the NIMBYs and their urban-growth boundaries and large-lot zoning disappeared, these regions would all become less dense, not more dense. Housing costs would become more reasonable, and the areas might actually become more productive, but not because of density.

The trouble is that, although Avent doesn’t support coercive government policies, many other density advocates do. And Avent specifically cites transit-oriented developments as a way of increasing density without realizing the huge amount of coercion that is required for many of those developments, including prescriptive zoning, tax subsidies, and artificial restraints on low-density development so that some people will have little choice but to live in the higher densities.

Frankly, I have no idea what kind of economic model he’s using here or what evidence he has in mind. I’m not sure what would happen if zoning rules were relaxed across the board. Home price data suggest that there would be a lot of new construction in dense, high-cost cities. Given the high cost of land, it seems likely that densities would rise. Based on the research I’ve examined, and that which O’Toole himself cites, it seems likely that this growth would result in more people living in dense areas and earning higher wages, thanks to rising labor productivity. Now, maybe O’Toole is arguing that cities would like to grow outward but are prevented in doing so by NIMBY rules, such that relaxed zoning would lead to sharply reduced densities. Again, there is scant evidence for this assertion, but I’m happy to admit that it’s a possibility. Either way, I can’t see why O’Toole wouldn’t be in favor of making it easier to build new housing — of whatever sort the market wants — in cities, especially those where home prices and productivity levels are high.

I discuss several different ways to make it easier for cities to satisfy demand for housing, and transit-oriented development is one of them. It seems clear that government intervention would be required to support such a policy, and O’Toole seems very unhappy about this. Whether he’s just as unhappy about the taxes, infrastructure spending, and zoning rules that typically accompany suburban growth isn’t clear. But in general, I’m interested in changes that increase, on net, a landowner’s ability to build as he wishes. It seems clear to me that TOD is a useful way to get local governments to accept upzoning, and landowners seem more than happy to build as much as the government will let them. Libertarians might not approve, but it seems to be a more liberal approach than the status quo.

I do appreciate O’Toole’s engagement with the book, such as it is. I sincerely hope that libertarians will consider the arguments I make and warm to the theme.

Housing Is About Supply and Demand

Rob Pitingolo’s response to The Gated City has now been posted at Greater Greater Washington under the headline, “Housing is more than supply and demand”. I actually think this headline is a little unfair to Rob, who is not in fact arguing that housing is more than supply and demand. Rather, he’s arguing that housing is not a commodity. That is, homes aren’t all perfect substitutes for each other:

There are many unique types and styles of housing, some of which are more desirable than others. When demand for housing rises in a neighborhood, rents will rise, regardless of the type or quality of the housing. A neighborhood might have century-old rowhouses, 1970s apartments, and brand new luxury buildings. If demand is rising in that neighborhood, rents for all types of these units will rise.

But what if a neighborhood doesn’t have any vacant land sitting around waiting to be developed? How do you increase the supply of housing when there’s no place to build new housing? Basically, you have to knock something down and replace it with higher density housing.

Let’s imagine that a developer is proposing to level some not-so-great ’60s-style townhouses in an urban neighborhood. In their place, the developer is going to build a multi-level apartment complex with a gym, pool on the roof, and ground-floor retail. Perhaps the developer is going to knock down 10 low-quality units and replace them with 50 high-quality units, for a net-gain of 40 housing units.

Even though the number of housing units in the neighborhood goes up, it’s virtually guaranteed that the market rents for those new units are going to be higher than the rents for the old units. So the folks who might have been able to afford one of the ’60s-style townhouses no longer can afford a luxury-apartment in the neighborhood.

This is a good point. Supply matters, but not all housing types are perfect substitutes. If you tear down cheap, low-quality housing to build lots of high-quality housing, then the cost of housing across the metro area as a whole may fall (or grow more slowly) along with the cost of high-quality housing. But you’ve reduced the supply of low-quality housing. Residents displaced from cheap housing who can’t afford the lower cost of high-quality housing may find themselves in a pickle. The solution, however, is not to restrict development (and to be clear, Rob isn’t necessarily suggesting that it is).

There are two key things to remember. The first is that the more development a metro area approves, the less development pressure there will be on any given piece of land. If most of the city is zoned to accommodate tall buildings, then a lot of new demand for the city will flow to those new buildings, placing less cost pressure on the land atop which sit low-quality units. Consider Brookland, in Washington. The land around the Metro station is mostly empty, and local residents are fighting “dense” developments on that land, by which we mean buildings up to 6 stories tall. But residence in the District on metro-accessible land is in very high demand. The less of that demand is accommodated by new, high-quality buildings on empty land, the more will be shifted toward the older homes in the neighborhoods around the station.

The second thing to remember is that when I say I want looser housing regulations, I mean it. I’m bothered by rules that prevent owners from subdividing existing houses — an important means through which affordable housing can be provided. I wish it were easier to convert outbuildings, basements, and similar structures into housing. I wish people weren’t anxious to keep urban industrial land zoned for industrial uses in the misguided view that industry will be coming back to central cities. I’m not just interested in making the supply of high-quality housing more flexible. I want that flexibility to extend across housing types.

Rob is right; lower-income residents might well oppose new development out of a fear that it will mean replacing old, affordable homes with new, unaffordable ones. Those residents are mostly the victim of efforts to make it hard to build everywhere. And if those concerned about the poor succeed in blocking redevelopment of older homes, they’re not stopping displacement — oh, no. They’re simply shifting it elsewhere, to places where the poor aren’t quite as good at finding advocates for their interests.

On the Great Relocation

Noah Smith writes a post that touches on some themes in The Gated City, particularly those relating to the export advantages of healthy cities. I comment on his post here, but I wanted to add to the argument with a passage from the book. At Free Exchange, I write:

In other sectors, the gains to firm concentration remain. We observe concentration in these sectors where there are…concentrations of economic activity, as you might expect. If we look at the tradable sectors in American cities, we see things like finance, management, technical consulting, high-tech research and manufacturing, information, systems design, and so on—that is, human-capital intensive industries on the innovative frontier. Collections of smart people are good at taking new ideas, turning them into workable business models, and marketing those businesses around the world.

I elaborate on the point in Chapter 3:

Within this space, where the ideas are “in the air”, there is a greater level of “technological congruence”. Think of it like this: if you’ve been following and participating in a conversation between friends, then you’re likely to find a particularly witty crack funny — you were there for the build-up. If you later report the punchline to another buddy, he probably won’t get it. He had to be there.

Innovation is like that. You often can’t drop a new technology into a random situation and expect local firms and workers to exploit the technology to its fullest. They haven’t been participating in the conversation that led up to innovation, that often spelled out the principles on which it relies and the problems it’s meant to solve. Even if they can get it to work, they’ll frequently be at a loss if it breaks down, and there’s almost no chance that they can continue to refine and improve the innovation.

Cities aren’t simply a collection of raw human talent. Instead, they’re the processing power embodied in the conversation that talent conducts. Innovations and technologies that perform well in one processor might not be compatible with others. And large cities act as potent idea processors.

This processing role that cities play — supporting a conversation that drives innovation — applies almost exclusively to fledgling technologies. In the early years of a new technology, the conversation is of critical importance. Firms are attempting to understand its capabilities, to make it work better, and to reorganize modes of production and business models to optimize the new, big idea. At this stage, awareness of the context of the technological discussion is crucial. Among the firms and individuals within the innovative city circulates a large body of knowledge of what has worked and failed so far, and what is likely to bear fruit in the future. Transplant that technology to a tabula rasa at this stage, and it will be nearly useless.

As the technology matures, the importance of the technological community wanes. Much of the tacit knowledge is written down, codified, and taught in schools or training programs. Processes become standardized and easier to explain to newcomers. Consequently, it becomes easier to shift use of the technology abroad. This is a key way in which the city fuels growth across the economy, by domesticating new technologies and making them accessible enough for deployment outside the initial all-important technological community. But it’s also a process that diminishes the importance of the city, at least insofar as a particular technology is concerned.

Eventually, new ideas are tapped out. Every permutation is explored and every efficiency realized. The technology can be used anywhere; it’s simply a matter of plugging in the machines and passing out the manuals. At this point, a technology is entirely offshorable. There is no longer any need to pay the premium to be close to the center of discovery and development. This moment can be a destabilizing and painful transition point for the innovative city. But so long as it hasn’t lost its innovative capacity during the process of wringing ever greater efficiencies out of the initial technology, it should go on to tinker with and develop new and different ideas.

The difficulty of achieving technological congruence in places apart from innovating cities is one of the great challenges of developmental economics. Most of the technology the world has is fairly simple to use and reproduce (if not for you and I then for trained engineers), and yet there are massive wealth disparities across the world. Some of the gap is due to institutional differences and variation in education. But history is also littered with examples of economies that tried and failed to import and develop around existing technologies. Local populations unfamiliar with the deep (and often tacit) knowledge underlying the original process of innovation struggle to successfully deploy even basic technologies…

Economic growth is not a zero-sum game. That is not to say that Americans never face costs related to growth abroad; they certainly do. It is to say that growth abroad is clearly good for the welfare of the world as a whole and is generally good for the welfare of Americans, particularly when American government policy is set appropriately.

To the extent that Americans are interested in exporting goods or services or ideas rather than jobs, it’s worth reflecting on the role of the productive city. This processing machine, including the human capital of its residents and the tacit knowledge of the ongoing innovative discussion they share, is not easily replicable abroad. Believe me, firms would love to be able to do just that; moving away from the high land and labor costs of big American cities is very good for profits. But they often cannot afford to. The value of the urban agglomeration is simply too great.

China can’t easily reproduce our innovative cities. Right now, unfortunately, America seems more interested in chasing smokestacks than in strengthening its existing advantages.

On Attitudes

Kevin Drum pens a few thoughts on The Gated City here, and I’d like to respond to one of them. He writes:

It’s also unfortunate, I think, that his solution to restrictive building policies sounds an awful lot like the libertarian dogma that any kind of property restriction is a “taking” that should be fully compensated by the government. That’s pretty tough to swallow. There’s more to life than simply letting developers build anything and everything that pops into their minds.

I’d say my prescriptions go well beyond that, but I’ll let readers judge for themselves. Generally speaking, I’d say that this is a pretty good encapsulation of the problem we face. Using zoning rules and other measures to heavily restrict what can be done with a piece of private property does substantially reduce the value of that piece of private property; that’s not dogma, it’s common sense. Maybe Kevin thinks that the government is right to do this, and that the benefits of preventing private property owners from satisfying market demand are much greater than the costs imposed by the government regulation. It seems clear to me that this is often not the case, and Kevin doesn’t present any evidence to the contrary beyond an aversion to the idea that people might want to invest money in building things on their own land. How gauche!

We need to wrestle with the trade-offs here, not indulge our inclination to think of developers as nasty, uncouth brutes. Maybe developers will sometimes build things we don’t like. If that opens high-wage cities to millions of middle-class workers, and brings substantial environmental benefits in the process, isn’t that maybe, just maybe a worthwhile exchange?

During the industrial revolution, technological shifts led to a dramatic change in the optimal city size. Key industrial cities grew by upwards of 800% in some cases. This massive growth wasn’t always pleasant; in many cases it led to dreadful poverty and disease epidemics. But what would the world be like today if governments had prevented New York City, London, Chicago, Detroit and so on from growing beyond the tiny burgs they were in the early 19th century? The transformative effect of the computing and IT revolution on cities might not be that dramatic. It is nonetheless likely to be important, and the incredibly high cost of living in high productivity cities should strike us as a very worrisome sign. We’re failing to get people to places they’d very much like to be, and it’s reasonable to conclude that we’re doing substantial economic damage in the process.

But hey, there’s more to life than letting people build anything they want on their own land, right? That’s easy for those of us with the means to live in rich places to say.

The City and Growth: Papers of the Day

Since I published the book, there have been questions about the depth of the literature on the relationship between cities, density, and growth. It’s pretty deep! And to illustrate, I’ll resume posting interesting research results when I find them. Like this one:

This paper reassesses the relationship between density and productivity by using detailed geo-coded data on wages and employment in Sweden. The contribution is empirical and builds on an analysis of spatial units of exactly the same size in terms of geographic surface. The data divide Sweden into areas of one square kilometer, and describe each area in terms of wages, worker characteristics and industry structure. Since the geographic areas are of constant size, the sheer number of employees in an area is an ‘exact’ measure of employment density. We find a significant relationship between density and productivity across squares. The estimated elasticity is in the 9-10 percent interval, and is insensitive to whether we employ a standard OLS estimator or a panel fixed effects estimator. Neighbor characteristics also matter in ways consistent with the idea of positive agglomeration externalities, where a location in a square with high density neighbors reflects proximity to (or a location in) a larger agglomeration. Moreover, the estimated relationship between productivity and density is insensitive to the chosen spatial scale.

And this one:

Schumpeter claimed the entrepreneur to be instrumental for creative destruction and industrial dynamics. Entrepreneurial entry serves to transform and revitalize industries, thereby enhancing their competiveness. This paper investigates if entry of new firms influences productivity amongst incumbent firms, and the extent to which altered productivity can be attributed sector and time specific effects. Implementing a unique dataset we estimate a firm-level production function in which the productivity of incumbent firms is modeled as a function of firm attributes and regional entrepreneurship activity. The analysis finds support for positive productivity effects of entrepreneurship on incumbent firms, albeit the effect varies over time, what we refer to as a delayed entry effect. An immediate negative influence on productivity is followed by a positive effect several years after the initial entry. Moreover, the productivity of incumbent firms in services sectors appears to be more responsive to regional entrepreneurship, as compared to the productivity of manufacturing firms.

Within regions, entrepreneurship ultimately boosts the productivity of incumbent firms.

The Gated Valley

It seems clear to me that current weakness across the American economy is primarily a demand-side phenomenon; output is falling below potential because consumers and businesses are reluctant to spend and invest, and a demand-side government boost would make a big difference.

At the same time, I think there are significant structural problems in the economy that acted as a drag on growth prior to the recession and which constrain recovery now, not the least of which are the ones I describe in my book. New data on metropolitan GDP help make my case. In 2010, the San Jose metropolitan area enjoyed real GDP growth of 13.4% — well above the 3.0% growth rate for the economy as a whole and easily the best performance of any large metro. This was quality growth, too. The information sector contributed 3.62 percentage points to growth, professional and business services added 1.72 percentage points, and much of the rest is almost certainly attributable to high-tech manufacturing (based on the share of growth in manufacturing and the historically high share of high-tech, computer and electrical equipment manufacturing in San Jose GDP).

That 13.4% rate is eye-popping but not entirely shocking. Tech firms have done very well lately, including business service firms, retail-oriented businesses like Apple, and the rapidly growing social-networking sector. And among fast-growing portions of the economy, these are the businesses that contribute a large share of value added and which are a key component of American exports. The health care and education sector growth we see elsewhere doesn’t necessarily score well on those measures.

But there’s a problem: for all that growth, the San Jose area added just 16,000 private-sector jobs in 2010. Real output expanded by 13.4%, and private employment rose by just 2.1%. By comparison, the Houston metro area experienced comparable employment growth at 2.0%, despite real GDP growth in 2010 of just 1.6%. Why isn’t San Jose’s blockbuster growth translating into lots of new jobs?

The city of San Jose’s housing statistics report gives us a pretty good idea what the problem is. A San Jose household hoping to limit fair market rent to 30% of income must earn $5,673 per month. The owner-occupied housing market in San Jose remains dysfunctional, not least because of the difficulty of getting a loan. The rental market gives us a clear sense of what housing conditions are like for typical families, however. The rental vacancy rate in San Jose is about 4% — strikingly low. And average rents have been rising since 2009 and are up 11% year-over year.

Reuters has been kind enough to publish an excerpt from my book, and conveniently enough, it’s from a chapter analyzing Silicon Valley entrepreneurship rates during the Dot Com boom. The pattern is clear: when new technologies drive a Silicon Valley boom, wages and housing costs rise while entrepreneurship and employment lag. The San Jose metro area’s inability to allow enough new housing to accommodate housing demand translates rising incomes into rising rents rather than rising employment. This, in turn, prevents the American economy from taking full advantage of Silicon Valley’s economic potential.

TGC links

Some interesting comments on the book:

Alex Block

David Levinson

Josh Rothman, at the Boston Globe’s Braniac blog.

Reihan Salam discusses Yonah Freemark‘s response to the book.

And Wendell Cox responds to my response. I think the literature is more sophisticated on this subject than he lets on (see this summary, for instance). A county-level study is as close to a fine-grained, weighted-average employment density analysis as you can get given available data. Fine-grained population densities are a reasonable proxy for employment density. And don’t let anyone tell you San Jose isn’t dense.

I do appreciate hearing what people think. Let me know if I’ve missed your comments on the book.