Ryan Avent in the NY Times: One Path to Better Jobs – More Density in Cities (His article is apparently based on his new Kindle Single Gated City, which I have not yet read). I have several comments.
1. I like density. This is a personal preference. Being mostly (small “l”) libertarian, I feel no need to impose this on others. If lots of people like density, they should create dense places, or hire developers to do this for them. If they are more productive in dense places and earn more money, or are more consumptive and get more value for money, they are even more incented to do so. I am all for fewer regulations so long as negative externalities associated with density (congestion, pollution, parking spillover, etc.) are properly internalized. The lack of good property rights, road pricing, parking pricing, pollution pricing, etc., however, has led communities to develop a regulatory rather than pricing approach. See: Zoning and Externalities.
2. Population density of the earth, the United States, Minnesota, etc. increases over time if population increases and land area does not.
2a. Density of suburban areas is increasing over time as they are transformed from farms to not-farms.
2b. Local density may decrease as cities of fixed boundaries depopulate. Part of this problem is the fixity of boundaries, part is consumer’s desires as shaped by markets and policies to consume areas with locally lower density.
3. I think the most important part of this argument is backwards. The primary causality is that cities (metropolitan areas) with growth create density as people and firms are attracted to faster growing areas (NIMBYism is certainly a cost, but there is plenty of housing in the Bay Area that goes relatively undemanded, think of Oakland, much of which begs for gentrification. Clearly housing is not so dear as to cause people to do *that*.). Take away the growth and leave the built infrastructure and even the people and you have a decaying city which slowly (or quickly) depopulates.
4. Accessibility vs. Density. Density of itself does not create growth (despite economic correlations suggesting otherwise – correlation is not causation) or lead to more productivity without a useful transportation system. This is why fast growing areas are auto-oriented in many places, cars are faster than transit under many land use pattern/road network configurations and connect to more places in less time.
Accessibility is the product of both density and speed, how much one can reach in a given time. Certainly density helps increase accessibility to a point (where the benefits of more people are outweighed by the congestion/pollution costs, etc.), but so does a faster transportation network. At the point where connecting by personal transport fails, mass transport may be appropriate, and is certainly necessary for very large conurbations like New York. The pedestrian city maxes out at about 50,000 (see City of London c. 1800 – 1850).

5. Density is quite low in many places he describes (Silicon Valley, rather than San Francisco, is the job engine of the Bay Area, Oakland has plenty of density compared with say Cupertino or Mountain View). But there is still plenty of accessibility by auto (if not by walking, biking, or transit).
6. Some industries (finance, government, media) value the connectivity provided by accessibility more than others (agriculture, large scale manufacturing), and so are more likely to cluster together. It turns out, these are also presently the faster growing industries in the economy. Unfortunately the rules that help finance be more productive (concentrate everything in Manhattan or London) does not necessarily generalize to other industries or other cities. Those cities just don’t get as many financial headquarters and offices, and don’t see the growth in those sectors, and don’t need the accessibility as much (the premium they are willing to pay for accessibility), and aren’t as dense.
Similarly, within metro areas, the old, dense, downtown retain the regional financial, legal, and government headquarters while the new suburban areas that don’t value the propinquity get less, since the density is costly and the benefits for those in certain sectors is small.
7. There is also the inter-firm vs. intra-firm spillover argument. Many large firms, like universities and hospitals, build campuses to capture internally the spillovers from the random contacts that accessibility brings. This might lead to less vibrancy and turmoil in the market sector, as there are fewer independent actors competing, but it is also a natural stage of development from birth to maturity in sectors. To obtain economies of scale, firms consolidate, and fewer and fewer firms capture more and more market share. Ultimately gales of creative destruction come and the stagnancy of the few firms leads to their own demise, but that is the natural order of things. If firms could never capture their downstream profits by growing large, no one would ever invest. After all, the first few years are much less profitable than those after the monopoly/oligopoly has been established. (And capitalism is not about free markets and free trade, it is about profit (not that there is necessarily anything wrong with that, but let’s be honest)).
There are thus natural cycles. Firm formation is naturally higher in some sectors, in some periods, and in some places than others. But the successes of dynamism lead to consolidation. Increasing density exogenously (or removing some regulatory constraints in a few places where they are binding) has very little to do with this.