Richard Florida has a good, short post up arguing against what we might deem "blind density." In other words, in an effort to chase after density, we're simply building taller. Not more compact. And certainly not more efficient.
The diminishing returns comes from a few places. First, walkability and modal share of alternative transportation begins to jump around 20 units per acre. People are closer to the things they need and places they need to go. Other forms of transportation besides the car not only make sense but are more effective forms of transportation within dense places. Dense places will invariably have congestion. As Mumford said, "if you're trying to move 100,000 people around in a square mile area, the motor car is the worst possible solution." These gains in other, more efficient forms of transportation start to gradually decline from 40 to 60 units per acre and then plateau. There are no more gains to be made in terms of walkability and transit ridership over a certain density "saturation" point.
Another issue is that adding height often diminishes the quality and character of a place. Not everybody wants to live or work in a high-rise. By adding density only via height, you're effectively adding supply while diminishing your market, aka demand. Furthermore, because you're decreasing your market with that height, that means those who won't live in the area will have to commute in from further out areas increasing the overall vehicle miles travelled and overall load on infrastructure. Simply adding height rather than smarter, compact density is another way of supply-side urbanism. If you're selling the view, only a few people can afford views and when other buildings sprout up your view is of some dude in his underwear dancing to the oldies. Again, narrowing your target market.
A couple of tweets I, uh, tweeted a few months ago:
Barcelona = 41,000 ppl per sq mi. NYC = 27,000 ppl per sq mi. Lesson: it doesnt take skyscrapers to have high quality density.
If we want to cherrypick Manhattan: 71,000/sq mi. L'Eixample in Barcelona (well to do central district): 92,000/sq mi.
It should be noted that L'Eixample is the nicest, most desirable part of Barcelona. Barely a building over ten-stories as well. Density (should) = desirability. While I'm not totally against height or tall buildings (I live on the 19th floor currently), I am very wary of a rush towards adding height that might diminish the overall character of the place that makes it so desirable in the first place. Think DC or Paris with their strict height limitations. Are they too strict? Perhaps. But wouldn't you be skeptical of proposals from sociopaths like Le Corbusier want to destroy it and rebuilt in their singular vision?
Also, low-scaled, clustered, but high density buildings are the best way to protect buildings and the outdoor public spaces (and their microclimates) from harsh weather and climates, be they hot and sunny or cold and windy. Think about the way Penguins cluster. In hot climates, the close spacing of buildings decreases the amount of sun exposure to both buildings and public spaces. The ample square footages of rooftop surfaces serves as heat sink (or reflector) rather than walls, windows, and outdoor spaces.
Lastly, besides the role of density in transportation choice and reduced infrastructural load, the goal of density (mostly to economists) is to accelerate the internal combustion engines of cities, efficient and accelerated exchange of goods, services, and ideas within proximity. However, stretching buildings upwards has the same effect as stretching them outwards. I often lament living on the 19th floor (moving in 10 days - will post more on that later). I often walk to work. But I still experience rush hour: waiting for the elevators before and after typical work hours (often as much as 10 minutes if a few of the elevators are down, which invariably some always are).
As I said, in ten days I'm moving to another building. I'll be living on the 2nd floor right above Main Street. I'll be able to pop down the stairs in no time flat and get to that infamous interchange zone, the public street, far more efficiently (and desirably) than from 250 feet above the street.
So if the goal is proximity with density, the measure of proximity shouldn't simply be in the sprawling x- and y-axes, but also be measured upwards, in the z- or 3rd-dimension. Smart Density should be the goal. Not dumb (UP!) density.