Predicting the Future is Hard

About two years ago, the Urban Land Institute published Growing Cooler: The Evidence on Urban Development and Climate Change, which argued that it will be crucial to build cities in a more compact fashion if the country hopes to avoid substantial growth in vehicle miles traveled and carbon emissions over the next few decades.

At the time Sarah Goodyear summarized some of the findings:

The report cites real estate projections showing that two-thirds of development expected to be on the ground in 2050 is not yet built, meaning that the potential for change is profound. The authors calculate that shifting 60 percent of new growth to compact patterns would save 85 million metric tons of CO2 annually by 2030. The savings over that period equate to a 28 percent increase in federal vehicle efficiency standards by 2020 (to 32 mpg), comparable to proposals now being debated in Congress...

2883223507_b86ffc3f60_m.jpgLess of this, please... (Photo: SkilliShots/Flickr)
In other words, better development choices can and should be a key part of efforts to meet emission goals.

Then about a month ago, the National Academies' Transportation Research Board released its own report on the effects of compact development. In many ways it supported the conclusions of the Growing Cooler report. Increased density can produce significant reductions in VMT. And there's this:

The TRB report suggests that if 75 percent of this new construction is of a more compact variety, that emissions could be reduced 10 percent or more from the baseline scenario (and that is not taking into consideration the deployment of cleaner electricity generation and other potential sources of savings).

Of course, as some commenters noted at the time, the TRB report also quoted figures for a future, a "moderate" scenario, in which a far smaller share of new development was built compactly. On the whole, the authors were a bit more conservative in their view of the effects of increased density, and a good bit more conservative in their assessment of how much density might actually be increased.

The difference in conclusions is largely about the assumptions used to build models. Yesterday, Reid Ewing, Arthur C. Nelson, and Keith Bartholomew offered some comments on the TRB report's findings, focusing on ways in which their assumptions differ.

For instance, with respect to the 75% conclusion quoted above, they write:

Their “moderate” scenario assumes that 25 percent of residential development between now and 2050 will be compact, defined as twice the density of trend development.  Their “upper-bound” scenario assumes that 75 percent of residential development will be compact.  We, on the other hand, assume that between 60 and 90 percent of all new development through 2050 will be compact...

The NRC committee’s “moderate” assumption translates into as much as 80 percent of the built environment continuing to be sprawled, despite the forces described above moving us toward more compact development.  For instance, between 2010 and 2050, more single-person households will be added than households with children. Moreover, roughly two-thirds to three-quarters of the net gain in households between 2010 and 2050 will be among households without children. Housing demand functions of households without children and single-person households are different from households with children.

In other words, much of the growth in low-density suburban development was driven by an increase in the share of households with children (and with multiple children). From the 1970s to the 1990s, the massive baby boom generation was busy having and raising kids, and they overwhelmingly opted to do their child-rearing in suburbs.

But that same generation is now making the transition from suburban parents to empty-nesters to retirees. At the same time, their children are not getting married and having kids at nearly the same rate as they did.

Now, the decline in the share of households with children will eventually be offset by increases in the overall population, and so demand for traditional suburban development will hold steady and eventually increase. But since so much of the current built environment is low density, these demographic shifts mean that most of the new housing built in the coming half century will need to be more compact to satisfy household demand.

That is, household types that tend to prefer more compact development constitute a growing share of a growing population in a world where existing housing supply is not oriented toward such demand. It would be very strange indeed if only 25% of new development were of a more compact sort.

So why did the TRB report conclude, unrealistically, that levels of compact development might be so low? Well, because the process of modeling the future is hard, and scientists tend to be pretty conservative about it. In particular, they're often reluctant or unable to incorporate into their models things that will probably happen, but which are "exogenous" in nature -- that is, outside the scope of the model.

Take the above question. Just how compact new development can be will depend in part on how successful cities are at changing land-use rules that support low-density, car-oriented development. Given increasing demand pressure for more compact housing in walkable areas (and the developer money to be made building it) it stands to reason that local governments will increasingly change their zoning rules to accommodate denser building.

But there is basically no way to incorporate this assumption into a model. It involves the potential for action among thousands of disparate local government organizations, the sitting members of which have yet to be chosen.

A cautious modeler will be reluctant to go out on a limb on this point, and his results will be conservative as a result. That's part of the way reports like this are written, but it doesn't mean that we, as readers, have to dumbly accept that what the report's authors were forced to assume is what the real world will actually be like.

When talking about development forms, there are many such assumptions biasing results toward conservatism.

For instance, it has been very easy in recent decades for developers of suburban tract housing or strip malls to obtain financing for their projects. So common were such development types that Wall Street firms developed standardized financing terms for builders, which served to reduce financing costs.

Denser projects or in-fill projects, by contrast, were less common and more complex. As a result, financing for them required a certain amount of customization. Not only did this frequently increase borrowing costs, it also led many finance firms to abandon the business for less cumbersome opportunities.

Growth in the share of development that is of a compact form will increase the incentive to develop cheaper and more standardized financing options for compact projects, thereby making the construction of additional compact developments easier and more affordable.

But this process is essentially unmodelable. Things will likely work this way, but careful researchers simply can't guess how financial products will evolve in future decades and plug that guess into their models.

The moral of the story is this: whether the topic is the effect of climate change regulations, or high-speed rail system construction, or compact development, experts will produce analyses that try to predict likely outcomes and which will be widely circulated and quoted by policymakers and journalists.

These reports are not the final word. They have to be read critically, with an understanding of the limitations faced by their authors. If you don't get the constraints researchers face, then you can't understand their results.