Co$t$ of $prawl?

Is sprawl costly and smart growth cheaper? That would depend on what you mean by sprawl, smart growth, and costly.

If sprawl is defined as low density or fragmented development, smart growth is the opposite of sprawl, and costly is average per-capita public expenditures aggregated to the county level, then Carruthers and Ulfarsson report absolutely in their unpublished paper: “Does Smart Growth Matter to Public Finance? Evidence from the United States.” (2008 update: This paper has since been published in Urban Studies 45 (9), August 2008, pp. 1791-1823.)

A new costs-of-sprawl kind of study with the still alluring marquee “smart growth” thrown in for good measure, and a fine chance for an idle mind with a costless web site to weigh in on a sometimes messy research/practice literature.

From the abstract,

This paper addresses four fundamental questions about the relationship between “smart growth,” a fiscally motivated anti-sprawl policy movement, and public finance. 

1. Are low-density, spatially expansive development patterns more expensive to support?
2. If so, how large of an influence does sprawl actually have?
3. How does the influence differ among service types?
4. And, how does it compare to the influence of other relevant factors?

The analysis, which is based on the entire continental United States and uses a series of spatial econometric models to evaluate one aggregate (total direct) and nine disaggregate (education, fire protection, housing and community development, libraries, parks and recreation, police protection, roadways, sewerage, and solid waste disposal) measures of spending, provides the most detailed evidence to date of how the built environment affects the vast sum of revenue that local governments spend every year.

Estimates derived from the empirical models indicate that the value—increased cost—of a one standard deviation decrease in density nationwide would be $7.99 billion annually and that the value of a one standard deviation increase in the spatial extent of developed land would be $20.25 billion annually. These findings are unambiguous evidence that smart growth matters to public finance.

And being both weak and unsporting by nature, I am drawn to statements such as “unambiguous evidence” like a moth to a flame. Or, better for today’s lesson, like a sinner to a suburb.

Why bother?
Before troubling to read past the abstract, what am I expecting? Such studies tend to feature two kinds of results:

One, low density development leads to higher direct service costs because, well, it’s the density, stupid. Longer roads, more extensive sewage networks, and so on. Stands to reason. Very much an engineering exercise. They then offer illustrative calculations that, owing to their preciseness, inadvertently suggest utility and credibility (see Shoup’s masterful take on the illusion of precision): Lowering density x units per acre raises average water pipe costs y$ (presumably for any given performance standard), and so on (as in Table 1 here, calculated by the critical Cox and Utt from the benchmark Burchell, et al.). You can plug these right into a memo to council. And the primary source is the National Academy of Sciences!

Two, to press the larger point home further still, sprawl leads to higher indirect costs owing to various induced behaviors, such as more driving all around, or higher public health costs from less walking around. That is an economics exercise, making the fiscal impact calculations trickier and less consistent.

So we seem to have some answers in user-friendly form. Any need for more work on either? Yes, please.

Here’s the thing: As back-of-the-envelope funxercizes (a word I just made up and claim credit here for all time), these estimates are either useful or useless, with the inconvenient caveat that the dissimilarity is rarely apparent. Though often pitched as well nigh definitive:

• Even the simplest country reckoning suggests the need to account for a tradeoff between the higher and lower costs of density. Lower densities imply a lengthier network, but denser networks imply higher facility/planning costs of their own. The calculation of a higher (or lower) infrastructure cost in any one case, or worse in any average across places or times, does not separate these tradeoffs out, let alone acknowledge them. (Start with a series of papers by Ladd and her numerous respondents.)

• Add to that the problem of conflating expense with performance. Suburbs may well have higher income residents who in turn demand, and receive, higher levels of service. (For the difficulties and necessity of separating out the cost from the demand dimensions of local demographics alone, start with Schwab and Zampelli, 1987.)

• The related “impact fee” aka “pay as you go” literature has also identified a number of thorny obstacles associated with attributing specific cost shares to new growth in general, as in the discussion of how to portion out social overhead in chapter 6 of 1993′s excellent Regulation for Revenue, by Altshuler and Gómez-Ibáñez and subsequent studies of fiscal impacts in practice (such as this new draft by Kurt Paulsen).

So my prior is that using simple comparisons of per-capita budget outlays to claim that “sprawl is more costly to service” is a rather rough, and quite possibly wrong, way to go. They might or might not be close enough for government work, or not even wrong.

Does Smart Growth Matter to Public Finance?

What about highly evolved efforts? Which brings us back to the state of the art in the form of the Carruthers and Ulfarsson paper:

1st, and I realize I should have spit this out before I read the abstract, I have an issue with the title. The answer to the posed question is either yes, which we thought we knew, or no, which would be quite a stunner — in which case you would want to be less suspenseful and merely say, “Smart growth doesn’t matter for public finance!”

Yet, reading on, the abstract says the answer is yes, leaving us back in the “we knew that” category. Of course, the abstract also promises, “the most detailed evidence to date” without hinting why the details matter. This is partly a matter of personal taste, but since this site is all about my personal taste you should know that it runs more to the “How and why does smart growth matter to public finance?” or “Three shocking ways in which SG matters to PF” or “Amazing details about SG public finances” sort of teaser. These are just off the top of my head of course.

(How about, “Does Public Finance Matter for Smart Growth?” A different question altogether, I realize, but someone should write it anyway. For emphasis, I’ll conclude on that point.)

2nd, the literature review is rather good, meaning it’s both credible and purposeful. It’s up to date and organized in (a) a way that makes sense and (b) tells a story that more research of a certain kind would be valuable. Namely, their kind.

I like sensible stories, even those I don’t buy, and theirs is that sprawl results from market failures which are difficult to solve for various mundane reasons. This develops into a characterization of the policy regimes in use and at hand, where the authors’ own work is prominent. (I’ve heard this kind of lit review called “self-serving,” which is confusing. If you won’t go out on a limb and contend that your own work matters, perhaps greatly, then why should I?)

3rd, they plan to explain and thus estimate direct and indirect costs together. They also intend to separately examine two land development measures: Density and proportion of land that is developed.

4th, the authors construct a statistical model with a good pedigree, and apply this to … county-wide data. Well, that raises a flag. I’ve used county-wide data when the county is too big for the question at hand but at least I apologize, in the first draft anyway. Counties may be fine for some areas, less so for the west I think. Still, the fair standard is whether this moves our understanding forward rather than backward, not whether the data are ideal.

5th, results. Without going into the details of the specification or estimation strategies, which are not blog-friendly, the paper has a subsequent “policy evaluation” section set up in Q&A form. Here are (slighted edited-down) excerpts:

Are low-density, spatially expansive development patterns more expensive to support? The estimation results show that density has a negative effect on four key measures of local government spending: total direct, education, roadways, and sewerage. In sum, the results … provide unambiguous evidence that low-density, spatially expansive development patterns are more expensive to support than high-density, compact development patterns.

How large of an influence does sprawl actually have? These calculations indicate that the value—increased cost—of a one standard deviation decrease in density nationwide would be $7.99 billion annually and that the value of a one standard deviation increase in the spatial extent of developed land would be $20.25 billion annually.

How does the influence differ among service types? The elasticities reported for the individual expenditures … show that the specific magnitude of sprawl’s influence depends on the service in question. Here, sprawl has the largest influence on relatively specialized services that may have to be replicated when they otherwise would not, a more moderate influence on linear infrastructure systems that connect to centralized facilities, and the smallest influence on facilities/services that receive heavy day-to-day use.

How does the influence of sprawl compare to the influence of other relevant factors? [C]ompared to other relevant factors, the influence of sprawl is substantial.

[Overall, w]hile there is considerable variation in how density and the spatial extent of development influence different types of services, other things being equal, sprawl nearly always raises expenditures and the effects translate into very large dollar values when summed across the entire country. These findings strongly suggest that the reasoning behind fiscally motivated, anti-sprawl smart growth policy frameworks is sound.

Summary

This is very useful work. It addresses an increasingly topical policy issue and takes these data and methods further. It can’t help but serve as a handy reference for what comes next.

On the other hand, there is at least one important limitation to the “sprawl is really costly” conclusion that I wish was stated more plainly in the introduction and conclusion. Moreover, this work stimulates other questions about the public finances of sprawl and smart growth. In many respects, these are addressed to the literature as a whole more than this paper.

1. Is Demand adequately explained?
As mentioned up top, how does a focus on the per-capita service bill map to a world of Tieboutian migrants, purposefully shopping or voting for their preferring mix of cost, service, and housing circumstances?

The paper at hand does this via several demand controls, such as population, housing values (a measure of wealth, and possibly of benefits, though they list this as a built environment measure), income, age and race, along with some tax-price variables (tax revenue and source, intergovernmental revenue). (Not to mention their capture of marginal rather than average prices is dicey; hey, I just remembered I wrote a paper on this once.)

Which is about all you can do with these data. Still, the method applies averages drawn from internally diverse counties. I would prefer a disclosure that demand may remain incompletely explained and thus an unobserved factor in these results — especially since any remaining bias is probably in the direction of making high-demand, low-density suburbs appear to have higher costs when they may only have higher service demands.

Put another way, in practice a denser general plan may well attract a different pattern of per-capita service demands, on top of any production cost differentials. This may lower service costs in itself. This paper attempts to abstract from that but we can’t say how successfully due to data limitations.

2. Questions on the Public Finances of Sprawl

• I could question whether smart growth is the opposite of sprawl, or something else again.

• Can we say, generally, how far off are back-of-the-envelope calculations for different services?

• Cities are densifying in the west, where land is relatively valuable and populations climbing; is the opposite is true in declining metropolitan areas? In either case, are low average service costs explained more by growth or the pattern of growth?

• This paper measures at the county level. I wonder about counties that are densifying on average, where this masks urban areas that are losing population to their in-county suburbs. How would this affect the interpretation of the statistical results?

• More to come….