I recently read Daniel Treisman’s brilliant book, The Architecture of Government: Rethinking Political Decentralization. This book is particularly important for classical liberals who defend decentralization as an important institutional reform for promoting and protecting individual freedom. Treisman’s thesis is essentially that decentralization is overrated. He doesn’t argue that decentralization generally has bad consequences, even under readily identifiable circumstances, but that the consequences of decentralization are so unpredictable and case-specific that few generalizations, even highly conditional ones, can be made about them. The book is largely theoretical, and Treisman takes on standard justifications of decentralization like Tiebout sorting, the role of mobile capital in keeping government small, and keeping government “close to the people.” While Treisman’s counterarguments to decentralization’s defenders are well thought out and in many cases persuasive, I remain more optimistic about our ability to make valid generalizations about decentralization. Still, any defender of “competitive federalism” or more local governance will need to grapple with Treisman’s challenges. I’ll take some of the most important of these challenges in turn.
One common argument for decentralization comes from Charles Tiebout: competition among local governments providing public goods allows residents to reveal their true preferences for these goods and incentivizes local governments to act on those preferences. Treisman argues that key assumptions of the model are so thoroughly violated in reality that the predictions of the model are not likely to hold true in the real world.
First, he argues that if “public service differentials are capitalized into property prices, then pressure on governments may disappear completely” (79). Residents then won’t leave districts that provide poor public services, and local officials will not be disciplined. Indeed, economists often use “hedonic pricing models” to test how local amenities affect real estate values, under the assumption that these amenities do get capitalized into prices. I believe Bryan Caplan, whom he cites, was the first to draw the conclusion that Treisman also does. Treisman concedes that if homeowners control the local government, then there will still be discipline. But then, he says, homeowners could just discipline bad public service provision anyway, making Tiebout competition redundant. And homeowners do not always control local governments. (And there are sometimes other bad consequences when they do.)
You might think it is unrealistic to assume that home values would adjust immediately to a change in policies. But it seems they would. If your local government increases your taxes or starts providing less service, you could try to sell out and move to another jurisdiction. But if you did, you would have to accept a lower sale price. Either way, you will be punished. Caplan says that it’s local governments’ reliance on property taxes that creates this result, but it seems to me that it is fully general, insofar as it applies to homeowners. Even if it’s an income tax hike, you can’t escape it by selling out — the buyer will demand a reduction in the price of the house equivalent to the (amortized) new income tax liability to which the buyer will be subject (assuming perfect mobility among jurisdictions). Variation across earners in their losses due to the income tax could certainly cause some sorting, but not a net migration loss, on average. It could cause a decline in the tax base, however, as especially high earners leave. But you could design a regressive income tax to hit low earners harder (in theory).
What if there is asymmetric information? Existing owners in a jurisdiction might know more about policies than potential buyers, at least in the short run. That way, they can benefit by selling out and moving. Real estate markets would eventually adjust with informed buyers leading the way, but there would still be a net migration outflow. Normally, asymmetric info makes markets work less well, but here it could make the interjurisdictional public services market work better.
What about those who do not hold wealth in real estate? Here is where the form of tax seems to matter. A property tax is not paid directly by renters, but will show up in rents as the supply of rental housing adjusts. In the long run, a property tax hike has no net effect on rents (decreases property values, increases wedge between purchase and rental values). So renters don’t care about that kind of bad behavior. A drop in the quality of public service has exactly parallel effects if services are funded through property taxes. But an income tax hike would hit renters directly, as well as owners. Property values would fall by the amount of the income tax hike, but because it would take some time for the rental market to equilibrate, some renters would leave.
Treisman’s other principal objections to Tiebout sorting are that 1) there are likely to be interjurisdictional externalities, and 2) central government policies have variable effects across jurisdictions, inhibiting or undermining competition effects. On the first point, as Coase reminded us, the real issue is transaction costs. Are transaction costs high to interjurisdictional bargaining? Sometimes yes, but often no — at least, unless upper-level governments impose transaction costs. It should be easy for local governments to associate together to form special-purpose unions for regulating externalities. On the second point, Treisman certainly does have a point, but even if Tiebout competition doesn’t give us perfect demand revelation under such circumstances, it can still be an improvement over standardized services. Whether an area is advantaged or disadvantaged by the central government, it may still have incentives to try to attract residents.
Still, when all is said and done, the evidence for Tiebout sorting in the U.S. today is not great (but see this interesting recent book from William Fischel (PDF)). The “capitalization” problem seems to me to be the most compelling argument against Tiebout sorting. How well does it apply to the “mobile capital” argument of Weingast & others?
Barry Weingast has argued that federalism is “market-preserving”: that competition among local governments for mobile capital restrains potentially predatory local officials from creating and seizing rents. If we assume that capital owners own no property, then it does indeed seem possible to avoid predatory behavior by moving capital from one jurisdiction to another. But Treisman has three main objections to this conclusion.
First, as in his AER piece with Hongbin Cai, Treisman shows that if capital is far more productive in one jurisdiction than another, capital will always flow to the more productive jurisdiction, no matter what the tax rate is. The less productive jurisdiction will not even bother to compete.
The result holds given the assumptions, but I find the assumption of just two jurisdictions with different production functions implausible. The greater the number of jurisdictions, the more likely it is that there will be more than one jurisdiction with a similar production function. If those jurisdictions are near the productivity frontier, they will compete for capital against each other. Granted, it will still be the case that the least productive jurisdictions do not bother competing for capital: any capital that gets created in these places immediately flows to the most productive places. This provides a rationale for countries that are far from trade routes and have very low labor and land productivity to clamp down on capital flows. That is the only way these countries will not stay impoverished. (Or they could get rich countries to liberalize immigration. Somehow.) But the equilibrium tax rate under decentralization will still be lower than under centralization.
The second of Treisman’s objections is that under decentralization, the tax base can be a kind of “commons” that both central and local governments will “overgraze.” Complete centralization can solve the commons problem, resulting in lower taxes and rents, even if the central government is completely predatory.
But if central government policy responsibilities are kept extremely limited, the potential for central governments to overgraze under decentralization will also be limited. In the pre-1914 U.S., the central government accounted for less than 40% of all revenues, and local governments made up by far the most of the rest. So central and state governments had limited opportunity to overgraze. Today, of course, with almost two-thirds of tax revenues being raised by the federal government, the potential for overgrazing seems more significant.
Treisman then recounts several other objections to mobile capital competition familiar from the public finance literature: underprovision of public goods for consumption, imposing negative externalities on neighbors, “exporting” taxes to residents of other jurisdictions, resources wasted in trying to attract businesses, & the “state-corroding federalism” problem. My own familiarity with the U.S. data suggests that these issues, even combined, are quantitatively minor. For instance, on the tax exporting issue, you can compare the Tax Foundation tax burden data for states (which “rewards” states for exporting taxes) to the Mercatus data (which does not, except for severance taxes). The differences are small for the vast majority of states.
Finally, Treisman argues that if Tiebout sorting and mobile competition for capital are so great, a centralized, unitary state could implement them, simply by ordering its local functionaries to try different policies in different places. This argument seems too cute to me. Centralized governments have never done this (“because it doesn’t work in the first place!” Treisman might say), but the most plausible reason is that politicians are often wary of delegating power. Would such a scheme of administrative deconcentration and policy experimentation be credible? What if local functionaries start doing things central government politicians don’t like? The advantage of federalism is that it is institutionalized. Local officials can do their own thing even if the center doesn’t like it. More on this issue below.
In summary, when there are many jurisdictions that are not “too heterogeneous,” decentralization is very significant, mobility is guaranteed, and transaction costs for interjurisdictional agreements are low, the familiar competitive dynamic by which taxes on capital are kept efficiently low seems to hold firm.
Due to space, I will not go very deeply into Treisman’s other arguments against the purported benefits of decentralization, but I will address some of the most important ones briefly.
First, Treisman says that the “vertical fiscal imbalance” problem, whereby local governments deriving some or most of their revenues from grants become less efficient and overspend, is overrated. If local governments have to fund spending out of own revenues, they will become more fiscally responsible, he concedes, but we also have to look at the central government. If the central government loses some of its tax authority to local governments, it will become less fiscally responsible. There is a sort of “law of conservation of fiscal responsibility” for Treisman.
But the relationship between the proportion of expenditures funded out of own-source taxation and fiscal responsibility is non-monotonic. There is no added benefit to fiscal responsibility from raising more than 100% of direct expenditures from own-source taxation. So — relative to a situation in which there are between-level transfers — a system in which each level of government pays its own way should indeed have greater fiscal responsibility, as the evidence has long suggested. But granted: this is not an argument for decentralization per se.
Second, Treisman argues that decentralization doesn’t make government “closer to the people.” He points to anecdotes of intimidation at Vermont town meetings and to corrupt local governments like that of Chicago. People might learn political vices, not virtues, from participation in government. When it comes to listening to voters, representatives in a centralized system can do just as well, so long as electoral districts are small. (But how small can electoral districts be in a large country? I wonder.) I basically accept his claims here. There is no evidence that small countries are “better” or “more democratic” or “more efficient” than large countries, for instance. (Contra Scott Sumner.)
Third, Treisman argues that decentralization doesn’t protect liberty. Decentralization can make it easy for local majorities to oppress local minorities. While that’s true, it’s not necessarily true. In the U.S., for instance, a robust 14th-Amendment jurisprudence (largely) prevents state and local governments from oppressing local minorities. Decentralization can take place along different dimensions. You don’t have to decentralize human rights policies in order to decentralize economic policies. In fact, I have found that across the world, regional autonomy is statistically associated with erosion of governmental discrimination against local majorities (that are also state-level minorities) and local minorities! So the worry that, in general, decentralization will cause more discrimination against local minorities seems misplaced.
Fourth, Treisman argues that decentralization doesn’t reduce secessionism. He’s right about this, although I also don’t buy arguments that decentralization promotes secessionism. The relationships appear to be highly context-dependent, and my own view of the matter is unfortunately too complex to summarize briefly in this space.
In conclusion, Treisman’s book presents a robust and welcome challenge to fans of decentralization. I came away unpersuaded that we can make no general statements about the likely consequences of particular kinds of decentralization, and on balance I still favor particular kinds of decentralization tailored to particular situations, but I did come away with a radically different view on Tiebout sorting, for instance. I am much more pessimistic about decentralization than I was before I read this book.
Still, one robust empirical relationship we see with political decentralization/federalism is that larger countries, both in terms of population and in terms of area, have more. It’s hard to explain this if Treisman is right that unitary governments have the right incentives and opportunities to capture all the advantages of political decentralization with mere administrative deconcentration. In fact, principal-agent relationships within the central government are important, and un-modeled by Treisman. In large countries, the central government simply can’t do as good a job with delegation, in addition to the “institutionalization” issue discussed above.