Federal, state, and local governments invest more than $50 billion annually in public school construction and repairs. Despite large school infrastructure spending, resources are allocated unevenly among the almost 100,000 public elementary and secondary schools in the United States, and many are in need of renovation, expansion, and repair. For example, a third of public schools rely on portable or temporary classrooms (Figure 1) and a quarter report that environmental factors, such as overcrowded classrooms, poor ventilation and indoor air quality, temperature control, inadequate computer hardware and wiring, and broken windows and plumbing, are "moderate" or "major" obstacles to instruction.

Figure 1 · A dilapidated temporary classroom

A full evaluation of school investment decisions must capture all potential positive impacts, and at the same time account for the costs of investment. Those costs include the direct taxation necessary to fund school investments, and other potential hidden costs. For example, the local public jurisdiction may be inefficient or corrupt, resulting in a dollar of taxation translating into less than a dollar of investment value. If the local government overhead is large enough, such investments could be better provided by the private sector. Moreover, out-of-control bureaucratic overhead levels are unproductive.

Instead of evaluating each of the potentially positive and negative outcomes that come from funding and providing school infrastructure investments, we rely on parents' location decisions to identify their revealed preferences about spending levels. In essence, we assume that any shift in the desirability of a district—along either academic or non-academic dimensions—will be reflected in housing prices.

Residential values, and land values more generally, are a function of community amenities. Schools may be the most important amenity, as parents want to live in areas with good schools. A school that receives new investments is likely to attract high income and high education residents, and that may set in motion a process in which the composition of the households in the neighborhood also changes, giving rise to substantially larger price and stratification effects. Such movements benefit real estate owners and developers.

Whether in practice public investments in schools increase real estate values is an empirical question. Below we examine the California school finance system, and how bond referenda can be used to estimate parents' willingness to pay for school investments. We present the data used in our analysis, our results, and what they imply in terms of over- or under-investment in public school infrastructure. We conclude by addressing the policy implications of our findings, and what these implications mean to local developers and real estate owners.

California School Finance

California was known in the immediate post-war era for its high-spending, high-quality school system. By the 1980s and 1990s, however, California schools were widely considered to be mediocre and underfunded. In 1995, per-pupil current spending was 13 percent below the national average, and the state ranked 35th in the country despite its relatively high costs. The state was particularly tight with capital spending, which was 30 percent below the national average. California schools became notorious for their overcrowding, poor physical conditions, and heavy reliance on temporary, modular classrooms.

Much of the decline in school funding has been attributed to California's shift to a centralized system of finance under the 1971 Serrano v. Priest decision, and to the passage of Proposition 13 in 1978. In the regime that resulted, the property tax rate was fixed at 1 percent and the state distributed additional revenues using a highly egalitarian formula. Districts were afforded no flexibility and there was little provision for capital investment. In 1984, voters approved Proposition 46, which allowed school districts to issue general obligation bonds to finance capital projects, as long as these issues are approved by a local referendum. Initially, a two-thirds vote was required, but beginning in 2001 proposals that adhere to certain restrictions qualify for a reduced threshold of 55 percent. Authorized bonds are paid off over twenty or thirty years through an increment to the local property tax rate, typically 0.25 basis points. Under Proposition 13 assessed home values are based on the purchase price rather than the current market value. As property values in California have risen substantially in recent decades, recent homebuyers bear a disproportionate share of the burden, while long-time homeowners carry a low tax share.

Districts must specify in advance how the bond revenues will be spent. The ballot summary for a representative proposal reads: Shall _________ Unified School District repair, upgrade and equip all local schools, improve student safety conditions, upgrade electrical wiring for technology, install fire safety, energy efficient heating/cooling systems, emergency lighting, fire doors, replace outdated plumbing/sewer systems, repair leaky rundown roofs/bathrooms, decaying walls, drainage systems, repair, construct, acquire, equip classrooms, libraries, science labs, sites and facilities, by issuing $85,000,000 of bonds at legal rates, requiring annual audits, citizen oversight, and no money for administrators' salaries? Anecdotally, such bonds are frequently used to build new permanent classrooms replacing temporary buildings (Figure 2), although repair, maintenance, and modernization are also common. Overall, 629 of the 1,035 school districts in California voted on at least one bond measure between 1987 and 2006, and the average bond value was $6,300 per pupil.

Figure 2 · New modular classrooms, Watkinson School, San Francisco Bay Area, California

How to Value Capital Projects

Bond-funded investments are accompanied by an increased tax burden with an approximately equal present value. Thus, if funds are misspent or simply yield smaller benefits than the consumption foregone due to increased taxes, bond authorization will make a district less attractive, leading to reduced pre-tax housing prices. By contrast, if the effect on school output is valued more than the foregone consumption, home prices will rise when bonds are passed. In aggregate, it can be shown that finding positive effects on prices after a bond passage indicate that previous levels of school investments were inefficiently low.

We use this bond setting to estimate the impact of school investments on house prices. Although school districts that issue bonds are likely to differ in many ways from those that do not, some of these differences can be observed and controlled for, such as the average income of residents or the racial composition of the district. However, other differences, such as the residents' demand for education, cannot be easily observed or measured. To mitigate this problem, we focus on close elections: a district where a proposed bond passes by 1 percent of the vote, for example, is likely to be similar, on average, to one where the proposal fails by a similarly narrow margin. Thus, comparing districts that passed and failed a bond by small margins can identify the average impact of bond funding on house prices and other school district outcomes.

A final issue concerns timing. Capital projects take time to plan, initiate, and carry out, so bonds issued today will take several years to translate into improved capital services. Direct measures of school outputs will reflect the effects of bond passage only with long lags. House prices reflect the present discounted value of all future services less all future taxes, so house prices should rise or fall as soon as the outcome of the election is known. This may happen well before the election if the outcome is easy to predict, but when the election is close there is likely important information revealed on Election Day. Price effects may therefore be immediate. However, if house prices are sticky, or homebuyers have imperfect information, it may take several years for prices to fully reflect the impact of bond passage. We are thus interested in measuring the full sequence of price changes in the first few years after a bond has passed.

Data & Results

Our data on bonds come from a database maintained by the California Education Data Partnership. For each proposed bond, the data includes the amount, intended purpose, vote share, required vote share for passage, and voter turnout. Our sample includes all general obligation bond measures sponsored by school districts between 1987 and 2006. We merge these to annual district-level enrollment and financial data from the Common Core of Data (CCD). The house price data comes from a proprietary database compiled from public records by the real estate services firm DataQuick, and describes all housing transactions in California from 1988 to 2005. We use GIS mapping software to assign those housing transactions to school districts. We also test whether test scores were improved due to capital investments. That data comes from the California Department of Education. We find that passage of a bond measure causes house prices in the district to rise by about 3 percent in the first year, and then it gradually increases to 6 percent two or three years following the election and persists for at least a decade. Hence present benefits exceed the net present value of future tax burdens.

We then ask whether these increases in housing prices reflect achievement gains of students. However, we find that the effect of bond passage on student achievement is extremely imprecise and our estimates provide only ambiguous evidence for positive effects at long lags. Even our largest estimates for the achievement effects are too small to fully explain the impact of bond authorization on housing prices. Evidently, prices reflect dimensions of school output that are not reflected in student test scores, such as enhanced safety or aesthetics.

As discussed above, a substantial effect of bond passage on prices indicates that the marginal resident's willingness-to-pay (WTP) for school services exceeds the cost of providing those services and therefore that school capital spending is inefficiently low. It is thus instructive to compute the WTP implied by our estimated price effects. This calculation requires assumptions about interest and discount rates, the speed with which new facilities are brought into service, property tax shares, and the income tax deductibility of property taxes and mortgage interest payments. We outline our baseline calculations here.

Assume that the average house in districts with close elections (margins of victory or defeat of less than 2 percent) is worth $236,433, so a 3 percent effect on house prices raises the value of the average house by approximately $7,100. The average bond proposal in close elections is about $6,300 per pupil, and there are 2.4 owner-equivalent housing units per pupil. With a typical municipal bond interest rate of 4.6 percent, this implies a property tax increment of $163 per house per year, for a present discounted value of $1,950. Thus, the effect of passing a bond on the total cost of owning a home in the district (combining the house price effect with the present discounted values of future taxes) is approximately $9,050. That homebuyers are willing to pay this amount implies that their WTP for $1 in per pupil spending is $1.44 ($9,050/$6,300). When we account for the deductibility of mortgage interest and property taxes and for the higher tax share borne by new homebuyers, the WTP estimate can be as low as $1.13, but never as low as $1.

The WTP is generally higher when we measure the price effects several years after the election. The WTP based on the price effect in year four, for example, ranges from $1.31 to $1.89. The sensitivity of WTP calculations to the year in which price effects are measured may indicate that capitalization of school investments into housing prices is not immediate.

Conclusion

Infrastructure investment is an important component of government budgets, yet we have few tools to assess its effectiveness. We provide a new design to conduct such an evaluation. We further identify the effects of capital investments on housing prices by comparing districts in which school bond referenda passed or failed by narrow margins. Unlike districts where bond referenda garnered overwhelming voter support or opposition, the set of districts with close votes are likely to be similar to each other in characteristics both observable and unobservable by the researcher. Most important, this methodology allows us to control for these unobservable differences.

We find effects of 6 percent or more on house prices, and implied valuations of $1.50 or more for every $1 in school capital spending. As theory predicts, most of the price effect appears well in advance of the completion of the funded projects. The home price effects presumably reflect the anticipation of increased school output, though it appears that much of the effect derives from dimensions of output, such as safety or aesthetics, that are not captured by test scores.

Those results are important for local real estate owners and developers who are uncertain about the economic benefits of such public investments. They should capitalize those positive externalities generated by school investments. Moreover, it also points to the relevant social function performed by local developers who execute the infrastructure work.

Finally, our results demonstrate that California districts at the margin of passing a bond are spending well below the economically efficient level, with returns to additional spending far in excess of the cost. Those districts are spending less than the desired amount in part because they were constrained by flawed state policies that restricted the ability of districts to raise more revenues. Moreover, the referendum process erects too large a barrier to the issuance of bonds and prevents many worthwhile projects. Loosening California's constraints on local spending would yield substantial economic benefits. It also suggests that recent efforts to permanently constrain school expenditures in states such as Pennsylvania and New Jersey may backfire in the long run, by reducing local and social welfare.