School Accountability Ratings and Housing Values

DURING THE PAST DECADE, states have constructed elaborate systems for rating the performance of individual schools based on student test scores and then have released this information in the form of school "report cards." The federal No Child Left Behind Act of 2001 will accelerate that movement by requiring states to test all students in grades three through eight, publicly report each school's student test performance, and sanction schools when they fail to achieve specific standards. Earlier research has documented the cross-sectional relationship between housing values and student test scores at neighborhood schools. ¹ Given the magnitude of the relationship between test scores and housing prices in the cross section, one might expect the release of school report cards to have important effects on housing values and, indirectly, to provide incentives for schools to improve performance. However, there are also reasons to believe that the housing market would downplay the information in school report cards. In particular, school test scores are noisy measures of school performance and may provide homeowners with little new information about which schools are the best ones. ²

In this paper, we explore how school test scores, changes in school test scores, and categorical ratings are related to housing values. To control for any correlation between scores and neighborhood amenities, we use the approach of Sandra E. Blackand focus on homes near elementary school attendance boundaries. ³ The geographical detail in our data allow us to precisely determine the location of each home being sold and every school and school boundary and to focus our analysis on homes within a few thousand feet (or less) of school boundaries. Thus our empirical strategy is to compare sales prices for homes located in the same neighborhood and taxing municipality but that are assigned to different elementary schools. We control for a range of detailed observable characteristics of the house such as distance to school, lot and home size, and middle and high school assignments.

Our empirical analysis uses data from the housing market in Mecklenburg County, North Carolina, where various indicators of school quality were released by the state between 1993 and 2001 and reported each fall in the local paper. Between 1993 and 1996, average test scores were reported by grade level for the school overall and two racial categories (African Americans and whites/others). Beginning in August 1997, the state began explicitly placing schools in performance categories (ranging from "low performing" to "schools of excellence"), using a combination of the proportion of students that achieved proficiency on the test and a value-added measure (based on average improvement in each student's score from the prior year). With these data, we are able to explore how housing values are related to the variation in performance across schools, the year-to-year variation in performance for a given school, and the change to categorical performance measures in 1997.

We begin by estimating the relationship between long-run measures of school test performance (scores averaged over many years) and housing values using the regression discontinuity design proposed by Black—comparing sales prices of homes near elementary school boundaries. We find a significant positive relationship between test performance and housing values, somewhat larger than that found by Black, and which is robust across a variety of specifications. ⁴ Our estimates suggest that a one student-level standard deviation difference in mean school test score is associated with an 18 to 25 percentage point difference in house value, controlling for neighborhood amenities and housing characteristics. A student-level standard deviation in test scores is quite large relative to the between-school differences. The implied impact of a school-level standard deviation is smaller, approximately 4 to 5 percentage points. Nevertheless, if year-to-year changes in school performance had effects of this magnitude, we would expect large swings in...