«Does the Market Value Value-Added? Evidence from Housing Prices After a Public Release of School and Teacher Value-Added Scott A. Imberman∗ Michigan ...»
Does the Market Value Value-Added?
Evidence from Housing Prices After a Public
Release of School and Teacher Value-Added
Scott A. Imberman∗
Michigan State University and NBER
Michael F. Lovenheim
Cornell University and NBER
Value-added data are an increasingly common evaluation tool for schools and teachers.
Many school districts have adopted these methods and released the results publicly. In
this paper, we study the release of value-added data in Los Angeles by the Los Angeles Times newspaper to identify how measured value-added is capitalized into housing prices.
This analysis is the ﬁrst in the school valuation literature to examine property value responses to a value-added information shock, which is of interest as this measure is less correlated with demographics than typical school quality measures. Unique to this setting as well is the release of both school and teacher-level value-added data, which allows us to examine how property values respond to both types of information. Using a diﬀerence-indiﬀerences methodology surrounding the release, we ﬁnd that neither school nor teacher value-added scores are capitalized into home prices. Our results suggest that, despite the contentiousness following these data releases, homeowners do not consider value-added models as currently constructed to be a relevant school quality measure on the margin.
∗ We would like to thank seminar participants at the AEA Annual Meetings, the APPAM Fall Meetings, CES-Ifo, Georgetown and the University of Michigan along with Julie Cullen, Susanna Loeb, Stephen Machin, Steven Rivkin, Guido Schwerdt, Gary Solon, and Kevin Stange for helpful comments and suggestions. We would also like to thank the Los Angeles County Tax Assessor’s Oﬃce, the Los Angeles Times, the Los Angeles Uniﬁed School District, and Richard Buddin for providing us the data and for their assistance with the data. Finally we’d like to thank Margaret O’Rourke and Michael Naretta for excellent research assistance on this project.
⃝2013 by Scott Imberman and Michael Lovenheim. All errors and omissions are our own.
c 1 Introduction The push to expand test-based accountability in US K-12 education has led to a signiﬁcant rise in the amount of information available to the public about school quality. Such information typically consists of school-average test scores, which reﬂect many factors aside from the ability of schools to produce test score gains. Furthermore, information on individual teachers’ performance largely has been kept private. As a result, there recently has been a growing interest in providing results of teacher and school “value-added” assessments to the public. A number of school districts, such as Los Angeles, Houston, and New York City, have released such information, either voluntarily or by court order. Ideally, such measures isolate the teacher and school’s contributions to a student’s achievement by removing the inﬂuences of other observed and unobserved factors. This information thus provides new data to parents and homeowners on the local schools’ and teachers’ contribution to test score growth. However, to the extent that there is noise and bias embedded in value-added measures, release of this information has the potential to distort parental decisions about where to live as well as school staﬃng decisions. Furthermore, because value-added information typically comes from a complex statistical model, it is unclear how it is interpreted by parents. With more and more school districts and states providing value-added data to parents and local communities, understanding the extent to which these communities value this information is of primary policy importance. It also is important for school administrators and policy makers to understand the impact of value-added on land values as capitalization of value-added could aﬀect the tax-base for school districts.
In this paper, we examine whether value-added information that is released to the public is capitalized into home prices. We make two contributions to the literature. First, to our knowledge, there has been no work done examining how parents value the value-added data generated by schools.1 Due to the increasing prevalence of value-added information and the likelihood that such information will continue to be released to local communities, estimating the extent to which value-added aﬀects property values is of importance in its own right.
The second contribution of this analysis is to the school valuation literature. A large set of Jacob and Lefgren (2007) ﬁnd that parents do value teachers who raise achievement.
prior work examines how average test score diﬀerences across schools are capitalized into home prices (Bayer, Ferreira and McMillan, 2007; Kane, Riegg and Staiger, 2006; Figlio and Lucas, 2004; Black, 1999). Most of these analyses use boundary discontinuity methods at school attendance zone boundaries, comparing diﬀerences in home values across school boundaries with diﬀerent test scores.2 The results from these studies tend to ﬁnd that a one standard deviation diﬀerence in test scores is associated with two to ﬁve percent higher property values.
However, a drawback of using average test scores to measure school quality is that they are highly correlated with the composition of schools and neighborhoods. Indeed, both Bayer, Ferreira and McMillan (2007) and Kane, Riegg and Staiger (2006) show that neighborhood and housing characteristics change at school boundaries due to endogenous parental sorting, and the estimated eﬀects of test scores on home prices is reduced signiﬁcantly in these studies once they control for neighborhood characteristics. These ﬁndings suggest that part of the capitalization of test scores into property values is due to the high value placed on the composition of school and neighborhood peers rather than on the school’s ability to educate students.3 In order to isolate the capitalization of school quality as it relates to the production of learning, a school quality measure that is less related to demographic characteristics than are test score levels is needed. Value-added represents such a measure, as it typically is generated using statistical models that control for students’ prior test history in order to estimate the current teacher’s and school’s contribution to learning through the student’s growth in test scores.4 As we demonstrate, the value-added scores in our study are much more weakly correlated with student characteristics than are test score levels, and the most of the information contained See Black and Machin (2011) for a comprehensive review of this literature. Much international work also uses this method, such as Gibbons and Machin (2003, 2006) and Gibbons, Machin and Silva (2013) in England, Fack and Grenet (2010) in France and Davidoﬀ and Leigh (2008) in Australia.
Cellini, Ferreira and Rothstein (2010) show that investments in school facilities are highly valued by local communities. These results are consistent with residents placing signiﬁcant value on non-learning aspects of schools.
This is done in a few diﬀerent ways. Some models simply control for lagged achievement and/or student demographics and calculate residuals. Other more complex models include student ﬁxed-eﬀects or Bayesian smoothers. See Guarino, Reckase and Wooldridge (2012), Rothstein (2010), and Kane and Staiger (2008) for discussions of the beneﬁts and drawbacks of such models. We also note that value-added models have come under considerable scrutiny on statistical grounds, suggesting value-added models may yield a very noisy, and possibly biased, signal of school or teacher quality (Rothstein, 2010). Nonetheless, recent research has argued that if done correctly, value-added methods can produce accurate measures of teacher and school quality (Kane and Staiger, 2008; Chetty, Friedman and Rockoﬀ, 2011).
in these estimates was not predictable using observable school characteristics before they were published. Thus, our results provide new information about valuation of a school quality indicator that provides previously unknown information about a school or teacher’s contribution to test score growth rather than information about the demographic makeup of the school.
While some prior work has shown that property values are unresponsive to value-added as calculated by the researcher (e.g., Dills, 2004; Downes and Zabel, 2002; Brasington, 1999;
Haurin and Brasington, 2006),5 parents and homeowners do not have direct access to this information. It is thus unlikely they would respond to it because the information is not salient.
In contrast, we study a unique and unanticipated release of value-added information for schools in the Los Angeles Uniﬁed School District (LAUSD). The value-added measures were easily available and widely discussed in local and national news outlets, making them highly salient to home buyers and local residents. To our knowledge, this is the ﬁrst analysis to identify the responsiveness of home prices to the release of this type of school quality information.
The information experiment that forms the basis for our study began in August 2010, when the Los Angeles Times (LAT) published average value-added estimates for 470 elementary schools as well as individual value-added estimates for 6000 third through ﬁfth grade teachers in LAUSD. In April 2011, LAUSD released their own value-added estimates, and in May 2011 the LAT updated their estimates to reﬂect the receipt of additional data. Prior to the initial release, California already provided information on the eﬀectiveness of LAUSD schools through published passing rates on the California Standards Tests and Academic Performance Index (API) scores. The API scores are based on average school performance on standardized exams and thus provide a summary measure of school-average test score levels. When the LA Times released the value-added data, they also provided information about API scores and passing rates on the same web page. Although these were already publicly available, the LA Times intervention potentially increased public awareness of these scores.
Using home sales data we obtained from the LA County Assessor’s Oﬃce from April 2009 through September 2011, we estimate diﬀerence-in-diﬀerence models that identify how home Gibbons, Machin and Silva (2013) is the only study of which we are aware whose results suggest that test score levels and value-added are similarly valued.
prices change after the release of each set of value-added data as a function of the valueadded scores. We ﬁnd no evidence that the composition of home sales changed due to the information release, nor do we observe any change in foreclosure rates or diﬀerential pre-release trends as a function of value-added, which supports the use of our empirical methodology.
Furthermore, using boundary discontinuity methods, we show that API scores are similarly capitalized into home prices as has been reported in other studies. We ﬁnd no evidence, however, that value-added information is valued by local residents: the diﬀerence-in-diﬀerence estimates are universally small, are not statistically diﬀerent from zero, and are suﬃciently precise that we can rule out all but very small positive eﬀects.
Unlike previous work on school quality valuation, we are able to examine how within-school variation in teacher quality is capitalized into property values, rather than just the school-level mean. It could be the case that home prices react more to the presence of a set of very good or very bad teachers, which school-level value-added can miss. We identify how home prices change as a function of the standard deviation of teacher value-added and the proportion of teachers in each school in each quintile of the value-added distribution. Our estimates are inconsistent with any home price response to the distribution of estimated teacher quality.
Finally, we examine several potential sources of heterogeneity and ﬁnd suggestive evidence that value-added information is more highly valued in lower SES schools.
Overall, our results indicate that value-added information, as presently constructed, is not valued by local communities. One potential conclusion that could be drawn from our estimates is that marginal homebuyers do not value the aspect of school quality that is embedded in value-added, namely the ability of schools and teachers to raise test scores. Alternatively, the public could ignore value-added information due to its statistical complexity as well as the uncertainty amongst the research community as to the accuracy of these measures.
Our ﬁndings have important implications for the release of these data more broadly in the US. Typically, the public release of value-added is contentious, with teacher groups arguing value-added is ﬂawed and uninformative and with community advocates arguing that people have a right to know this information. Our results suggest that in their current form, the public does not respond to value-added information, and that while this information may not be causing the distortions about which the opponents of publishing value-added data worry, they also are not being valued as relevant school quality information that constitutes the main reason for publishing these data.
In 2010, the Los Angeles Times newspaper acquired individual testing records of elementary students in Los Angeles Uniﬁed School District via a public information request. The achievement scores were linked to teachers so that a teacher and school value-added analysis could be conducted. The LA Times hired Dr. Richard Buddin to conduct the statistical analysis.
Details on the methodology can be found in Buddin (2010), but the basic strategy is to use a linear regression model with teacher ﬁxed-eﬀects to calculate teacher value-added. Teacher ﬁxed eﬀects are replaced with school ﬁxed eﬀects to calculate school value-added. All models control for lagged test scores and controls for student characteristics. Following completion of the analysis, the newspaper wrote a series of articles explaining the methodology and other issues in LAUSD throughout the month of August 2010 as a lead in to the release of the data in a simpliﬁed form on August 26, 2010. The value-added data were presented through an online database and could be accessed by anyone with a computer without charge or registration.6 The database was searchable by school and teacher name and people also could access information through various links oﬀ of the main web page.
Figure 1 shows an example of how the information was presented for a given school. Schools were categorized as “least eﬀective,” “less eﬀective,” “average,” “more eﬀective,” and “most effective,” which refer to the quintiles of the value-added score distribution for LAUSD. However, as Figure 1 demonstrates, the black diamond shows each school’s exact location in the distribution, providing parents with the ability to easily estimate the school’s percentile. Although The current version of the database can be accessed at http://projects.latimes.com/value-added/. The web portal is similar to the one that was available in August 2010 but now provides information for more teachers and more detail on the value-added measures. In most cases one can access the original August 2010 release through links on the teacher and school web pages.