{"title":"Different Data, Different Measures: Comparing Alternative Indicators of Changes in Neighborhood Home Values","authors":"Dan Immergluck, Adria Hollis","doi":"10.1080/10511482.2023.2254749","DOIUrl":null,"url":null,"abstract":"AbstractUrban scholars and practitioners have used changes in neighborhood-level home values to serve as indicators of neighborhood change, including gentrification and disinvestment. A common measure is the “median home value” variable from the American Community Survey (ACS). However, household-level research suggests that self-assessed home values, such as those of the ACS, differ significantly from market-based measures, and medians can be affected by changes in the mix of homes . Transaction-based home price indices are unaffected by such changes and are based on market sales rather than self-assessments, but also have limitations. Moreover, self-assessments of home values might be desired if the intention is to measure the value households place on their homes or to avoid potential biases baked into market values. Comparing changes in the ACS median home value to a common market-based home price index (HPI), we find that the ACS median tends to fall more slowly than the HPI when values are falling and increase more slowly than the HPI when values rise. The differences between the measures are large and are not randomly distributed across space, tending to be larger in neighborhoods where values fall or rise more steeply. They are also related to a variety of neighborhood characteristics.Keywords: Neighborhoodneighborhood changehousinghome valuesgentrificationdisinvestmentproperty values AcknowledgementsWe would like to thank the editor and the three anonymous reviewers for their very helpful comments on this paper.Disclosure StatementNo potential conflict of interest was reported by the author(s).Notes1 Regression is used to account for varying periods between paired transactions. The variation in changes in housing values is assumed to increase with the time between transactions, because variables other than market appreciation are expected to influence the values of housing units as this period increases. For more detailed information on the general FHFA repeat-sales methods, see Calhoun (Citation1996).2 The census-tract-level FHFA Home Price Index is provided here: https://www.fhfa.gov/DataTools/Downloads/Documents/HPI/HPI_AT_BDL_tract.csv. More information on how the index is constructed is provided in Federal Housing Finance Agency (Citation2023).3 Moreover, the HPI is a relative measure of home value compared to other points in time and does not provide dollar-value estimates of median or typical value at one point in time.4 The Missouri Census Data Center release of the 2005–2009 ACS median home value variable was spatially interpolated using owner-occupied housing units as the weighting variable. More information can be found at https://mcdc.missouri.edu/data/acs2009/Variables.html.5 The two exceptions are the 2007–2012 HPI change and the initial median home value, which is taken from the 2005–09 ACS, centered on 2007.6 Tables 3–5 show a slight difference in sample size of one tract between the 2007 to 2012 period regression and the 2012–2017 period regression. This is due to a slight difference (one tract) in the missing ACS variables for the regression across the two different time periods.7 Heteroskedastic-robust standard errors were used in both regressions. Variance inflation factors (VIFs) were calculated for both regressions. The average VIFs were under 3.0, with no variables (other than those involving polynomial terms) having VIFs larger than 7.5.Additional informationNotes on contributorsDan ImmergluckDan Immergluck is Professor of Urban Studies at Georgia State University. His research concerns housing, race, neighborhood change, gentrification, segregation, real estate markets, and urban political economy. Dr. Immergluck is the author of five books and over 120 scholarly articles, book chapters, and research reports. He has consulted for the federal agencies, philanthropic foundations, and nonprofits. His latest book is Red Hot City: Housing, Race, and Exclusion in Twenty-First Century Atlanta (UC Press, 2022).Adria HollisAdria Hollis is a graduate student in the Urban Studies Institute at Georgia State University. She is currently a program and policy intern for Enterprise Community Partners. She was previously a graduate research assistant in the Urban Studies Institute. She received her B.A. from Rust College.","PeriodicalId":47744,"journal":{"name":"Housing Policy Debate","volume":"51 1","pages":"0"},"PeriodicalIF":2.8000,"publicationDate":"2023-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Housing Policy Debate","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1080/10511482.2023.2254749","RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"DEVELOPMENT STUDIES","Score":null,"Total":0}
引用次数: 0
Abstract
AbstractUrban scholars and practitioners have used changes in neighborhood-level home values to serve as indicators of neighborhood change, including gentrification and disinvestment. A common measure is the “median home value” variable from the American Community Survey (ACS). However, household-level research suggests that self-assessed home values, such as those of the ACS, differ significantly from market-based measures, and medians can be affected by changes in the mix of homes . Transaction-based home price indices are unaffected by such changes and are based on market sales rather than self-assessments, but also have limitations. Moreover, self-assessments of home values might be desired if the intention is to measure the value households place on their homes or to avoid potential biases baked into market values. Comparing changes in the ACS median home value to a common market-based home price index (HPI), we find that the ACS median tends to fall more slowly than the HPI when values are falling and increase more slowly than the HPI when values rise. The differences between the measures are large and are not randomly distributed across space, tending to be larger in neighborhoods where values fall or rise more steeply. They are also related to a variety of neighborhood characteristics.Keywords: Neighborhoodneighborhood changehousinghome valuesgentrificationdisinvestmentproperty values AcknowledgementsWe would like to thank the editor and the three anonymous reviewers for their very helpful comments on this paper.Disclosure StatementNo potential conflict of interest was reported by the author(s).Notes1 Regression is used to account for varying periods between paired transactions. The variation in changes in housing values is assumed to increase with the time between transactions, because variables other than market appreciation are expected to influence the values of housing units as this period increases. For more detailed information on the general FHFA repeat-sales methods, see Calhoun (Citation1996).2 The census-tract-level FHFA Home Price Index is provided here: https://www.fhfa.gov/DataTools/Downloads/Documents/HPI/HPI_AT_BDL_tract.csv. More information on how the index is constructed is provided in Federal Housing Finance Agency (Citation2023).3 Moreover, the HPI is a relative measure of home value compared to other points in time and does not provide dollar-value estimates of median or typical value at one point in time.4 The Missouri Census Data Center release of the 2005–2009 ACS median home value variable was spatially interpolated using owner-occupied housing units as the weighting variable. More information can be found at https://mcdc.missouri.edu/data/acs2009/Variables.html.5 The two exceptions are the 2007–2012 HPI change and the initial median home value, which is taken from the 2005–09 ACS, centered on 2007.6 Tables 3–5 show a slight difference in sample size of one tract between the 2007 to 2012 period regression and the 2012–2017 period regression. This is due to a slight difference (one tract) in the missing ACS variables for the regression across the two different time periods.7 Heteroskedastic-robust standard errors were used in both regressions. Variance inflation factors (VIFs) were calculated for both regressions. The average VIFs were under 3.0, with no variables (other than those involving polynomial terms) having VIFs larger than 7.5.Additional informationNotes on contributorsDan ImmergluckDan Immergluck is Professor of Urban Studies at Georgia State University. His research concerns housing, race, neighborhood change, gentrification, segregation, real estate markets, and urban political economy. Dr. Immergluck is the author of five books and over 120 scholarly articles, book chapters, and research reports. He has consulted for the federal agencies, philanthropic foundations, and nonprofits. His latest book is Red Hot City: Housing, Race, and Exclusion in Twenty-First Century Atlanta (UC Press, 2022).Adria HollisAdria Hollis is a graduate student in the Urban Studies Institute at Georgia State University. She is currently a program and policy intern for Enterprise Community Partners. She was previously a graduate research assistant in the Urban Studies Institute. She received her B.A. from Rust College.
期刊介绍:
Housing Policy Debate provides a venue for original research on U.S. housing policy. Subjects include affordable housing policy, fair housing policy, land use regulations influencing housing affordability, metropolitan development trends, and linkages among housing policy and energy, environmental, and transportation policy. Housing Policy Debate is published quarterly. Most issues feature a Forum section and an Articles section. The Forum, which highlights a current debate, features a central article and responding comments that represent a range of perspectives. All articles in the Forum and Articles sections undergo a double-blind peer review process.