Abstract
The divergence in housing price growth in the US in coastal cities relative to inland cities has been thought to occur, in large part, due to severe housing regulations and restrictions on development. Researchers have posited that this trend implies that these heavily regulated cities are experiencing higher incidences of gentrification. However, the gentrification of lower-income communities may be negatively influenced by restrictive regulations rather than positively, as is the case with overall housing price growth. This may occur if restrictions make it more difficult to improve housing structures and engage in new housing projects. We use data from over 12,000 census tracts to analyse the relationship between land use regulations and the probability an area will undergo gentrification in the years 2000 to 2010. By separating the influence of higher levels of regulation on overall housing price growth from the likelihood that a lower-income neighbourhood will gentrify, we find that regulation has opposing forces. While increased levels of regulation are associated with an almost 10% greater increase in overall housing prices, they are also associated with a three to four percentage-point lower probability that a lower-income tract will experience gentrification, contrary to previous conclusions.
Introduction
Land use restrictions have been shown to play a large role in explaining housing price growth differentials between regions (Glaeser and Ward, 2009). Recent research has demonstrated that higher levels of regulation are associated with higher growth in housing prices, although the mechanism by which regulation may influence house prices is difficult to establish. This influence of regulation on the growth in overall housing prices has been taken as evidence that these areas are experiencing a greater degree of gentrification (Kahn et al., 2010). However, determinants of gentrification need not be consistent with determinants of overall housing price changes. Gentrification is typically defined as the transformation which occurs when a lower-income, neglected/decaying neighbourhood experiences increased investment in housing and growth, which attracts higher-income residents and potentially displaces lower-income residents. This revitalisation process can take on many forms but has the common experience of describing a change in lower-income neighbourhoods in particular. Consequently, an increase in average housing prices need not correspond with an influence on gentrification patterns or the overall rate of gentrification.
Land use restrictions typically take the form of local regulations which involve housing construction or renovation. These restrictions include any hurdles or constraints in the regulatory process involved in undertaking housing projects as well as restrictions imposed by the local community (minimum lot size restrictions, open space requirements, etc.) (Gyourko et al., 2008). Intuitively, the regulatory forces that drive up all housing prices may have the opposite influence on any particular neighbourhood’s likelihood of experiencing gentrification. Regulatory burdens which dampen housing supply may also dampen the ability or profitability of renovating decaying housing stock and improving the surrounding facilities. If true, neighbourhoods in areas with higher levels of land use restrictions may be less likely to experience the investment in improving existing housing stock and commercial spaces that accompanies gentrification than areas with lower levels of land use restrictions.
Although considerable empirical analysis has been devoted to the effect of land use regulation on housing supply and prices, less attention has been paid to how these regulations may influence gentrification in particular. Gentrification of downtown areas can increase the attractiveness of such areas for residential and business activity, ultimately serving as a catalyst for economic growth. If regulation reduces the likelihood that an area will gentrify, then attempts to capture the economic consequences of land use regulation may underestimate the effects of these types of regulation. We use data from over 12,000 census tract areas from 2000 to 2010 to consider the influence of land use restrictions on the probability that a particular neighbourhood will experience gentrification, controlling for other influential factors. The Wharton Residential Urban Land Use Regulation Index (WRLURI) (Gyourko et al., 2008) is used to measure the intensity of land use regulation in the larger geographic areas and we use above average growth rates of socioeconomic indicators in low-income neighbourhoods to identify gentrification.
With respect to overall housing prices we find that, similar to previous findings, higher levels of regulation are indeed positively associated with a higher growth in house prices. More specifically, an increase in the intensity of land use regulation is associated with an 8–10% increase in housing values. However, when we classify a neighbourhood as gentrified during our time period, we find that a one standard deviation increase in the level of land use regulation is associated with a reduction in the probability that a tract will gentrify by approximately four percentage points. A similar association is found when we loosen our classification criterion of a ‘gentrification eligible’ neighbourhood.
This finding of opposing influences was not apparent in previous analyses which did not differentiate between overall housing price growth and gentrification in particular. That higher levels of regulation increase house price growth rates while decreasing the probability of gentrification suggests that land use regulations may be costlier to local areas than previous estimates suggest. Furthermore, our findings suggest that higher levels of land use regulation may be stifling the ability of neighbourhoods to gentrify, offering policymakers a potential tool to revitalise lower-income neighbourhoods by reducing local regulations related to housing supply and construction.
Literature review
Land use regulations may influence both the supply of and demand for housing, making it difficult to estimate the direct influence of regulation on housing supply and prices. Demand-side factors can broadly be categorised as being motivated by political influences, a desire to offset negative externalities or a desire to engage in exclusionary zoning (Quigley and Rosenthal, 2005).
The regulation which arises from these demand-driven forces often restricts the type and/or quantity of housing construction or increases the cost associated with development, which reduces the potential housing supply. However, in addition to the demand-side forces prompting increased regulations, land use regulations may also act to increase the desirability of the area, increasing demand for housing in regulated areas (Irwin and Bockstael, 2002; Irwin et al., 2014; Walsh, 2007; Wu et al., 2004). This may spur demand further through a ‘social multiplier effect’, as higher-income households desire to live closer to other higher-income households, bidding up house prices and prompting increases in housing supply (Waldfogel, 2008). Consequently, houses in areas with higher levels of regulation are expected to have higher prices, while the effect on housing supply density is ambiguous (Ihlanfeldt, 2007; Kahn et al., 2010). While several studies have found evidence that housing prices are higher in areas with more land use regulations, others have found small or insignificant effects (Quigley and Rosenthal, 2005).
Saiz (2010) incorporates detailed geographic data to analyse the relationship between house prices and regulations with a model of endogenous regulations tied to geographic constraints. He finds that areas with high land values (due to constraints on supply) are more likely to have restrictive land use regulation, exacerbating the effect on housing supply and prices. Regulation is measured broadly using the WRLURI (Gyourko et al., 2008), an index comprised of different components of restrictive land use policies, practices and formal regulations.
More recently, Kahn et al. (2010) consider the influence of land use regulation on house prices and supply of housing in Californian coastal zones. Unlike broad measures of regulation intensity, the use of regulations which target designated coastal zones allows Kahn et al. (2010) to estimate these effects more precisely. They find evidence that these regulations have resulted in higher housing prices and increased housing supply (albeit with declining relative population density), consistent with general predictions. 1 Kahn et al. (2010) also explore the influence of land use regulation on gentrification patterns in these areas. They define gentrification as the relative proportion of high-income census tracts residing in the regulatory zone compared to similar tracts outside of the regulatory zone. They find an increased share of ‘rich’ tracts within the bounds since the regulation was enacted in the 1970s. This is consistent with a general trend of higher incomes and house prices in the regulatory zone during this time period. The authors conclude that the coastal zone regulation appears to have had a positive influence on gentrification in the affected areas.
However, gentrification is typically described as a process in which decaying or neglected neighbourhoods are revitalised through renovating, re-purposing or constructing new commercial and/or residential spaces by new residents who are higher-income than existing residents (Levy et al., 2006). Although the particular designation of what constitutes gentrification may vary, the measurement is typically constructed to capture the experience of lower-income neighbourhoods undergoing a transformation to the point that the neighbourhood is no longer considered lower-income. Consequently, the finding by Kahn et al. (2010) that housing regulations are positively associated with the proportion of higher-income residents may not be evidence that gentrification is also positively associated with housing regulations. Although regulation may increase the desirability of a given locale, there is reason to believe that these effects may not be felt evenly in the affected area. The same forces which increase the desirability of areas with already high-income households may increase the cost of renovating or re-purposing lower-income neighbourhoods, dampening the ability of these areas to undergo gentrification.
Previous research which focuses on gentrification and land use regulation has largely concentrated on the experience of lower-income households in areas that have already gentrified. While gentrification offers communities the opportunity to enjoy improved housing stock and potentially experience economic revitalisation, the resulting increase in housing prices and rents can potentially result in displacement of low-income families who can no longer afford to reside in neighbourhoods that are undergoing gentrification. Although some studies have found evidence of displacement due to gentrification (Atkinson, 2000), other studies have found little evidence for residential turnover due to gentrification (Freeman and Braconi, 2004). Ding et al. (2016) consider resident mobility in areas undergoing gentrification and find that more advantaged residents had a slightly higher mobility rate than similar households in non-gentrifying neighbourhoods. The most vulnerable residents have similar rates of mobility to equivalent households in non-gentrifying neighbourhoods, although of those who did move, they were more likely to move to lower-income neighbourhoods. However, Newman and Wyly (2006) note that even limited displacement exacts a significant toll on those affected.
Studies which explicitly consider the relationship between land use regulation and gentrification typically consider the effect of policy-related regulations which provide or impede low-income housing solutions for individuals who are priced out of recently gentrified areas (Ghose, 2004; Newman and Wyly, 2006). In this way, regulation related to gentrification takes on the form of affordable housing options for lower-income residents. These forms of regulation occur as a response to gentrification rather than an influence on the propensity of a neighbourhood to gentrification. In this study, we wish to contribute to the understanding of regulation as a potential determinant of gentrification rather than a response.
The literature examining the determinants of gentrification, rather than the effects, has been given theoretical and empirical consideration. Brueckner and Rosenthal (2009) form a model of gentrification, originally proposed by Alonso (1960) and Muth (1969), in which households prefer newer construction in the suburbs and accept a longer commute as a trade-off. As the housing stock near the Central Business District (CBD) ages, renovation of these older homes is a profitable endeavour and households move back to these renovated, centrally located neighbourhoods. In this way, the age of housing stock and distance to the CBD are taken to be primary determinants of gentrification. Guerrieri et al. (2013) incorporate a spillover effect of living near wealthier neighbourhoods to a model of gentrification. In this model, higher-income areas experience lower crime levels and increased access to amenities, which spill over into surrounding neighbourhoods. Lower-income neighbourhoods bordering these areas are relatively more desirable and, consequently, are more likely to gentrify. They use zip code-level housing prices from 1988 to 2008 to examine the influence of a housing demand shock, and find evidence that bordering a higher-income area does increase the likelihood that an area will gentrify, consistent with their model predictions.
Previously, Kolko (2007) also considered these predictions empirically and found evidence which is consistent with the predictors of gentrification proposed by Brueckner and Rosenthal (2009) and Guerrieri et al. (2013). Kolko (2007) uses census data to examine gentrification from 1980 to 2000 and finds evidence of gentrification patterns consistent with a household’s desire to reside close to the CBD, an increased likelihood of gentrification in areas with an ageing housing stock and evidence of a preference to be located near wealthier census tracts.
These extensive strands of research highlight the importance of regulation in influencing housing supply and prices and, separately, the many determinants and consequences of gentrification. While there is strong empirical evidence that land use regulation positively influences housing values (Kahn et al., 2010; Saiz, 2010), it is unclear how the intensity of regulation will influence an area’s propensity to gentrify. The empirical research on gentrification has historically focused on predictors of gentrification unrelated to regulation or the effect of regulation post gentrification. To our knowledge, this study is the first to directly consider how regulation may have opposing influences on overall house prices compared to the ability of lower-income neighbourhoods to experience gentrification.
Data and methodology
We examine data from 42 Metropolitan Statistical Areas (MSAs), which include 12,576 census tracts in our sample for the period of 2000–2010. We employ neighbourhood and regional characteristics to estimate the influence of the intensity of housing regulations in 2000 on the probability that a census tract will experience gentrification by 2010. Our census tract and MSA data come from the 2000 Decennial Census and the 2008–2012 American Community Survey. One issue with using this data is that some census tract classifications are not consistent between years. To account for these discrepancies, we use the census data set created by Logan et al. (2014), which recalculates 2000 census data to the 2010 census tract definitions used in the 2008–2012 American Community Survey.
Census tracts are designed to include around 1500 households (4000 individuals), but may vary based on local population density and geography. Although census tracts do not perfectly capture neighbourhood designations, researchers use the census tract borders to analyse neighbourhood effects and influences. Consequently, we classify households as belonging to the same neighbourhood if they are in the same census tract. So that we may potentially highlight differential outcomes related to regulation on overall housing prices and gentrification, we also use this data to estimate the influence of regulation on overall housing prices during this time period.
Although the description of gentrification is broadly adopted as a social and economic transformation/rejuvenation of a decaying neighbourhood, there is not a generally accepted classification to identify an area as having gentrified. We follow previous descriptions which consider gentrification to be ’the socioeconomic upgrading of a previously low-income, central city neighborhood, characterized by the influx of higher socioeconomic status residents and an increase in housing prices’ (Ding et al., 2016: 38). Different classifications of gentrification may yield very different conclusions, necessitating careful construction of the criterion (Barton, 2016). While smaller-scale case studies of gentrification typically rely on multidimensional shifts in neighbourhood composition, amenities and housing stock quality, large-scale analysis generally relies on changes in average income and house prices (Freeman, 2005). We also restrict our sample to neighbourhoods located in the central city to ensure that we are capturing the experience of urban gentrification rather than rural transformations.
Wyly and Hammel (2004) consider field survey classification markers of identified gentrification and find that these identified neighbourhoods are closely predicted when shifts in average income are used as a classification for gentrification, but that income as a sole marker may miss gentrification occurring due to an influx of a low-income creative class (artists) or of recent college graduates (Owens, 2012). Whether we employ an income change measure or a multidimensional socioeconomic and housing measure, if we simply consider an increase in these average characteristics to be evidence of gentrification, we would expect that areas with higher levels of these characteristics may demand more regulation, consistent with previous findings. Our primary interest requires a measure which will identify lower-income neighbourhoods that are improving. To identify whether a lower-income neighbourhood is experiencing an increase in higher socioeconomic residents, we employ a measure of the change in demographic composition coupled with changes in housing costs. More specifically, we classify a low-income area as having gentrified if it has experienced above MSA average growth in the college-educated population and either house price growth or rent growth that was positive and above the MSA average. In doing so, we are considering how regulation influences the propensity of lower-income neighbourhoods to gentrify, differentiating from previous strands of research which consider how land use regulation affects all neighbourhoods (rather than focusing on the experience of lower-income neighbourhoods in particular).
The appropriate income designation to identify areas that are eligible to be gentrified has not been settled in the literature. As previously mentioned, Kahn et al. (2010) consider gentrification to be a rise in average house prices and, consequently, all census tracts would be eligible for gentrification. McKinnish et al. (2010) and Ding et al. (2016) consider a stricter requirement that census tracts must have an average income below the 50th percentile to be eligible for gentrification, while other research has imposed an even stricter definition that an eligible tract be in the bottom 25th percentile (Christafore and Leguizamon, 2019). An important consideration in our study is the separation between overall economic improvement and gentrification in particular, so we employ the more stringent definition used by Christafore and Leguizamon (2019). More specifically, we identify census tracts as eligible to be gentrified if they are in the bottom 25% of average income in the MSA in 2000. These tracts are identified as gentrified if the tract experiences above the MSA average growth in college-educated residents and either house price growth or rent growth that is positive and above the MSA average. Of the 12,576 census tracts in our sample, 5078 are eligible for gentrification. A list of MSAs included in our data set and summary statistics on the percentage of census tracts eligible for gentrification can be found in Table 1. On average, 41% of eligible tracts within each MSA experienced gentrification, with as much as 70% of low-income tracts gentrifying in Washington DC, while only 29% in Phoenix, Arizona. Overall, 42.5% of low-income tracts experienced gentrification between 2000 and 2010.
Proportion of low-income tracts gentrified by MSA.
Notes: All city tracts included in the analysis and gentrifiable tracts are those in the first income quartile.
Prop. Gent. is the proportion of gentrifiable tracts that gentrified.
Our measure of land use regulation intensity is the WRLURI (Gyourko et al., 2008), whose MSA values are also reported in Table 1. The use of a regulatory index to identify the intensity of land use regulation has been employed most prominently by Glaeser and Ward (2009) and Saiz (2010). The index was created from a 2005 nationwide survey of over 2600 municipalities on local land use regulations, and uses responses from a questionnaire designed to capture three aspects of the regulatory environment: (1) the number and type of agencies involved in the process of zoning requests, (2) the current local regulation rules and (3) the effects of the regulation on development costs and time delays. The regulatory environment for each category was computed from the responses from 15 survey questions, which were used to create a final index comprised of 11 sub-indices, i.e. local political pressure, state political involvement, state court involvement, local zoning approval, local project approval, local assembly, density restrictions, open space index, exactions index, supply restrictions and approval delay. Gyourko et al. (2008) also incorporate state-level regulatory action and policies, as well as a measure of ‘community pressure’ to adopt further regulation.
Since gentrification involves a physical transformation of residential and/or commercial spaces, additional time or monetary cost associated with changing the physical landscape will act as a direct increase in the costs of this investment. In this way, a measure of the intensity of land use regulations may be taken as a measure of an additional cost of gentrification. Although data exist to consider the influence of each separate component of regulation intensity, Gyourko et al. (2008) recommend the use of the index which combines all of these features because the measures are highly correlated with one another. However, since our data are an aggregated measure of the original index, the sub-indices of the index are not highly correlated. 2 Consequently, we use the index which combines the sub-indices as our primary measure of regulation intensity but we also consider the influence of the sub-indices separately. The cumulative score is standardised to have a mean of zero and a standard deviation of one across communities. It reflects the overall regulatory environment, with a lower score implying less regulatory interference.
The index is provided at the community level, and we restrict our sample to MSAs with at least 10 community observations, providing us with a data set comprised of neighbourhoods in 46 MSAs. 3 We average the community-level regulatory index to the MSA level and re-standardise the index so that the values for MSAs in our sample have a sample mean of zero and a standard deviation equal to one. 4 Consequently, the index may be interpreted as the deviation from our sample mean rather than a directly interpretable value. The regulatory index ranges from a high of 1.79 standard deviations above the average in Providence Rhode Island to a low of 0.8 standard deviations below the mean in Kansas City. Metropolitan areas in Texas consistently experience the lowest regulatory intensity, with all areas having regulatory index values below the mean.
In addition to the regulation index (WRLURI), we also include a control for the physical land unavailability from Saiz (2010). This unavailability measure is the estimated percentage of land area that is undevelopable due to geographic constraints (oceans, mountains, lakes, steep slopes, etc.), and MSAs in our sample report an average of 30% undevelopable land. Because Saiz (2010) finds that higher levels of undevelopable land are positively associated with higher levels of regulation, it is necessary to include both factors in our analysis. The physical land unavailability index provides a measure for 42 of the 46 MSAs that are included in the regulation data. Consequently, our data set of 12,576 observations is comprised of central city neighbourhoods in 42 MSAs that have at least 10 community-level regulatory values and a measure of physical land unavailability.
We first consider the influence of regulation intensity on overall housing prices, similar to Kahn et al. (2010). During this time period, house values increased by approximately 44% on average. We allow this influence to vary along the income distribution of census tracts within the MSA so that we can identify potentially different outcomes associated with regulation for higher- and lower-income areas. That is, we break up tracts into four subsets based on income quartile and employ OLS to estimate the influence on higher-income neighbourhoods separately from lower-income neighbourhoods. The model is then:
where the neighbourhood demographics commonly used in housing price determinants, regional economic controls and the regulatory index value are included. The change in log home values between 2010 and 2000 is the dependent variable, and our neighbourhood-level control variables include the log population, log median household income, log population density, percent white, percent with a bachelor’s degree, percent foreign born, percent unemployment rate, percentage of houses more than 30 years old, percentage of housing units that are vacant and percentage of housing units that are owner occupied. Regional-level controls include the log MSA population, log MSA median household income and MSA unemployment rate, and robust standard errors clustered at the MSA are employed.
All controls take on the value in 2000, and Summary Statistics for these controls can be found in Table 2.
Variable definitions and statistics.
Notes: All city tracts are included in the analysis. Gentrifiable tracts are those in the first income quartile.
In order to distinguish between rising house prices overall and gentrification of lower-income neighbourhoods in particular, we consider the influence of regulation on the propensity of a neighbourhood to gentrify. We restrict our data to low-income census tracts located in the central city that are eligible to be gentrified, leaving our sample with 5078 tracts. Our Probit model takes on the following form: P(gentrified = 1) = G(β0 + xB), where x includes all our control variables and G is the standard normal cumulative function. In the results tables, we report the estimated marginal effects calculated at the means of the control variables.
We incrementally include controls that have been shown to affect the likelihood that an area will gentrify, following Brueckner and Rosenthal (2009), Kolko (2007) and Christafore and Leguizamon (2019). Since the identification of our estimates comes from differences across MSAs rather than census tracts, we employ MSA-clustered standard errors. We begin by considering how regional characteristics may influence the probability of gentrification. Specifically, we include the MSA unemployment rate, the log MSA population and the log MSA median household income. Controls for geographic regions are then added, followed by a consideration of how demographic controls may influence this observed relationship. These demographic controls include neighbourhood log population, log median household income, log population density, unemployment rate, percentage of households that are foreign born, percentage of households with a college degree and percentage of households that are white. Our final, and full, specification includes housing characteristics previously found to influence gentrification. These are the log median house value, the percentage of houses over 30 years old, the percentage of houses that are vacant and the percentage of houses that are owner occupied.
Our primary specification uses the values of the control variables in the year 2000 rather than the 2000–2010 change in the values of these variables. This is done to ensure that the mechanisms that allow a census tract to gentrify are not being controlled for by changes in other mechanisms that promote gentrification. Table 3 provides summary statistics of our control variables for low-income neighbourhoods which gentrified separately from neighbourhoods which did not gentrify. Although most variables exhibit little difference between neighbourhoods which gentrified and those that did not, a few had noticeable differences. Areas which gentrified had a larger share of college-graduated residents in 2000 (16.5% versus 12.33%) and a larger share of white households (28.8% versus 26.8%) but a smaller share of owner-occupied residences (32.7% versus 37.2%).
Proportion of low-income tracts gentrified by MSA.
Notes: Only gentrifiable tracts are included in the analysis. Gentrifiable tracts are those in the first income quartile.
Results
We first consider how regulation intensity may influence overall housing prices. In our approach, we consider the relationship between land use regulation and house prices separately for neighbourhoods in the bottom 25th percent, the 25th−50th percent, the 50th−75th percent and the 75th−100th percent, with respect to income distribution within the MSA. The reported estimations, provided in Table 4, are generally consistent with previous findings. Our analysis suggests that census tracts with a higher percentage of residents with a college degree are associated with a greater increase in housing values in 2010 for neighbourhoods along the entire income distribution; a higher percentage of foreign-born residents is positively associated for all but the top quartile of the income distribution; and a higher percentage of houses over 30 years old is associated with higher housing values for the middle 50% of the income distribution. With respect to regulation intensity, our estimation is consistent with previous findings of a positive, large and statistically significant influence on housing prices.
OLS results with change in log home value between 2010 and 2000 as the dependent variable, by income quartile.
Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Standard errors clustered by MSA in parentheses.
We find that a one standard deviation level of regulation intensity in 2000 is associated with an 8–9% greater increase in house prices in 2010. This is true for all houses along the income distribution, including those in the lowest quartile that are eligible for gentrification. Indeed, neighbourhoods in the lowest quartile have the highest estimated influence, with a one standard deviation higher level of regulation being associated with an almost 10% greater increase in house prices in 2010. Since the homes in the bottom quartile are the only homes eligible for gentrification, these findings point to a particularly strong, negative association between regulation and gentrification. Given that regulation has been found to decrease the supply and/or increase the costs of additional housing stock, it is not surprising that higher levels of regulation are associated with higher levels of housing price increases.
However, as previously stated, the forces that may be increasing house prices (reduced housing stock) may act as a damper on the ability of low-income neighbourhoods to experience gentrification. Considering only an increase in overall house prices rather than a compositional change in the neighbourhood relative to the ongoing changes in neighbourhoods across the MSA will be a reflection of overall growth rather than a neighbourhood transformation. To consider whether higher levels of regulation intensity may increase the costs associated with renovating decaying neighbourhoods, and consequently reduce the probability that a lower-income area may gentrify, we next employ our primary specification. Table 5 reports the marginal Probit effects estimation when the 2000 values of the control variables are included.
Probit marginal effects on gentrification: Tracts in bottom income quartile eligible to gentrify.
Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Standard errors clustered by MSA in parentheses.
As previously stated, we define gentrification as occurring when a neighbourhood experiences a growth in college graduates higher than the MSA average and above-average house price or rent growth between 2000 and 2010. We first consider only the influence of MSA-level controls (unemployment rate, log median household income and log population), the unavailability land use index and the regulatory index (column 1 in Table 5). At the MSA level, only the log median household income has a statistically significant influence, which is positive. This suggests that neighbourhoods in metropolitan regions which have higher median income levels are more likely to experience gentrification. With respect to regulation intensity, land use regulation is found to have a negative, statistically significant influence on the probability that a neighbourhood will have gentrified. We estimate that a one standard deviation higher level of regulation intensity in 2000 is associated with a 2.75 percentage point lower probability that the census tract will have gentrified by 2010.
It is possible that regional attitudes towards regulation may be associated with unobserved characteristics that may be negatively associated with the forces of gentrification. To account for this possibility, we include controls for a neighbourhood being located in the South, Northeast or West Census Regions of the US (column 2 in Table 5), with the omitted region being the Midwest. Being located in the South or West increases the probability of gentrification relative to the probability of neighbourhoods in the Northeast or Midwest. The log MSA median household income remains positive and significant, and the log MSA population and unavailability of physical land use index now also exhibit a statistically significant positive influence. Having less physical land to develop appears to increase the propensity of lower-income neighbourhoods to gentrify, while neighbourhoods in MSAs with higher populations are also more likely to gentrify. Note that these are still lower-income neighbourhoods but are in MSAs with a relatively higher average income. Regulation intensity continues to have a negative, statistically significant influence that is even more pronounced. With this specification, a one standard deviation higher level of regulation intensity is now associated with a 5.3 percentage-point lower probability that a neighbourhood will gentrify.
Initial neighbourhood demographic composition has been found to affect the likelihood that a neighbourhood will experience gentrification (Kolko, 2007). If metropolitan areas systematically differ in neighbourhood composition by region, the coefficient on our regulatory index may be reflecting these different demographics. Adding neighbourhood demographic controls slightly reduces this associated influence of regulation, but it remains statistically significant (column 3 in Table 5). We find that higher levels of neighbourhood median income and population in 2000 appear to be negatively correlated with the probability that a neighbourhood will gentrify by 2010, while a higher percentage of the neighbourhood with a college degree is positively associated with the probability that a neighbourhood will gentrify, consistent with previous findings. The influence of the unavailability of physical land index is now statistically insignificant, but all other previously significant control variables report similar influences. The neighbourhood log population density, unemployment, percent foreign born and percent white are estimated to have a statistically insignificant influence.
Finally, we consider the influence of neighbourhood housing stock characteristics. This full model specification, column 4 in Table 5, incorporates all previously considered characteristics. We control for the neighbourhood log home value, the percentage of homes over 30 years of age, the percentage of homes which are vacant and the percentage of homes which are owner occupied. We find that higher initial house prices (relative to low-income neighbourhoods in general) are positively related to the probability that an area will gentrify. Having a higher percentage of homes over 30 years old and having a higher percentage of vacant homes are also associated with an increased probability of gentrification. That an ageing housing stock is positively related to gentrification is consistent with theoretical models of gentrification patterns (Brueckner and Rosenthal, 2009). All previously considered control variables exhibit a similar influence, with the exception of log median household income (which is still negative but statistically insignificant).
Higher levels of regulation remain associated with a reduced probability that a low-income neighbourhood will experience gentrification, although this is now significant at the 10% significance level rather than the 5%. A one standard deviation higher level on the regulatory index is associated with a 3.8 percentage point decrease in the probability that a neighbourhood will gentrify.
Overall, low-income neighbourhoods in higher-income, larger MSAs which are located in the South and West are more likely to experience gentrification from 2000 to 2010. Low-income neighbourhoods with a relatively higher percentage of college graduates, higher home values, an ageing housing stock and more vacancies are also estimated to have an increased propensity to gentrify, while those located in relatively populated neighbourhoods are less likely to gentrify. These estimations are consistent with previous findings that neighbourhoods in higher-income areas, those with a higher percentage of college graduates, those with a higher percentage of vacancies and older homes are more likely to gentrify (Kolko, 2007). For all specifications, higher levels of land use regulation are consistently associated with a reduced probability of gentrification.
In Table 6, we have expanded the criterion of neighbourhoods eligible to experience gentrification to include neighbourhoods in the bottom 40% of the income distribution (within the MSA) instead of the bottom 25%. These specifications yield results similar with respect to direction of influence and magnitude to those of the full specification with our more stringent criterion for all models except Model 1 with MSA level controls only. Higher levels of regulation are associated with a reduced probability that an area will gentrify to approximately four percentage points for the models incorporating regional and neighbourhood demographics. For the full model specification, this associated influence is reduced to 2.5 percentage points and is insignificant (but with a p-value of 0.14). This reduced significance of regulation may suggest that the detrimental influence of regulation on gentrification is concentrated in the lowest-income neighbourhoods.
Probit marginal effects on gentrification: Tracts in first four income deciles eligible to gentrify.
Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Standard errors clustered by MSA in parentheses.
To consider which specific regulatory features may be influencing this relationship, we decompose the WRLURI into the 11 sub-indices previously mentioned. These are: local political pressure, state political involvement, state court involvement, local zoning approval, local project approval, local assembly, density restrictions, open space index, exactions index, supply restrictions and approval delay (Gyourko et al., 2008). 5 The Probit marginal effects estimates are provided in Table 7. For simplicity, only the coefficient on the regulatory index is provided for each estimation. 6 Of the 11 regulatory sub-indices, the local assembly index, the density restrictions index and the exactions index report a statistically significant influence on the probability that a low-income area will gentrify. The local assembly index and the density restriction index are negatively associated with the propensity that a neighbourhood will gentrify, while the exactions index is positively associated with gentrification. The local assembly index represents a democracy requirement and captures whether a community meeting or assembly must take place before any zoning changes are voted on, with a popular vote dictating whether the zoning or rezoning request is granted. It is possible that our estimated negative relationship between an assembly requirement and gentrification is capturing the propensity of communities with more power to veto zoning changes using that power to reduce the ability of developers to repurpose or renovate existing housing stock or to construct new housing stock in a pattern consistent with gentrification.
Proportion of low-income tracts gentrified by MSA.
Notes: Each marginal effect from a regression including the full set of controls.
The density restriction sub-index takes on a value of one if the community requires a minimum lot size of at least one acre anywhere within the jurisdiction, and zero if it does not. Our finding of a negative relationship may be capturing the additional cost of new construction due to restrictive minimum lot sizes. If these costs act as a deterrent to potential developers, an ageing housing stock may not attract potential residents interested in renovated or new construction housing in up-and-coming neighbourhoods, one of the hallmarks of gentrification. The one measure that is positively associated with gentrification is the exactions index. This index takes on a value of one if developers are required to pay a portion of new infrastructure resulting from the new development, and zero if they are not. These findings suggest that requiring developers to share the burden of improving the infrastructure yields a positive influence on the propensity of low-income neighbourhoods to gentrify. Coupled with the local assembly findings, one potential explanation is that communities may support new developments in which the developer is sharing some of the costs imposed on communities. If developers have community support and are more easily able to influence zoning, this may have positive spillover effects on gentrification efforts. In this way, the limiting aspect of a costly regulation may be outweighed by the community support benefit to developers.
Overall, these findings are consistent with previous findings that land use regulation is associated with higher house prices. As previously hypothesised, higher levels of regulation may increase the growth in housing prices but decrease the propensity of low-income areas to gentrify, which may be due to similar influences of regulation. If higher levels of regulation increase the cost of construction and renovation, this may have the twin influences of driving up home prices overall (via reduced supply and/or increased demand due to increased amenities) but simultaneously stifling the ability of neighbourhoods to engage in housing investment to the extent necessary to improve their relative standing within the region (via increased cost of remodelling/renovating/rebuilding).
Conclusion
High levels of land use regulations have been found to be a strong explanatory factor of the divergence in real estate prices in coastal cities relative to inland cities (Glaeser and Ward, 2009). Regulation typically acts as a restrictive force in the development of new housing stock, driving up the price of housing in these markets. As a result, the cost of high levels of regulation has been taken to be the reduced affordability of housing for residents. However, the forces which increase overall housing prices may have additional influences that have not been previously considered.
The ability of lower-income neighbourhoods to undergo economic transformation plays a vital role in the long-term economic health and development of a region. Revitalisation of decaying neighbourhoods may improve the economic health of struggling cities and promote economic growth in the region. This process of gentrification need not be correlated with an overall increase in house prices in the larger region. If land use regulations hinder the ability of developers and households to renovate existing housing stock and add new housing stock, this may result in higher house prices even as it reduces the likelihood that a relatively low-income neighbourhood experiences gentrification.
Previously, Kahn et al. (2010) considered the influence of regulation on overall housing prices and found that the highest growth in high-income residents occurred in the most heavily regulated areas. They argue that this suggests that regulation positively influences the propensity of an area to gentrify. However, if gentrification describes a transformation of lower-income and decaying neighbourhoods, allowing rising house prices in areas that were not lower-income to be classified as gentrified may obfuscate opposing influences. Consistent with previous findings, to our knowledge, our analysis is the first to consider the influence of land use regulation on the propensity of lower-income neighbourhoods to gentrify separately from the influence of land use regulation on housing prices. We find that higher levels of regulation intensity are associated with higher house prices along the entire income distribution with the MSA. When we define gentrification as a low-income neighbourhood which experiences above average growth in share of college graduates and above average growth in either house prices or rent prices, we find the opposite influence of regulation. A one standard deviation higher level of regulation intensity is associated with a reduction in the probability that a low-income neighbourhood will gentrify by almost four percentage points, even while increasing housing prices over 8%.
This present, and diverging, influence may contribute to our understanding of both the determinants of gentrification and the unintended consequences of higher levels of land use regulation. The determinants of gentrification in the literature have largely focused on housing and neighbourhood characteristics, but our findings suggest that regional policy factors may play a role as well. While researchers have long recognised the potential power of land use regulation as a response to gentrification, its effect on the propensity to gentrify (or not) has not been afforded as much attention in the literature. One limitation of our data is that we do not have information on the movement of households at the individual level and so we must classify based on changes in the average neighbourhood characteristics. Future research may advance our understanding with more detailed, individual-level data.
The effects of land use regulation, like many urban planning policies, are difficult to fully predict and measure. Land use regulation is often enacted so that policy makers may direct growth and development consistent with the preferences of constituents or to address physical, geographical limitations on housing development. Previous research on the relationship between land use regulation and gentrification largely focused on regulation which occurred as a consequence of gentrification rather than as a potential determinant of it (Ghose, 2004; Newman and Wyly, 2006). That land use regulation may increase house prices while also reducing the propensity of decaying neighbourhoods to experience gentrification suggests that the effects of these policies may be understated. Although our analysis suggests a strong relationship between levels of regulation and gentrification, more research is needed to establish the causal mechanism underlying this trend.
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
