Zoning ordinances and the housing market in developing countries: Evidence from Brazilian municipalities

Journal Pre-proof

Zoning Ordinances and the Housing Market in Developing Countries: Evidence from Brazilian Municipalities Ricardo Carvalho de Andrade Lima , Raul da Mota Silveira Neto PII: DOI: Reference:

S1051-1377(18)30208-0 https://doi.org/10.1016/j.jhe.2019.101653 YJHEC 101653

To appear in:

Journal of Housing Economics

Received date: Revised date: Accepted date:

14 August 2018 30 July 2019 13 September 2019

Please cite this article as: Ricardo Carvalho de Andrade Lima , Raul da Mota Silveira Neto , Zoning Ordinances and the Housing Market in Developing Countries: Evidence from Brazilian Municipalities, Journal of Housing Economics (2019), doi: https://doi.org/10.1016/j.jhe.2019.101653

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier Inc.

Zoning Ordinances and the Housing Market in Developing Countries: Evidence from Brazilian Municipalities. 

Ricardo Carvalho de Andrade Lima Catholic University of Brasilia, Brazil Brazilian Federal Public Ministry



Raul da Mota Silveira Neto Federal University of Pernambuco, Brazil

ABSTRACT: Due to recent urbanization and the increased population density around major Brazilian cities, the real estate market in the country has appreciated greatly. In addition to demand-related factors, mechanisms with the potential to limit the housing supply may have also affected the price increase in that market. The objective of this study is to investigate the impact of zoning ordinances (land-use regulations) on the average rental prices and the growth in the number of dwellings in Brazil. Through an intercity analysis and using identification methods that rely both on selection on observable and unobservable covariates, the set of evidence indicates that zoning ordinance is associated with an increment ranging from 5.3% to 6% in average rents, but does not affect the dwellings growth. Sensitivity analysis suggests that these results are not simply driven by unobservable confounders. Our evidence, thus, indicates that even being a usually well-intentioned policy, zoning ordinances tend to generate social costs that need to be taken into account in the analysis of the real-estate market in Brazil. Keywords: Land Use Restrictions, Urban Zoning, Urban Development, Public Policy. Código JEL: R31, R52, L51.

1

1. Introduction In recent decades, a sharp increase in population concentration has taken place in major Brazilian metropolitan areas, continuing the rapid urbanization process that has been occurring in Brazil over the last century. For example, the population density of Brazilian urban areas grew 16.65% between 2000 and 2010. In the new metropolitan areas of Brasilia and Goiania, the population densities increased by 23.7% and 22.7%, respectively (Brazilian Demographic Census 2010). Due to higher population density around cities and macroeconomic stability, which increased housing credit, the Brazilian real-estate market underwent a price boom. According to the Fipe Zap index, which measures the selling price of properties and rentals in the main Brazilian cities on a monthly basis, from 2010 to 2015, housing prices increased 223.5% in Sao Paulo, whereas rents increased at a rate of 100.8%. This variation was significantly higher than inflation in the same period (which reached 54.6% as measured by IPCA, the official Brazilian consumer price index). In addition to demand-related factors, supply-related factors such as the higher cost of real estate development and restrictions in urban land use may have also a connection with this great appreciation in the Brazilian real-estate market. One of the main ways to regulate and restrict urban land use in Brazil is through zoning ordinances. Zoning ordinances are usually established by local governments and can regulate both the type of use (establishing residential areas, commercial areas, industrial or mixed) and the intensity of use, restricting the size of buildings, their weight or the floor-area ratio (McDonald and McMillen 2012). Note that, in addition to traditional zoning, there are other regulations that can be used to restrict urban land use, such as rules governing septic systems, wetland regulations, green-area requirements, and others (Fischel 2004; Glaeser and Ward 2009). In Brazil, local governments (municipalities) are in charge of the adoption and the management of land-use controls. These instruments gained greater relevance in the late 80s, when the new Brazilian constitution gave more fiscal and administrative autonomy to municipalities. In addition, federal laws began to encourage local urban planning, such as Law 6766/79, which regulates the urban land parceling, and Law 10527/01, which establishes a series of guidelines for integrated urban planning in cities. In this context of the federal regulation of urban planning instruments, a large increase in the number of municipalities that have their own zoning ordinances occurred: while in 1978, only 64 municipalities had regulations for the land use, in 2013 this number had risen to 1,724 (Munic, 2013). Despite the recent increase in the adoption of these regulations and the vast empirical literature considering their effects elsewhere, few studies have investigated their impact on the Brazilian housing market and most of them focused on the impact of urban land regulation on housing informality (Lall et al. 2006; Biderman, 2008; Ávila, 2015). According to McDonald and McMillen (2012), the objective of the supporters of the law is to protect the residents' well-being from the adverse effects of industrialization, rapid and intense urbanization and high-impact buildings. Therefore, zoning would be useful to reduce the negative externalities that are associated with certain types of land use or degrees of use. However, by restricting the land use and limiting new developments, urban zoning ends up bringing changes to the land market and consequently may increase the prices of existing buildings and residences (Quigley 2007; McDonald and McMillen 2012). Thus, it creates a great incentive for homeowners to support, defend and fight for zoning laws (or other restriction instruments) with the aim of increasing the price of their houses, even in the absence of actual negative externalities (Quigley 2007; Fischel 2004). Many empirical studies investigate the impact of zoning and land use restrictions on the housing and urban land market. Quigley and Rosenthal (2005), Quigley (2007) and McDonald and McMillen (2012) made detailed reviews of these studies. Even with different empirical strategies for different metropolitan areas in the United States, these papers show a similar conclusion: the degree of land use restrictiveness is positively correlated with property price and negatively correlated with new real-estate developments (Quigley and Raphael 2005; Ihlanfeldt 2007, Glaeser and Ward 2009; Zabel and Dalton 2011). Additionally, there is evidence that lot price is higher in more regulated areas (Kok, Monkkonen and Quigley 2014) and that cities with stricter regulations suffer from a greater amplitude in the house price cycle, with more intense booms and bursts (Huang and Tang 2012). 2

The major empirical challenge of those papers is to solve the potential endogeneity in the relationship between zoning and real-estate prices. Typically, to overcome this problem, some studies add a wide range of control variables (Glaeser and Ward 2009; and Kok, Monkkonen and Quigley 2014), add fixed effects (Zabel and Dalton 2011) or employ an instrumental variable approach (Ihlanfeldt 2007). However, not all omitted variables can be measured or are fixed in time, which weakens the first two approaches. Moreover, it is extremely difficult to find an instrument that satisfies the exclusion restriction. For example, Ihlanfeldt (2007) received criticism for using past demographic characteristics as an instrument for the current level of land-use restrictiveness. It turns out that past demographics can be correlated with other current characteristics that affect property price, weakening the exclusion restriction (Glaeser and Ward 2009). In addition to this unresolved empirical difficulty, another limitation of those studies is that they focus mostly on the U.S. housing market. Due to the recent urbanization process, the real-estate market of developing economies can behave in a very divergent way (Alterman 2013), so that the conclusions of such papers cannot be easily extended to these countries. In this study, we will take advantage of the variability of the law for the different Brazilian municipalities to assess their impact on the average rental prices and on the growth of the number of dwellings. The contribution of this paper is to analyze the Brazilian case, where, unlike the United States, the urbanization process only took off in the second half of the twentieth century. In addition, this study deals with endogeneity in different ways. First, rather than trying to solve the problem thoroughly, we will evaluate how robust our evidence is in the presence of endogeneity caused by omitted variables. In this sense, in order to obtain the effect of zoning on average rental prices and on the growth of the number of dwellings, we initially employ a Propensity Score Matching (PSM) estimator, that is based on selection on observable variables, and a set of sensibility analyses proposed by Rosenbaum (2002), Inchino et al. (2008) and Oster (2016) are applied. Second, considering the possibility of selection on unobservable factors, we also apply more recent approaches based on heterocedasticity of error terms proposed by Klein and Vella (2010) and Lewbel (2012) that use, respectively, a control function and generated instrumental variables. Furthermore, another contribution of our study is to use a measure of rental prices, which is adjusted for the property’s characteristics. Our results reveal that municipalities that adopt zoning ordinances as an urban planning policy have average rental prices 5.32% to 6.03% higher than those of similar municipalities that do not. Although a causal interpretation is not fully guaranteed, the sensitivity analysis based on Rosenbaum bounds indicates that for our evidence to be invalidated there must be an unobservable variable that increases the odds of zoning adoption by 95% and is strongly correlated with the rental price. Additionally, the sensitivity analysis proposed by Inchino et al. (2008) shows that even if there are important and specific confounders, our result remains valid. The Oster (2016) procedure also points in the same direction: for our results to be invalidated, the unobservable confounders would need to be twice as important as the observable ones. Finally, these results are also supported when using the estimators of Klein and Vella (2010) and Lewbel (2012). However, we found that the zoning law is not able to affect the number of dwellings, suggesting that higher rental prices are not simply a reflection of a lower housing supply. The paper is organized as follows: in Section 2, we discuss evidence from the previous literature that assesses the impact of land use restrictions from the perspective of developing economies. In Section 3, we present the institutional background and origins of zoning in Brazilian municipalities. Section 4 details the empirical strategy. Section 5 describes the data, the covariates and the procedure for constructing a constant-quality rental price; Section 6 presents the main results and robustness tests, and Section 7 presents the final remarks.

3

2. Zoning ordinances in Developing Countries: Empirical Evidence Most of the motivations for the regulations of urban land use which are present in developed countries, generally associated with negative externalities arising from an intense use of the urban land and with the preservation of the natural environment or historical/cultural buildings, are also present in developing countries; but the frequent poverty and inequality, push to urbanize, and lack of institutional development strongly condition their application and effectiveness (Lall et al., 2007; Ávila, 2006; Brueckner and Lall, 2015). As highlighted by Alterman (2013), in developing urban environments, these policies generally present consequences for housing markets that can differ substantially from the reactions observed in developed countries’ markets. A first reason for the differences is that strict regulations with low enforcement power, which is common in developing countries and reflects their more fragile institutional environments, may receive a low level of compliance and therefore exert a weaker influence on property prices. Second, stricter regulations, by increasing the price of real-estate development and reducing the supply elasticity of available lots, can cause an additional incentive to low-income households to enter the informal housing market; this may in turn increase slum formation and the number of low-quality settlements (Ávila, 2015; Cavalcanti et al., 2018). Furthermore, as an adequate urban infrastructure (public systems of piped water and sewer, for example) and formal job locations are generally situated in central areas, the sprawl associated with urban land regulation tends to be doubly hard on poor families. Note that, given the limited availability of the urban infrastructure and public transport, urban sprawl also tends to increase the costs associated with it, generating inefficiency (Dowall and Clark, 1996; Lall et al., 2006). Finally, since most developing countries experience a steady migration to the cities, through their effect on the affordability of housing, the policies can significantly affect the population, and thus the urban housing markets, and generate unexpected effects (Schuetz et al., 2011). The available empirical evidence tends to confirm these concerns. Some works highlight the potential consequences of a weak-enforcement environment. For example, in a study conducted on Argentinian municipalities, Monkkonen and Ronconi (2013) detected that in municipalities with more stringent regulations, both the levels of compliance and lot prices were lower. Although the negative correlation between the restriction of urban land use and prices is counterintuitive in the face of the evidence from U.S. cities, the authors argue that, by stimulating the growth of low-quality houses and slums, land-use restrictions cause negative externalities in neighborhoods, and therefore the lots suffer devaluation. Additionally, in a descriptive study for the Offinso South Municipality in Ghana, Boamah (2013) showed that the land-use regulations had little practical effect, since most of the developers did not comply with the orders. According to the author, this is a common situation in African cities. Not only are the planning laws poorly enforced, they are also poorly adapted to local conditions and supervised by corrupt officials. Similar results were found by Arimah and Adeagbo (2000) for a Nigerian city. Although the occurrence of low applicability and quality in planning laws are important points in the debate about the effects of zoning in cities in developing countries, the consequences for housing and affordability, slum formation, and excessive sprawl are even more pertinent to the well-being of the citizens. In this respect, the Brazilian experience appears to support this positive association between urban land-use restrictions, the growth of the informal housing market, and urban sprawl. Biderman (2007), for example, argued that the formal housing market is very connected to the informal ones, and presented evidence that there has been an increase of the informal ones in the Brazilian municipalities that adopted zoning rules in the 1990s, results similar to those obtained more recently by Ávila (2015). Even more recently, Cavalcanti et al. (2018), through a structural-general equilibrium model, showed that much of the 1980-2010 slum growth in Brazil was connected with rural-urban migration, an increase in income levels, and the restrictiveness of the land-use regulations. In a study of Indian cities, Brueckner and Sridhar (2012) found that municipalities with stricter building height limits are more spread out. The authors showed that a unit increase in the floor-area-ratio (FAR) limit can generate a reduction of approximately 20% in the area of an average city, which generates a yearly savings in commuting costs of about 0.7% of the annual income of a typical household. 4

Interestingly, because of the regular migration to the cities, a positive association as seen between urban land regulation and housing informality can also result from inclusionary zoning, which has an important impact on housing informality. Actually, investigating Brazilian cities, Lall et al. (2007) showed that in municipalities that had adopted urban policies attempting to reduce the minimum lot size, there were increases in the formal housing supply and population growth (via migration). When the population growth exceeded the available formal housing, the policy had an undesirable effect—an increase in slum formation. Note that this result is in line with the one obtained by Schuetz et al. (2011) for the U.S. that showed that policies that aim to produce affordable houses through inclusionary zoning can also generate unexpected effects (higher prices and lower production rates). Evidence about the effects of urban land regulation on production costs and housing prices is available, but not abundant. Analyzing the government interventions (taxes, subsidies and land-use restrictions) in the Malaysian housing market, Malpezzi and Mayo (1997) showed that the costs of regulation outweigh the benefits of subsidies embedded in housing programs, leading to higher housing costs and a less elastic housing supply. In a companion paper, Bertaud and Malpezzi (2001) developed a cost-benefit framework for analyzing the land use regulations in Malaysia and concluded that regulations have led to inefficient use of land and increased the risks of housing projects and land prices. More recently, considering the case of Chinese cities, Brueckner et al. (2018) estimated the elasticity of land price with respect to the floor-area ratio (FAR) and showed that, although there is variation across cities, the effect of this specific restriction is positive and significant for both residential and commercial uses. As for the Brazilian case, although some evidence provided by Ávila (2006) suggests a positive association between urban land regulation and property value for three Brazilian cities (Curitiba, Brasília and Recife), as far as we know, the study of Dantas et al. (2018) is the unique rigorous study about the impact of urban regulations on housing price. Specifically, considering the city of Recife (the center of the fourth largest metropolitan area of Brazil), the authors investigated the price effect of a law that restricts the height of the buildings in a particular area of that city. Through a geographic-discontinuity difference-in-differences strategy, the authors showed that the ordinance caused an increase of about 7% in the price of the property located in the area. As can be seen by the above set of evidence, the majority of empirical studies focus on assessing the impact of urban land regulations on indirect outcomes such as the informal housing market or urban sprawl, and very few consider the impact of regulations on the formal housing market prices. The evidence indicates that urban land regulations appear to increase housing informality and urban sprawl in areas with poor infrastructures. However, note that, as housing informality and urban sprawl generally tend to reflect prevalent market conditions, the available evidence provide limited insights into the effects of public interventions. Regarding the Brazilian case, the above-mentioned study of Dantas et al. (2018) is the only exception. However, since that paper analyzed a particular municipal policy implemented in a small area of a city, it lacks external validity. In order to generalize the effects arising from land-use restrictions in Brazil, it is therefore necessary to adopt a broader perspective. This study aims to fill this gap. 3. Zoning and Urban Planning in Brazil: Historical and Institutional Backgrounds Brazil has experienced one of the most dramatic urbanization processes in the world: in 1940, only 31.2% of the population lived in urban areas, and in 2010 this number reached 84.4% (IBGE, 2010). In an interval of just 30 years, (from 1940 to 1970), the country went from a rural society to a predominantly urban one. This change was driven by the industrialization of the country's large urban centers and by mechanized agriculture, which together led to a reduction in rural income opportunities, and consequently an intense rural-urban migration of the labor force. Thus, from 1940 to 2010, the urban population grew at a rate of approximately 1,149%, while the rural population grew at a rate of 5.17%. This rapid and intense growth of the cities generated persistent urban problems associated with the lack of adequate infrastructure, including strong spatial segregation and a high degree of housing informality. The housing deficit in Brazil, which represents the number of housing units necessary to replace precarious or heavily cohabited shelters, is approximately 6.35 million units (Fundação João 5

Pinheiro, 2016). Additionally, according to the 2010 census, about 11.4 million Brazilians (approximately 6% of the total population) lived in slums, characterized as low-quality housing, with insecure land tenure and poor access to public infrastructures. In this context, urban planning instruments and comprehensive zoning orders began to be systematically adopted 1 by Brazilian cities in the mid-1960s. In general, the land use policies were designed more to accommodate and legitimize trends of urban occupation ex-post facto than to plan urban occupation beforehand in a structured way (Avila, 2006). Some historical examples suggest that zoning was an instrument adopted mainly in wealthier, centralized, denser urban areas where the pressures on land use were higher (Nery Júnior 2013, Borges, 2007). In these places, the households experienced negative externalities generated by rapid urbanization and attempted to limit them to protect their housing values. At the same time, developers tried to take advantage of the high population density to install commercial corridors and high-rise buildings. The combination of these opposing pressures resulted in the creation of zoning ordinances to resolve the conflicts in the central areas, while neglecting the planning of the suburban regions. As discussed by Nery Júnior (2013), the first comprehensive zoning law in São Paulo (Law 7805/72) exemplified this point, since there were successive modifications to the boundaries of the zones to accommodate demands from the wealthy citizens and construction companies. Note that this common practice of prioritizing the adoption of zoning in expensive areas may contribute to increase the price of land in those places generating an additional incentive to housing informality increase and slum formation in the suburbs, as evidenced in the studies discussed in the previous section. It should also be noted that in the period of high urban growth and the diffusion of urban planning instruments (60s to 80s), Brazil was under the regime of a military dictatorship. During this period, although municipalities were responsible for adopting land use instruments within their territorial boundaries, the federal government effectively assumed the role of formulator and coordinator. Federal Law 6,799/1979, which established general rules and guidelines for the division and subdivision of urban land, clearly illustrates the centralized nature of the period. Among other rules, the law determined a minimum lot size of 125 m², with a minimum frontage of 5 m, and a compulsory donation of 35% of the developed area for public use and open spaces. This centralization generated a similarity in the zoning ordinances of different municipalities that persisted over time. After the democratization of the country and the new Constitution in 1988, there were significant changes in the formulation of urban planning in Brazil. The 1988 Constitution gave more administrative and fiscal autonomy to the municipalities, endorsed principles of the democratic management of cities, made housing a social right, and introduced an original chapter on urban policies (Fernandes, 2007). In this new setting, municipalities obtained greater autonomy to adopt zoning orders appropriate to their local needs and to implement land use regulations more integrated with the development of their cities. The Federal Law 10257/2001 (known as the ―City Statute‖) was also an important institutional landmark that broke definitively from the long-standing tradition of urban-planning policies that simply accommodated the trends in urban occupation and served the interests of the landowners. The new law reinforced the trend of the 1988 Constitution. Thus, the urban-planning policies began to emphasize land issues related to the interests of the whole society, seeking to bring more security to tenure titles and the regularization of informal settlements (Macedo, 2008). In addition, the City Statute established principles for the creation of guidelines (synthesized in the master plans) that aimed at a sustainable balance between land use, urban mobility and environmental issues. Currently, zoning ordinances are a land use instrument used widely by Brazilian municipalities. In 1999, 1,188 cities adopted them, and in 2013, this number reached 1,724 (MUNIC, 1999 and 2013). The number of adopters grew by approximately 45% over a period of just 14 years, suggesting that zoning was gaining greater relevance as an urban-planning instrument. Brazilian municipalities tend to adopt zoning ordinances that have similar contents to each other’s. Basically, local zoning orders involve the following tools, which vary by area within each city: minimum permeability area, maximum occupancy 1

Land use regulations have existed since the end of the 19th century in Brazil. However the first regulations were oriented to very specific areas of cities, and did not cover all of the municipal territory. In this way, these instruments resembled the current urban perimeter laws and building codes.

6

rate, maximum floor-area ratio (FAR), height limits, special zones of social interest 2 (ZEIS), environmental and historical patrimony-protection boundaries, and exclusively industrial zones. The delimitation of exclusively residential areas is not a common policy instrument, and there are no distinctions between self-built housing and developer-built housing relative to allowable density. Issues related to land development, such as the determination of minimum lot size, the compulsory donation of land for public use, partitioning and merging of lots, and the respective titling processes are regulated by local laws on land subdivision (Avila, 2015). Table 1A of the Appendix presents details of zoning ordinances for a set of twenty Brazilian municipalities. It can be observed that, in spite of possible variations in stringency, the zoning ordinance contents are similar across the cities. Figure 1 presents the spatial distribution of the municipalities with zoning ordinances, considering official information form the Brazilian Institute of Geography and Statistics. Consistent with the most advanced urbanization being located in the richest regions (South and Southeast) of the country, a clear geographical pattern can be immediately observed: municipalities adopting urban land regulation are relatively more present in the South (54.35%) and Southeast (26.05%) regions than in the Northeast (9.35%), North (11.13%), or Central-West (23.76%) regions. Figure 1 - Spatial Distribution of Municipalities by Zoning Adoption

Source: Survey of basic municipal information (MUNIC), 2004.

2

This municipal instrument consists of demarcated areas with more flexible parameters to facilitate land titling for low-income households and the implementation of social-housing programs (Macedo, 2008). This type of inclusionary regulation was strongly encouraged by the City Statute.

7

The adoption of zoning ordinances in Brazil, thus, although not limited to, appears to be associated with the levels of economic development of different locations in the country. In this respect, Table 1 presents a chart of the municipalities with zoning ordinances according to their distribution by decile for income, urbanization, density, distance to the state capital (measure of city centrality), population, share of blacks, share of immigrants, and presence workers in the manufacturing sector. Table 1 – Distribution of the municipalities with zoning by decile for specific variables Decile Interval

By Income (1)

Until the 1st 1st to 2nd 2nd to 3rd 3rd to 4th 4th to 5th 5th to 6th 6th to 7th 7th to 8th 8th to 9th Above the 9th

2.36% 3.27% 5.45% 12.39% 21.60% 26.32% 31.82% 33.94% 48.64% 66.00%

By By Urbanization Density (2) (3) 13.61% 13.79% 13.48% 16.00% 16.55% 16.36% 23.23% 30.91% 39.56% 68.36%

11.80% 12.34% 17.06% 19.82% 24.18% 24.73% 22.91% 23.59% 34.00% 61.45%

By Dist. By By By By to Workers in Population Black Immigrants Capital Manuf. (5) (6) (7) (4) (8) 44.10% 16.88% 32.00% 4.00% 9.62% 23.45% 13.79% 29.95% 9.26% 8.70% 24.86% 16.18% 30.13% 15.27% 10.00% 23.13% 17.64% 32.55% 19.45% 13.09% 16.15% 16.88% 29.82% 23.96% 18.87% 21.60% 17.82% 27.04% 29.58% 24.68% 23.27% 21.27% 23.45% 36.55% 28.88% 25.95% 24.14% 17.09% 34.91% 36.12% 29.27% 35.45% 16.33% 38.29% 47.01% 20.00% 71.82% 13.45% 40.55% 54.91%

Note: the calculations were based on information from MUNIC 2004 and the Demographic Census of 2000, both collected by the Brazilian Institute of Geography and Statistics.

The figures in Table 1 indicate that the adoption rate tends to be greatest in the highest deciles for each distribution, and at the upper extremities the proportions of adopters become significantly high. For example, while in the subset of municipalities with the lowest population density (up to the first decile) zoning adoption is only 11.8%, in the subset with the highest density it is above 60% (61.45%). The only exception is the distribution of Distances to Capital (Column (4)), where, as expected, the proportion of municipalities with zoning is highest in the lowest deciles. Note also that zoning occurs more frequently in areas with a lower proportion of blacks, higher proportion of immigrants, and with higher share of workers in Manufacturing activities. It should be noted that currently all large Brazilian cities (officially defined as those with populations over 500,000) and state capitals adopt zoning ordinances as an urbanplanning instrument. Thus, the numbers in Table 1 essentially mirror the historical examples discussed earlier, which indicate that urban-planning instruments have mainly been adopted to accommodate and legitimize trends in urban growth and to mediate land-use conflicts in central, affluent areas. Nevertheless, these numbers also suggest that zoning adoption is not exclusively driven by socioeconomic characteristics. Therefore, the situation creates both an opportunity and a challenge regarding the objectives of this study. We take advantage of the variability in the land-use regulations to verify their impacts on average rents and the growth of the number of dwellings, but we have to be aware of the influence of unobservable factors. In the next section, we present our strategies. 4. Empirical Strategy 4.1 Propensity-score matching The aim of this study is to investigate the effect of municipal zoning policies on average rental prices and the growth of the number of dwellings. The great empirical challenge of this kind of analysis is to eliminate the possibility of endogeneity by reverse causality and omitted variables. Although it is assumed that property prices can be explained by zoning ordinances, the reverse is also possible: maybe zoning ordinance occurs just in areas that have more highly valued houses (Quigley, 2007). To reduce 8

these concerns, we will apply a matching-based strategy in conjunction with a series of sensitivity analyses. Hence, our goal is to make matches between treated municipalities (those that have implemented zoning ordinances) and control municipalities3 (ones that have not done so) such that the control group can be considered a counterfactual – what would have happened to the treatment group had they not adopted zoning. Consider a binary variable that describes zoning. Thus, if the municipality has implemented the ordinance, and otherwise. and describe the potential outcomes that vary according to the treatment (zoning order). In real life, only one of these outcomes can actually be observed for each municipality. However, a causal effect is defined as the difference between and . We are interested in obtaining an estimate of the average effect of a treatment on treated (ATT), as follows: (

|

)

( )

Various methods can be used to estimate the ATT. One of these is Propensity Score Matching (PSM), developed by Rosenbaum and Rubin (1983). The propensity score is the probability of receiving a treatment (in our case, implementing a zoning ordinance) given the observed covariates (defined as the vector), i.e. ( ) ( | ). The key identification assumption for the ATT is that and are ( ) . In other words, , independent of , given | ( ) . This is known as Conditional Independence Assumption (CIA). Additionally, it is necessary that each treated unit have a matching control unit with similar . This assumption is known as the overlap condition, and can be written as: ( | ) . Finally, assuming that the CIA and the overlap condition hold, the impact of a zoning ordinance on the average rental price of a city is given by: (

|

) , ( , ( | ( )

| ( ) )) ( | ( )

)|

-

( )

That is, under these assumptions, we can use the outcomes of the matched control municipalities as a counterfactual for the estimation of ATT. A two-step procedure is employed to obtain the empirical counterpart of (2). Firstly, a probabilistic regression model is used to estimate the propensity scores ( ). Secondly, based on algorithms, we find treated and non-treated municipalities with similar propensity scores. For robustness purposes, we will apply three types of matching algorithms commonly used in the empirical literature: nearest neighbor, radius matching and kernel matching. Caliendo and Kopeinig (2008) described these algorithms in detail and discussed their respective advantages and disadvantages (in terms of bias and efficiency). 4.2 The plausibility of the CIA and sensitivity analysis The CIA assumption implies that the adoption of zoning is completely explained by observable variables. That is, the matching procedure is constructed of covariates that all generate differences in the distribution of treated and non-treated units, making the probability of treatment equal for both members in a pair, as if there had been a random assignment (Caliendo and Kopeinig 2008). In our case, as shown in section 3, certainly, traditional observable socioeconomic variables play an important role in determining the adoption of zoning ordinance by Brazilian municipalities. But, it is essential that the vector contain detailed information about the relevant characteristics of the municipalities, which is possible using the Brazilian Demographic Census and the Survey of Basic Municipal Information (MUNIC). We will describe the variables used to obtain the propensity scores in the next subsection. However, even with detailed information on the municipalities, there is a possibility that the CIA is not valid due to the existence of unobservable variables (confounders) that affect both the adoption of zoning and the outcome variable. The existence of confounders can make the ATT estimator biased and therefore hardly credible – this is the weakness of the PSM strategy. Although the CIA is, by 3

In Figure 1A of the Appendix, we show the distribution of municipalities with zoning ordinances.

9

construction, a non-testable assumption, and the absence of confounders is uncertain, we can, through a sensitivity analysis, either evaluate the strength that an omitted variable must have to invalidate our ATT estimate, or simulate potential confounders. Accordingly, we will use three different kinds of sensitivity analysis (Rosenbaum 2002; Inchino et al. 2008 and Oster 2016) to verify the robustness of the ATT when the Conditional Independence Assumption fails. Firstly, we will apply the sensitivity analysis developed by Rosenbaum (2002), which provides an indication of the magnitude of the omitted variable bias that would be necessary to invalidate the associations initially observed by the estimated ATT. This method is based on the parameter , which measures the difference in the odds of receiving the treatment between observations with the same observable characteristics. Therefore, a random experiment ensures , and when that grows, the experiment becomes less random. For example, in an observational study, if , then one of the units is three times more likely to receive the treatment due to unobservable factors, since the treated and untreated units are identical regarding the observable variables. Thus, the basic procedure for the Rosenbaum sensitivity analysis is to select a series of values and, for each one, calculate the bound of the significance level (p-values) for the ATT in the case of an endogenous treatment selection (Diprete and Gangl 2004). In addition, we will also apply the sensitivity analysis proposed by Ichino et al. (2008). This strategy aims to verify the bias of the estimated ATT when the CIA fails in some specific and relevant way. It is first necessary to establish values for the parameters that characterize the distribution of a specific confounding factor . It is assumed that this variable is binary, independent and uniformly distributed. Thus, the distribution of is completely characterized by the choice of four parameters, denoted by : ( | ( ) ( | ( ) ̅) ̅) (3) where * +, is an indication function and ̅ is the mean of the outcome variable. A value for is determined for each treated or control observation. The simulated is considered as a new observable covariate and is then included in the set of variables used for the matching process. Finally, the ATT is computed again with the inclusion of the confounder . Therefore, by modifying the parameters that characterize the distribution of , we can evaluate the robustness of the ATT against different assumptions related to the nature of the confounding factor (Nannicini, 2007). We will follow the simulation exercises proposed by Ichino et al. (2008) and choose in order to make the confounder distribution similar to the empirical distribution of the important covariates. Furthermore, the simulated confounders will be threatening to the baseline ATT: they will have a positive correlation with the treatment status variable ( ) and the outcome of the non-treated units. The third method we will apply is the sensitivity analysis developed recently by Oster (2016). Unlike the methods of Rosenbaum (2002) and Ichino et al. (2008), this method is not grounded in PSM and can be applied to OLS estimates with continuous treatment variables. Following the idea proposed by Altonji et al. (2005) that the reduction in selection bias resulting from an additional unobservable variable would be similar to that from an additional observable variable, Oster (2016) developed a consistent estimator of the treatment effect adjusted by the omitted variable bias, considering the assumption that there is a proportional selection between observable and unobservable variables: (

) ( )

(

) ( )

( )

That is, the correlation between the observable variables (X) and the treatment variable (Z) is proportional to the correlation between the unobservable variables (U), and the treatment variable; and the parameter is the coefficient of proportionality. Before implementing the estimator, Oster (2016) defined the coefficient resulting from the simple regression of on as and the corresponding R-squared as . Similarly, the coefficient from the intermediary regression of on and is denoted by , while the Rsquared is given by . Finally, the R-squared of a hypothetical regression of on and is denoted 10

as . These parameters are used in the following relationship, which allows us to approximate the adjustment bias of the OLS regression: (

)(

)

(

)

( )

From expression (5) it is possible to obtain an unbiased estimator for the treatment effect by calculating the bias given by the second term on the right side of the expression. Although and are unknown parameters, Oster (2016) proposed some upper bounds. It is claimed that the selection of unobservables should not exceed that of observables, so the maximum value of tends to be less than or equal to one. In addition, Oster (2016) showed that the value of 1.3 is a suitable upper limit for . To check the robustness in a practical way, we can construct a bounding set as , ( )] and verify if it contains coefficients that move toward zero. Another way to check robustness is to find the value of that would produce , corresponding to the degree of selection of unobservables relative to that of observables that would be required to fully explain the treatment effect. 4.3 Selection on non-observables and identification through heteroscedasticity In order to obtain an estimate of the impact of the zoning on the property price, the above approaches rely on CIA, or try to measure the influence of the selection of unobservables on the estimate. However, we also apply recent strategies that rely on heteroscedasticity in the error terms to identify this impact. Specifically, in our situation of no exogenous excluded instruments, we consider both the Lewbel (2012, 2018) and Klein and Vella (2010) approaches, which use presence of heteroscedasticity in the error terms to generate, respectively, instruments and a control function for identifying when the treatment variable is endogenous. The two approaches both rely on the presence of heteroscedasticity in the error terms, but use it differently to identify the effect of an endogenous regressor. To summarize the strategies, consider the triangular system given by equations (6) and (7): (6) (7) where y is the dependent variable (property price), z is a variable measuring the intervention (zoning), X is a set of exogenous regressors, and and are corresponding error terms. Note that we are interested in knowing the value of ; but traditional OLS and 2SLS approaches cannot be applied in the presence of selection on unobservable with no exogenous excluded instruments (equation (7)), because and tend to be correlated. Lewbel (2012, 2018) proposed using instruments generated from the error term and some additional conditions to obtain the effect of z on y. More specifically, his estimator is based on the existence of a set of exogenous variables W (may be the set of variables X or a subset of it) and the following conditions: (

)

(8)

(

)

(9)

((

)

((

) )

)

(10) (11)

11

Note that the first two conditions only require the exogeneity of the set of regressors in equations (6) and (7); conditions (10) and (11) are less regular: they require, respectively, that the covariance between the product of the errors and W be zero, and that the error tem be heteroscedastic in terms of ) as an W; but note that these two last conditions also imply that we can use the product ( instrument for z (the intervention variable) in equation (7). Thus, it is possible to obtain an estimate for ̅ ) ̂ as the instrument (two-stage least squares) and by using the Generalized both by using ( Method of Moments (GMM). Klein and Vella (2010) also use the presence of heteroscedasticity in the error terms for identification when no exogenous excluded instruments are available, but they employ it to build a nonlinear control function. To summarize the approach, consider the presence of endogeneity associated with the z variable in equations (6) and (7) and write the association between the error terms as , ( ) where and measures the degree of association between the terms (a measure of the endogeneity). A naive control function can be obtained by substituting this value for in equation (1): (12) Because is a perfect combination of z and X, however, an OLS estimation is not possible. Nevertheless, it is possible to rewrite as: (

) (

(

where

(13)

) )

is the correlation coefficient between

and

, and

and

are the respective

standard deviations. By assuming a correlation between the error terms and a specific hypothesis about their heteroscedastic structure, Klein and Vella (2010) constructed a control function and a reliable estimator for Specifically, assume that the error terms contain a multiplicative structure composed of the heteroscedastic parts ( ) and ( ) and the homoscedastic parts and in the forms: (

)

(14)

(

)

(15)

( ) and ( ) are heteroscedastic functions in X, where , and the error correlation ( ) is constant. These two hypotheses allow equation (12) to be rewritten as:

As

(

)

(

)

(

)

(

)

(16)

varies non-linearly across the variables X, and

The new control function

(

) (

)

is constant, the regressors are not collinear.

provides a reliable estimate of the parameters of equation (16) .

Implementation of the Klein and Vella (2010) estimator thus needs an estimate for the control ( ) by function. This is acquired in three steps: i) estimate ( ̂ ) ii) obtain an estimate for ( ) regressing ̂ on X; and iii) use the estimates for and , and non-linear minimization to obtain ( ) the estimate for . 5. Variables and Data

12

As previously discussed, our empirical strategy will take advantage of the fact that some municipalities have adopted zoning ordinances while others have not. Information on the urban policies that each city adopts is obtained from the survey of Basic Municipal Information (MUNIC), a database updated annually by the Brazilian Institute of Geography and Statistics (IBGE) that gathers information on the dynamics, structure and functioning of local institutions. It is important to note that the survey collect information about the laws and regulations that exist in all Brazilian municipalities, but it does not go deep into the specificities of each particular law. Thus, our treatment variable is dichotomous: it takes the value of one if the municipality had a zoning law in 20044, and takes the value of zero otherwise. Since zoning ordinances may involve a set policy instruments with wide range of specific rules and parameters, this is a limitation of our treatment variable and a consequence of our general and aggregated data. However, in order to atenuate this fragility and verify the robustness of the results, we also use a treatment intensity variable. This variable will be calculated by just summing up the number of specific land-use regulations adopted by municipalities. Regarding the outcome variables, we will use two throughout the paper: the average rental prices (measured in 2010) and the growth of formal houses in the 2000-2010 period. In relation to the first variable, we follow the correction proposed by Quigley and Raphael (2005), which seeks to adjust the rental prices in relation to the properties’ characteristics in order to obtain average rents that reflect only the cities’ characteristics. Therefore, the first step is to estimate a hedonic regression for each Brazilian municipality (5,507 separate regressions), where the dependent variable is the rental price and the independent variables are the property characteristics available from the Demographic Census of 2010: number of bedrooms, bathrooms, other rooms, and a dummy that takes the value of one if the house is made of bricks, and zero otherwise (Seabra and Azzoni 2015). The dwelling is the observation unit in this analysis. In the Appendix, Table 2A summarizes the means of the estimated coefficients. The last step is to combine the estimated coefficients of each city with the average hedonic characteristics of Brazil in order to estimate the average rental price with constant quality. This will be our main outcome variable. We will follow the recommendation of Caliendo and Kopeinig (2008), and add only the covariates that simultaneously affect the outcomes and zoning adoption. To ensure that these variables are not directly affected by the treatment, they will be fixed in time or measured before the treatment (year 2000). The outcomes and control variables were obtained from the Brazilian Demographic Censuses of 2000 and 2010. It is necessary to point out that since our outcome variable was measured only in 2010, it is impossible to build a panel database. The choice of the control variables that will be used to match the municipalities that adopt zoning with those that do not was based on the recent empirical literature (e.g. Ihlanfeldt 2007, Glaeser and Ward 2009; Zabel and Dalton 2011; Monkkonen and Ronconi 2013; Kok, Monkkonen and Quigley 2014). We split the covariates into six not mutually exclusive sets: i) demand factors: the characteristics that drive the intense use of urban land, such as population density, population growth (1991-2000) and per capita income; ii) externality factors: activities that inflict damage on the natural and social environments of cities, causing greater pressure for the adoption of urban planning laws; these include the urbanization and industrialization rates; iii) demographics (since zoning orders are set by municipal legislators, the local citizens have a strong influence on the direction of urban policy, justifying the inclusion of demographic variables): average years of schooling, percentage of blacks, working age, immigrants, voter turnout, and percentage of homeowners; iv) urban infrastructure: as discussed in Section 2, policies that restrict urban land use can simply be a way to avoid high densities in places with poor urban infrastructure; thus we include the following variables in the control set: the percentages of households with sewer, electricity and running water; v) fiscal conditions: the fiscal dependence on property tax collection can encourage the zoning adoption to preserve the revenues or enhance them; thus we include 4

We opted to choose this particular year because the information available in the MUNIC 2004 had a higher degree of compliance with the local laws because of personal inspections. Information about zoning was also collected from MUNIC in 1999, 2005 and 2009. However, for these other years there were some degrees of disagreement with the local zoning orders. We evaluated the robustness of the results to the changes made by 1999, 2005 and 2009 MUNIC Surveys and the estimates were relatively similar. These estimates can be made available upon request.

13

the share of property taxes 5 in relation to total municipal revenues to capture this effect and vi) environmental and fixed factors. Urban planning may also be established to protect cities with a natural or cultural heritage from real-estate development. Cities that have a coastline, nature-conservation site or are historical-cultural capitals may be more likely to adopt zoning orders. Therefore, we will use the year the city was founded as a proxy for the historical-cultural capital, coastline and conservation-area dummies; and finally, we include in Group v the distance to the nearest capital city. In addition to these sets of covariates, we also include state fixed-effects in the propensity-score estimation. The purpose of using this broad set of variables to characterize the municipalities is to make the distribution of the treated and untreated municipalities as similar as possible. However, it is very likely that there are unobservables that affect the zoning assignments. Our argument in this study is that these variables are not important enough to invalidate our results. 6. Results In this section, we will present and discuss the results obtained by applying the methodology described above. Firstly, we will evaluate the matching quality and estimate the ATT considering different algorithms (Subsection 6.1); then we will conduct sensitivity analyses (Subsection 6.2); and finally, we will present the results of identification through heteroscedasticity (Subsection 6.3). 6.1 Baseline results Table 2 shows the means of the dependent variables and covariates by treatment status before and after the application of the matching procedure. Initially, we can see that cities that adopted zoning orders were substantially different from those that did not. For example, regulated cities had greater rental prices, higher house growth, higher per capita income, higher population density, were more urbanized, supplied a better infrastructure, were closer to the state capital, and also had more natural amenities. It is noteworthy that the mean difference for all observable characteristics between the two types of cities is large and statistically significant. This shows that a simple mean difference for the rental prices between municipalities with and without zoning is useless for impact evaluation. Columns (4) and (5) of Table 2 show the variables’ means after applying a nearest-neighbor matching algorithm with replacement and imposing a common support. By applying this expedient, we found that 9 treated municipalities (0.2% of the total group) were outside of the common support region, and therefore were excluded. We can see that after the matching, the characteristics of both kinds of municipalities become so similar that there are no statistically significant differences 6. Figure 1A of the appendix shows the distribution of propensity scores across treated and controls groups; it is observed that off support municipalities are characterized by those with very high probabilities of zoning adoption. Table 2 – Mean characteristics of cities that adopt zoning (treated) and that do not (control) Variable Name

Average Rental Prices House Growth Log Income (Per Capita) Log Population Density Population Growth (1991-2000) % Workers in Industrial Sector

(1) Treated 5.652 0.336 6.104 3.866 0.130 0.213

Before Matching (2) (3) Control Difference 5.220 0.432*** 0.272 0.064*** 5.511 0.593*** 2.908 0.958*** 0.077 0.053*** 0.137 0.077***

(4) Treated 5.648 0.343 6.097 3.841 0.130 0.214

After Matching (5) (6) Control Difference 5.595 0.054 0.346 -0.004 6.077 0.020 3.774 0.068 0.150 -0.020 0.213 0.000

5

In Brazil, the property tax is defined as Imposto Predial e Territorial Urbano (IPTU). For a comprehensive discussion about this property tax, see Afonso et al. (2013). 6

We also use 27 state dummies to build the matching sample of Table 1. The results have been omitted for space constraints, but reveal that the mean difference (with respect to these variables) is not statistically significant between the two kinds of municipalities.

14

Urbanization Rate % Immigrants % Homeowners % Working-Age population % Black Population % Households with PipeWater % Households with Electricity % Households with Sewage Avg. Schooling of 25 years old Voter Turnout in Mun. Elections % Local Property Tax Year of Mun. Foundation Distance to the State Capital Coastal Municipality (0/1) Conservational Area (0/1) Number of Observations

0.715 0.436 0.752 0.433 0.047 0.691 0.952 0.302 5.121 0.874 0.033 1957.4 233.52 0.088 0.264 1386

0.546 0.338 0.774 0.376 0.062 0.544 0.841 0.204 3.673 0.865 0.008 1963.3 259.86 0.036 0.081 4112

0.169*** 0.098*** -0.022*** 0.057*** -0.016*** 0.147*** 0.111*** 0.099*** 1.448*** 0.010*** 0.026*** -5.900*** -26.34*** 0.053*** 0.183*** -

0.713 0.435 0.752 0.433 0.047 0.690 0.951 0.300 5.099 0.875 0.033 1957.5 234.69 0.079 0.260 1377

0.708 0.440 0.750 0.431 0.049 0.687 0.954 0.304 5.078 0.876 0.039 1960.5 231.87 0.097 0.261 709

0.005 -0.005 0.002 0.002 -0.002 0.003 -0.003 -0.004 0.021 -0.001 -0.006 -3.000 2.820 -0.018 -0.001 -

Note: For the described matching, we used the nearest-neighbor propensity score with replacement and imposing a common support area. The average rental prices were corrected for their hedonic attributes (Subsection 5.2). *** p< 0.01, ** p < 0.05, *** p < 0.1.

Table 3 summarizes the results of the propensity-score estimation for unmatched and matched samples. We used a logit probability model with state fixed-effects. Although the choice of a probabilistic model is not critical to the ATT estimation, when compared to the probit function, the logit function has denser mass in the bounds (Caliendo and Kopenig 2008). In columns (1) and (2) we note that most of the estimated coefficients take the expected signs based on what was discussed above and are consistent with the descriptive statistics of Table 2 and the discussion of Section 3. Cities with higher incomes and moreeducated individuals were more likely to adopt zoning as an urban planning law, which conforms with Gyourko, Saiz, and Summers (2008). Areas that had a greater demand for land use (measured by population density), were more urbanized, had a higher dependence on property taxes, or a precarious urban infrastructure also adopted zoning more readily. This last fact suggests that Brazilian cities may be using the zoning law as a legal instrument to limit development in areas with poor infrastructures. Additionally, we can observe a positive correlation between the use of zoning and the existence of natural amenities. In columns (3) and (4) of Table 3, the logit model was estimated considering only the sample of matched municipalities. The results show that the matching strategy performs well in simulating the randomness of the treatment variable, since the degree of explanation of the covariates based on zoning drops significantly, such that R² reaches only 2.4%. At the bottom of Table 3, we also report other measures of matching quality suggested by Rosenbaum and Rubin (1983). Note that after the matching, the mean standardized bias dropped by about 76% (from 33.3 to 3.5). These results support those obtained in Table 2, indicating that, in fact, it is possible to build groups of reasonably similar municipalities. Table 3 - Propensity score estimation: determinants of zoning Dependent Variable: Zoning Log Income (Per Capita) Log Population Density Population Growth (1991-2000) % Workers in Industrial Sector Urbanization Rate % Immigrants % Homeowners % Working-Age population % Black Population

Unmatched Sample (1) (2) Coefficient Std. Err. 0.669*** 0.220 0.293*** 0.054 0.230 0.213 0.617 0.534 0.581* 0.362 -0.285 0.463 -0.675 0.554 -0.661 1.043 0.144 1.194

Matched Sample (3) (4) Coefficient Std. Err. 0.570 0.439 0.204 0.135 -0.254 0.294 -0.740 0.767 0.565 0.511 -0.088 0.693 -0.098 0.861 -0.929 1.571 -0.962 2.164 15

% Households with PipeWater % Households with Electricity % Households with Sewage Avg. Schooling of 25 years old Voter Turnout in Mun. Elections % Local Property Tax Year of Mun. Foundation Distance to the State Capital Coastal Municipality (0/1) Conservational Area (0/1) Constant State Fixed-Effects N. Observations Pseudo R² Mean Bias Median Bias

-0.120 -1.197** -0.278 0.675*** -2.650*** 3.524** 0.001*** -0.001 0.511*** 0.558*** -4.425*** Yes 5498 0.32 33.3 24

0.299 0.649 0.266 0.093 0.846 1.977 0.000 0.003 0.204 0.114 5.405

-0.193 -1.563* -0.742 -0.086 0.498 -0.984 0.000 -0.008*** -0.701*** -0.036 13.627*** Yes 2086 0.024 3.5 1.7

0.433 1.009 0.583 0.138 1.307 2.010 0.001 0.004 0.342 0.166 7.725

Note: The dependent variable is an indicator of zoned municipalities. The estimation was carried out with a logit function. *** p< 0.01, ** p < 0.05, *** p < 0.1.

After obtaining the propensity scores, we estimate the impact of zoning on the average rental prices and on the growth of the number of dwellings using four different matching algorithms and the basic OLS. For the rental prices, Table 4, Panel A, shows the estimated ATT and the corresponding measures of matching quality described above. The estimated ATTs vary from 0.053 to 0.060, indicating that cities that adopted zoning as an urban planning policy had an average rent 5% higher than the nonadopters. This effect is substantial, since it corresponds to about 12.47% of the unconditional difference in the average rental between the two kinds of municipalities. Notice that our estimates are in line with the evidence for U.S cities, but with present smaller magnitudes. For example, Ihlanfeldt (2007) showed that land use restrictions generate a price effect of around 15.5% in cities of Florida state and Zabel and Dalton (2011) evidenced that that minimum lot size restrictions (MLRs) impacted on Boston house prices up to 20%. As discussed in section 2, developing countries have an urban structure marked by spatial segregation and a large informal housing market. These urban environments tend to favor non-compliance with local ordinances and urban planning instruments. Actually, our results suggest that, in the Brazilian case, these characteristics tend to attenuate the price effect of land use restrictions, which may explain the smaller magnitude of the estimates. Table 4 – The impact of zoning on average rental prices and on growth of the number of dwellings Panel A: Average Rental Prices

(1)

OLS 0.0542*** (0.009) N. Observations 5498 Bias After Matching Pseudo R² Panel B dwellings growth (1) ATT

ATT

N. Observations Bias After Matching

OLS 0.0306 (0.012) 5498 -

(2) Nearest Neighbor 0.0532*** (0.023) [0.016] 2086 3.5 0.024 (2) Nearest Neighbor -0.0036 (0.037) [0.035] 1905 3.7

(3) 10 Nearest Neighbor 0.0549*** (0.012) [0.012] 3752 2.8 0.014 (3) 10 Nearest Neighbor 0.0020 (0.028) [0.031] 3437 2.4

(4)

(5)

Radius 0.0548*** (0.022) [0.010] 5477 2.9 0.013 (4)

Kernel 0.0603*** (0.021) [0.008] 5477 3.1 0.013 (5)

Radius 0.0091 (0.028) [0.026] 5127 2.7

Kernel 0.0098 (0.026) [0.0264] 5127 2.6 16

Pseudo R²

-

0.019

0.009

0.01

0.008

Note: Analytical and bootstrap standard errors (200 interactions) are reported in round and square brackets, respectively. The radius matching uses a 0.01 caliper, and the kernel matching is constructed using an Epanechnikov function. *** p < 0.01, ** p < 0.05, *** p < 0.1. The average rental prices were corrected for their hedonic attributes.

Panel B of Table 4 shows that, regardless of the matching algorithm used, the zoning laws do not affect the growth of the number of dwellings. Therefore, we cannot associate the higher rental prices simply with a reduction in housing supply. Mechanisms related to the demand side can better explain the causes of higher rental prices in regulated cities. For example, households may prefer to live in places with land use controls in order to avoid negative externalities associated with higher density, such as congestion and noise. This is consistent, for example, with the results obtained by Dantas et al. (2018) in the particular case of the city of Recife. In addition, zoning can function as a kind of insurance against future land-use conflicts (Zhou, McMillen and McDonald, 2008). In both cases the local community is willing to pay the premium for neighborhood protection provided by zoning ordinances, which translates into higher prices. In relation to the growth of the number of dwellings, the increase in the expected return of real estate assets generated by zoning can encourage new construction. In contrast, the needs to comply with density parameters and building codes tend to increase development costs and land prices, discouraging construction (Bertaud and Malpezzi, 2001; Quigley, 2007). Our results suggest that these two effects annul each another. While understanding the mechanisms underlying the relationship between zoning and housing outcomes is important, the aggregated nature of our database does not allow making definitive inferences about them. In Table 5, we check the robustness of our results using alternatives measures of land-use restrictions and considering a treatment variable that captures the intensity of regulation. In column (1) to (4), we show the ATT estimates of the following specific land-use regulations commonly adopted by Brazilian municipalities: (1) Minimum lot size (MLS) higher than that recommended by federal law 6799/79 (greater than 125 m²), (2) Urban growth boundary regulation, (3) Land subdivision regulation, and (4) building codes. All these treatment variables are binary and assume the value of one when the municipality adopts it. In column (5) and column (6), we employ a continuous treatment variable (regulatory index) to capture the stringency of all local land-use controls. This index is constructed by summing up the number of specific regulations adopted by the municipalities and it was commonly used as a measure of land-use restrictiveness in previous studies (see Ihlanfeldt 2007 and Quigley and Raphael 2005, for examples). The basic idea is that municipalities with larger numbers of regulations tend to have a more stringent regulatory environment. Table 5 – The impact of alternative measures of land-use restrictions on average rental prices Average rental prices

(1)

(2)

(3)

(4)

(5)

(6)

Treatment Definition

Higher MLS

Urban Growth Boundary

Land Subdivision

Building Code

Regulatory Index I

Regulatory Index II

ATT

0.0381**

-0.0359

0.0282**

0.0261

0.0227***

0.0276***

(0.0211)

(0.0346)

(0.0176)

(0.0212)

(0.0035)

(0.0044)

N. Treated

956

2974

662

1453

4802

2681

N. Observations

1566

3,506

2,683

2,317

5498

5498

Note: Standard errors are reported in parenthesis. *** p < 0.01, ** p < 0.05, *** p < 0.1. Coefficients from Column (1) to (4) are estimated using a nearest-neighbor propensity score matching and before these estimations, we dropped the municipalities that already had some zoning ordinances. The coefficients of columns (5) and (6) are estimated by ordinary least squares. Estimations consider all set of control variables.

The results of Table 5 show that the average rental prices are higher in municipalities that adopt higher MLS and in those that adopt land subdivision requirements (columns (1) and (3)). However, 17

building codes and urban growth boundaries are policy instruments that do not seem to affect the housing rental prices (columns (2) and (4)). These differences may be associated with the wider character of these last two tools. Column (5) shows the regulatory index summing up all set of land-use controls and column (6) shows the regulatory index summing up only those that effectively influence rental prices (zoning ordinance, higher MLS and land subdivision law). The coefficients in columns (5) and (6) indicate that higher levels of regulation stringency are associated with higher rental prices. These last results are in line with the previous one obtained using our general zoning variable as measure of land-use restrictions. 6.2 Sensitivity analyses In order to investigate how reliable our estimates are, we also consider how the estimated ATTs change if the Common Independence Assumption (CIA) fails. We recognize that, even including important covariates in the propensity-score estimation, there are potentially unobservable variables that affect both the outcome variable and zoning adoption. The degree of political power of homeowners is a classical confounder. According to the Homevoter Hypothesis developed by Fischel (2001), the home is the main financial asset for most people; thus, homeowners try to influence local politics in order to boost the value of their residences and prevent depreciation. Thus, municipalities with a higher proportion of homeowners with political power (more homevoters) tend to be more favorable to the construction of parks, open spaces, improvements in public services, and the adoption of stringent zoning ordinances. In the same way, interest groups such as developers, renters, and environmentalists can also play a central role. As evidenced by Solé-Ollé and Viladecans-Marsal (2013), the political party controlling the local government makes a great difference in the land use policies that are adopted. There is also evidence suggesting that local residents’ political ideologies may influence the decisions about housing policies (Kahn, 2011). Besides these political variables, municipal differences related to city management and urban planning expectations are also important confounders, especially in the case of Brazil, where there are strong spatial heterogeneities in local institutions. These characteristics include, for example, the degree of concern of the local governments about urban planning issues, the decision to adopt land use controls to just accommodate or to effectively plan the urban development, and the administrative capacity to enforce regulations. Thus, in municipalities with better urban planning management and a greater capacity to apply regulations, the adoption of zoning tends to be more likely. At the same time, the quality of the urban environment also tends to be better, positively impacting the housing market. In the next two subsections of the paper, we will investigate issues related to the existence of unobserved variables and failures in the CIA. Firstly, we employ the Rosenbaum (2002) bounds approach to measure the sensitivity of the baseline specification when our key assumption (CIA) is relaxed. Thus, we use different values of – which measures the difference in the odds of receiving the treatment between observations with the same observable characteristics – to verify changes in the inference due to the existence of unobservable confounders. Table 6 shows the results for ranging from 1 to 2 and the corresponding p-value bounds. It can be seen that up to , the estimated ATT remains statistically significant at the usual level of 5%. This indicates that a city may have the chance to adopt zoning 90% higher than the others due to unobservable characteristics, yet our initial evidence remains reliable. Only when is the ATT no longer significant at the usual levels. Thus, for higher rental prices to be associated with an unobservable variable instead of a zoning order, it is necessary that this confounder increase the odds of a municipality implementing zoning by 95%. Thus, the results of Table 4 cannot be explained by a moderate omitted-variable bias. Table 6 - Sensitivity analysis I: Rosenbaum bounds Gamma 1

Average Rental Prices p-crit+ p-critGamma 0.000 0.000 1.55

p-crit+ 0.000

p-crit0.000 18

1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4 1.45

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

1.5

0.000

0.000

1.6 1.65 1.7 1.75 1.8 1.85 1.9 1.95 2

0.000 0.001 0.000 0.001 0.003 0.010 0.026 0.060 0.119

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

Note: the parameter measures the odds of receiving the treatment considering municipalities with the same observable characteristics. The method is based on nearest-neighbor matching without replacement.

The results of the sensitivity analysis proposed by Inchino et al. (2008) can be seen in Table 7. We chose values of in order to simulate confounders with distributions analogous to those of the observed covariates, and that had a positive correlation with the treatment and outcome variables 7. This way, we simulated unobservables with a similar distribution of the following variables: per capita income, population density, urbanization rate, average schooling, and percentage of local tax property, conservation area and coastal municipality dummy. It is worth mentioning that the Inchino et al. (2008) procedure assumes that the distribution of the simulated confounder is discrete 8 . Thus, to mimic the continuous covariates, we used the following criteria: the simulated binary variable took the value of one if its value was above the mean (or median) of the distribution, and zero otherwise. Table 7 - Sensitivity analysis II: simulated confounders Baseline Neutral Confound to Mimic Income (Mean) Income (Median) Average Schooling (Mean) Average Schooling (Median) Density (Mean) Density (Median) Urbanization Rate (Mean) Urbanization Rate (Median) % Local Tax Property (Mean) % Local Tax Property (Median) Coastal City Conservational Area

p11 p10 p01 p00 0 0 0 0 0.5 0.5 0.5 0.5

Out. Eff. Selec. Eff. 1.009 1.006

ATT 0.053 0.053

SE 0.021*** 0.024**

0.94 0.91 0.92 0.92 0.71 0.7 0.77 0.76 0.65 0.92 0.09 0.29

17.611 17.787 15.123 15.268 0.989 0.982 3.059 3.117 29.656 11.556 1.246 2.952

0.041 0.042 0.040 0.041 0.057 0.060 0.050 0.051 0.042 0.039 0.057 0.053

0.009*** 0.011*** 0.009*** 0.010*** 0.012*** 0.010*** 0.011*** 0.012** 0.01*** 0.01*** 0.01*** 0.012***

0.44 0.35 0.43 0.42 0.49 0.49 0.46 0.45 0.08 0.5 0.08 0.11

0.76 0.71 0.71 0.69 0.45 0.44 0.59 0.58 0.29 0.66 0.04 0.12

0.16 0.13 0.14 0.13 0.45 0.45 0.32 0.31 0.02 5.122 0.03 0.05

3.805 3.376 3.997 4.144 2.325 2.274 2.197 2.197 4.364 5.122 2.499 2.969

Note: The method is based on nearest-neighbor matching with 200 interactions. The outcome effect measures the impact of the unobserved variable on the untreated outcome, while the selection effect measures the impact of the unobserved variable on treatment assignment. *** p < 0.01, ** p < 0.05, *** p < 0.1.

7

This would be an unobserved variable with potential to explain the positive impact of zoning on rental prices.

8

Through Monte Carlo simulations, the authors concluded that the assumption that the confounder assumes a discrete distribution, when in fact it is continuous, is not able to generate erroneous conclusions regarding the sensibility of the ATT. Moreover, Wang & Krieger (2006) showed that binary variables are more of a threat to causal inference than continuous ones.

19

From Table 7, it can be seen that when the CIA does not hold due to the existence of a specific unobservable confounder, the magnitude of the ATT changes but remains qualitatively important and statistically significant. For example, even if we are omitting an unobservable that is very important to the adoption of zoning, such as per capita income or population density (Alterman 2013, Gyourko, Saiz and Summers 2008), our initial evidence remains valid. Finally, Table 8 presents the results of the Oster (2016) procedure for different values of . We estimated two types of OLS specification to generate movements in the R-squared in order to calculate the selection bias given by equation (5): the partially controlled model, which includes only the treatment variable and state fixed-effects, and the fully controlled model, which includes the full set of controls. The first line of Table 8 shows the bounding set considering equal selection between the observable and unobservable variables ( ), and the second line shows the values that would be necessary to invalidate the conclusion that zoning affects the rental prices. Table 8 - Sensitivity analysis III: Oster’s procedure

Bounding set with

Value of

for

[0.048, 0.054]

[0.043, 0.054]

[0.038, 0.054]

[0.032, 0.054]

[0.025, 0.054]

6.419

3.661

2.561

1.968

1.599

Note: The method is based on OLS estimates. For the partially controlled model, and , and for the fully controlled model, and .

Firstly, we noted that for any maximum value of R-squared, the bounding set does not include zero. Therefore, the ATT remains valid in all situations assuming that the unobservable confounders are as important for the selection as the observable ones. Moreover, even considering the threshold of suggested by Oster (2016), which is equivalent to , it would be necessary that the selection on the unobservable was 1.968 times higher than that of the observables to invalidate our conclusions. This evidence reinforces the conclusion reached by applying the Rosenbaum bounds and simulated confounders – that the positive impact of zoning on the average rental price does not seem to be driven by some omitted variable. 6.3 Identification through heteroscedasticity As a final check of our results, Table 9 presents estimates of the impact of zoning on the average rental prices for Brazilian municipalities using the Lewbel (2012) and Klein and Vella (2010) estimators, which rely on heteroscedasticity in the error terms. To make comparisons easier, OLS estimative is also presented. For the Lewbel (2012) approach, we generated instrumental variables for all of the control variables in Table 3. For the Klein and Vella (2010) estimator, we followed Farré, Klein, and Vella ( ) and ( ) , with (2013), assuming exponential functions for both = , and used the X 9 variables from Table 3. The results yielded by the Lewbel (2012) estimator (2SLS and GMM) are similar, although the GMM estimate is more precise, and indicate an impact of around 4.5% by zoning on the rental price. This value is not so different from the estimate obtained with OLS, and it is very similar to some estimates obtained from a sensitivity analysis of Table 7 using the approach suggested by Inchino et al. (2008). Using a Breusch-Pagan test for the residues of equation (7), we obtained a Chi-2 statistic of 572.25, strongly rejecting the hypothesis of homoscedasticity. Note also that the first stage F statistic is high and

9

The results of the first stage for estimator of Lewbel (2012) and of the different steps for the estimator of Klein and Vella (2010) are available upon request.

20

significant, indicating the relevance of the instruments and that the reported J statistic does not indicate a rejection of the null hypothesis about the validity of overidentifying restrictions. The estimate obtained using the estimator of Klein and Vella (2010) is presented in the fourth column of Table 9. Here we obtained that zoning adoption is associated with a 5.3% increase of the rental prices of Brazilian municipalities, a value similar to the estimate obtained using OLS and PS matching (see the second column of Table 4). Thus, these two new estimates for the impact of zoning on the rental prices of Brazilian municipalities bring additional support for a positive effect of the policy on rental prices. These estimates also appear to suggest, once more, that unobservable variables have almost no substantive influence on the effect. Table 8 - The impact of zoning on average rental prices: identification through heteroscedasticity (1)

(2)

OLS

Treatment Controls Generated Instruments Control function

0.054*** (0.008) Yes No No

F first stage

-

Hansen’s J

-

N. Observations

5,408

(3) Lewbel

(4) Klein and Vella

2SLS

GMM

0.046*** (0.014) Yes Yes No 21.848 (0.000) 39.370 (0.5871) 5,408

0.045*** (0.011) Yes Yes No

0.053*** (0.008) Yes No Yes

-

-

-

-

5,408

5,408

Note: standard error and p-value, respectively, are in parentheses for parameters and test statistics. Control variables include the variables presented in Table 2. Generated instruments are obtained according to Lewbel (2012, 2018) for all control variables. The control function is obtained using a ( ) and ( ). *** p < 0.01, ** p parametric non-linear function (the exponential function) for < 0.05.

7. Conclusions In recent decades, due to the increased autonomy of the local governments, the rise of income and a greater federal concern with urban-planning issues, the idea of urban land-use regulations has become widespread in Brazil. In this context, the number of cities with zoning laws grew from 64 in 1978 to 1,724 in 2013. However, little is known about the impact of these local policies on the functioning of the Brazilian real-estate market. Our study sought to contribute to this debate by providing evidence on the impact of zoning on the average rental prices and on the growth of the number of dwellings of Brazilian cities. To reduce concerns about endogeneity in the relationship between zoning and prices, we used strategies based both on selection on observable and on unobservable factors. Firstly, it as adopted abased on Propensity-Score Matching (PSM). The results indicate that zoning is associated with an increase in the average rental price of 5.3% to 6%. Although the empirical strategy relies on selection on observable characteristics (Conditional Independence Assumption, CIA), we showed that the results survive to different kind of failures in CIA (Rosenbaum 2002, Inchino et al. 2008 and Oster 2016). In addition, we showed that using the generated instruments proposed by Lewbel (2012) or a control function indicated by Klen and Vella (2010) for dealing with potential endogeneity does not change significantly these initial estimates. The fact that the cost of housing is higher in cities that restricted land use is consistent with the previous studies that showed that regulations favored the development of the informal housing market 21

and slum formation (Biderman 2007, Cavalcanti et al., 2018). Facing a higher price in the formal housing market, low-income households may migrate to the informal ones (characterized by poor urban infrastructures). In the case of Brazilian municipalities, it is thus possible that zoning ordinances may be generating not negligible social costs. More investigation about these possible indirect consequences of zoning, however, is necessary. From the public policy perspective, note also that our results also suggest that Brazilian municipal land-use restrictions may negatively affect the more recent social-housing financing program Minha Casa Minha Vida, once its positive influence on prices makes the recipients of the federal program move farther away from their jobs.

22

References Afonso JR, Araújo, EA, Nóbrega, MAR (2013) IPTU no Brasil: um Diagnóstico Abrangente, FGVProjetos. Available for download in: www.fgv.br/fgvprojetos. Alterman R (2013) Planning Laws , Development Controls , and Social Equity : Lessons for Developing Countries. World Bank Legal Review 5: 1–32. Altonji JG, Elder T and Taber TEE (2005) Selection on observed and unobserved variables: Assessing the effectiveness of Catholic schools. Journal of Political Economy 113(1): 151–184. Arimah BC and Adeagbo D (2000) Compliance with urban development and planning regulations in Ibadan , Nigeria. Habitat International 24(3): 279–294. Avila P (2006) Urban Land Use Regulation in Brazilian cities – Impact on Urban Land Markets and access of Low Income People to Land and Housing. In: Brazil Inputs for a Strategy for Cities: A contribution with a focus on cities and municipalities. World Bank, pp. 216–255. Avila P (2015) Land subdivision legislation in Brazil—guilty of promoting insecure housing tenure? Economic Analysis of Law Review 6(2): 304–321. Bertaud A and Malpezzi S (2001) Measuring the costs and benefits of urban land use regulation: A simple model with an application to Malaysia. Journal of Housing Economics 10(3): 393–418. DOI: 10.1006/jhec.2001.0293. Biderman C (2007) Regulation and Informal Settlements in Brazil : A Quasi-Experiment Approach. Lincoln Institute of Land Policy. Boamah NA (2013) Land use controls and residential land values in the Offinso South municipality, Ghana. Land Use Policy 33. Elsevier Ltd: 111–117. DOI: 10.1016/j.landusepol.2012.12.016. Borges M (2007) Zoneamento na Cidade do Rio de Janeiro: Gênese, Evolução e Aplicação. UFRJ. Brueckner JK and Lall S (2015) Cities in Developing Countries: Fueled by Rural-Urban Migration, Lacking in Tenure Security, and Short of Affordable Housing. Handbook of Regional and Urban Economics (5). Brueckner JK and Sridhar KS (2012) Measuring welfare gains from relaxation of land-use restrictions: The case of India’s building-height limits. Regional Science and Urban Economics 42(6). Elsevier B.V.: 1061–1067. DOI: 10.1016/j.regsciurbeco.2012.08.003. Caliendo M and Kopeinig S (2008) Some Practical Guidance for the Implementation of Propensity Score Matching. Journal of Economic Surveys 22(1): 31–72. Cavalcanti T, Da Mata D and Santos M (2018) On the Determinants of Slum Formation. The Economic Journal Forthcomin. Dantas R, Duarte G, Silveira Neto R, et al. (2018) Height restrictions and housing prices: A difference-indiscontinuity approach. Economics Letters 164. Elsevier B.V.: 58–61. DOI: 10.1016/j.econlet.2018.01.002. Diprete TA and Gangl M (2004) Assessing Bias in the Estimation of Causal Effects: Rosenbaum Bounds on Matching Estimators and Instrumental Variables Estimation With Imperfect Instruments. Sociological Methodology 34: 271–310. Fernandes E (2007) Implementing the urban reform agenda in Brazil. Environment and Urbanization 19(1): 177–189. DOI: 10.1177/0956247807076724. Fischel W (2004) An economic history of zoning and a cure for its exclusionary effects. Urban Studies 41(2): 317–340. DOI: 10.1080/0042098032000165271. Fischel WA (2001) The Home-voter Hypothesis. How Home Values Influence Local Government 23

Taxation, School Finance, and Land-use Policies. Harvard University Press. Glaeser EL and Ward B a. (2009) The causes and consequences of land use regulation: Evidence from Greater Boston. Journal of Urban Economics 65(3). Elsevier Inc.: 265–278. DOI: 10.1016/j.jue.2008.06.003. Ihlanfeldt KR (2007) The effect of land use regulation on housing and land prices. Journal of Urban Economics 61(3): 420–435. DOI: 10.1016/j.jue.2006.09.003. Kahn ME (2011) Do liberal cities limit new housing development? Evidence from California. Journal of Urban Economics 69(2). Elsevier Inc.: 223–228. DOI: 10.1016/j.jue.2010.10.001. Kok N, Monkkonen P and Quigley JM (2014) Land use regulations and the value of land and housing: An intra-metropolitan analysis. Journal of Urban Economics 81. Elsevier Inc.: 136–148. DOI: 10.1016/j.jue.2014.03.004. Lall S V, Wang HG and Da Mata D (2007) Do Urban Land Regulations Influence Slum Formation? Evidence From Brazilian Cities. Proceedings of the 35th Brazilian Economics Meeting. No. 119. ANPEC-Brazilian Association of Graduate Programs in Economics. Macedo J (2008) Urban land policy and new land tenure paradigms: Legitimacy vs. legality in Brazilian cities. Land Use Policy 25(2): 259–270. DOI: 10.1016/j.landusepol.2007.08.001. Malpezzi S and Mayo SK (1997) Getting Housing Incentives Right: A Case Study of the Effects of Regulation, Taxes , and Subsidies on Housing Supply in Malaysia. Land Economics 73(3): 372–391. McDonald JF and McMillen D (2012) The Economics of Zoning. In: Brooks N, Donaghy K, and Knaap G-J (eds) The Oxford handbook of urban economics and planning. 1st ed. Oxford University Press, p. 1006. Monkkonen P and Ronconi L (2013) Land Use Regulations, Compliance and Land Markets in Argentina. Urban Studies 50(10): 1951–1969. DOI: 10.1177/0042098012471982. Nannicini T (2007) Simulation-based sensitivity analysis for matching estimators. Stata Journal 7(3): 334. Nery Júnior J (2013) Um Século de Política Para Poucos : O Zoneamento Paulistano 1886-1986. Anais: Encontros Nacionais da ANPUR 10. Oster E (2017) Unobservable Selection and Coefficient Stability: Theory and Evidence. Journal of Business & Economic Statistics 0(0): 1–18. DOI: 10.1080/07350015.2016.1227711. Quigley JM (2007) Regulation and Property Values in the United States: The High Cost of Monopoly. Program on Housing and Urban Policy W06-004. . Quigley JM and Raphael S (2005) Regulation and the high cost of housing in California. American Economic Review .(.): 323–328. Quigley JM and Rosenthal L (2005) The effects of land regulation on the price of housing: What do we know? What can we learn? Cityscape 8(1): 69–110. Rosenbaum P and Rubin D (1983) The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika 70(1): 41–55. Rosenbaum PR (2002) Observational studies. New York: Springer. Schuetz J, Meltzer R and Been V (2011) Silver Bullet or Trojan Horse? The Effects of Inclusionary Zoning on Local Housing Markets in the United States. Urban Studies 48(2): 297–329. DOI: 10.1177/0042098009360683. Seabra D and Azzoni CR (2015) Custo de Vida Comparativo para os Distritos das 100 Maiores Cidades Brasileiras. Temas de Economia Aplicada - FIPE: 12–24. Solé-Ollé A and Viladecans-Marsal E (2013) Do political parties matter for local land use policies? Journal of Urban Economics 78: 42–56. DOI: 10.1016/j.jue.2013.07.003. 24

Wang L and Krieger AM (2006) Causal conclusions are most sensitive to unobserved binary covariates. Statistics in Medicine 25(13): 2257–2271. DOI: 10.1002/sim.2344. Zabel J and Dalton M (2011) The impact of minimum lot size regulations on house prices in Eastern Massachusetts. Regional Science and Urban Economics 41(6). Elsevier B.V.: 571–583. DOI: 10.1016/j.regsciurbeco.2011.06.002. Zhou J, McMillen DP and McDonald JF (2008) Land Values and the 1957 Comprehensive Amendment to the Chicago Zoning Ordinance. Urban Studies 45(8): 1647–1661. DOI: 10.1177/0042098008091495.

25

Appendix Table 1A – Main Characteristics of Zoning Ordinances in Selected Brazilian Cities

Municipality

City Size (per thousand)

Current Local Law

Minimum Permeability Area

Maximum Occupancy Rate

Maximum FAR

Height Limit

Protection Areas

Inclusionary Zoning (ZEIS)

Exclusively Residential Zones

Exclusively Industrial Zones

São Paulo - SP 12,110 16402/2016 15 to 20% 50 to 85% 1 to 4 10 to ―n‖ meters Yes Yes Yes Yes Fortaleza - CE 2,610 7987/1996 20 to 40% 40 to 60% 1 to 3 15 to 95 meters Yes Yes No Yes Manaus - AM 1,793 1836/2014 15% 1 to 5.4 4 to 25 floors Yes Yes No Yes Curitiba - PR 1,765 9800/2000 25 to 50% 30 to 50% 0.4 to 5 2 to ―n‖ floors Yes Yes No Yes 16176/1996 20 to 60% 2 to 7 24 to ―n‖ meters Yes Yes No Yes Recife - PE 1,555 B. Horizonte - MG 1,433 7166/1996 10 to 20% 20 to 50% 1 to 5 Yes Yes No No Campo Grande - MS 774 74/2005 0 to 12.5% 50 to 80% 1 to 6 Yes Yes No No Ribeirão Preto - SP 590 2505/2012 20 to 35% 75 to 80% 3 to 5 21 to ―n‖ floors Yes Yes Yes Yes Aracaju - SE 571 42/2000 5% 40 to 90% 1 to 6 6 to ―n‖ floors Yes Yes No Yes Maringá - PR 342 331/1999 30 to 80% 1 to 6 650 meters Yes Yes No Yes Dourados - MS 210 205/2012 10 to 20% 60 to 80% 1.2 to 9 2 to ―n‖ floors Yes Yes No Yes 349/2016 20 to 30% 50 to 70% 1 to 4 30 to ―n‖ meters Yes Yes No Yes Anápolis - GO 325 Rio Branco - AC 319 1611/2016 10 to 40% 50 to 70% 1 to 3 3 to 4 floors Yes Yes No Yes 926/2006 20 to 30% 60 to 70% 1.5 to 2 4 to 12 floors Yes Yes No Yes Boa Vista - RR 278 Vitória - ES 211 6705/2006 10% 30 to 75% 1.2 to 2.7 2 to ―n‖ floors Yes Yes No Yes Petrópolis - RJ 186 5393/1998 10 to 15% 25 to 70% 1 to 2.5 10 to 15 meters Yes Yes No No 003/2004 20 to 50% 1.6 to 2 3 to 15 floors Yes Yes No Yes Imperatriz - MA 234 003/2014 50 to 70% 1 to 4.5 15 to 24 meters Yes Yes No Yes Uruguaiana - RS 116 4% 052/2007 10 to 40% 40 to 70% 0.8 to 4 2 to 25 floors Yes Yes No No Serra Talhada - PE 85 Viçosa - MG 72 1429/2000 10 to 30% 50 to 80% 1 to 3.5 3 to 10 floors Yes Yes No Yes Note: The legal limits of density constraints (permeability area, occupancy rate, FAR and Height Limit) vary according to each zoning boundary and building height can be unlimited in some cities. The protection areas are characterized as construction constraints in areas with environmental assets or historical-cultural patrimony.

26

Table 2A – Average Regression Coefficients of the Hedonic Models. Variable Brick Dummy (1/0) Number of Rooms Number of Bathrooms Number of Other Rooms Constant

Average Coefficient 74.4631 30.9680 218.7929 30.0606 -128.0507

Note: The figures reported in the table display the average coefficient for each variable. 5507 municipalityregressions were estimated, where the observation unit is the dwelling. The number of observations varies for each municipality.

Figure 1A – Histogram of propensity scores across treated and control municipalities.

Note: The distribution is characterized by the nearest-neighbor propensity scores.

27