MACROECONOMICS, TRADE AND INVESTMENT E Q U I TA B L E G R O W T H , F I N A N C E & I N S T I T U T I O N S N OT E S Gender and Property Taxes in São Paulo May 2023 EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 1 2 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE >>> TABLE OF CONTENTS ACKNOWLEDGEMENTS 4 EXECUTIVE SUMMARY 5 1. INSTITUTIONAL CONTEXT 8 1.1 Property taxes in Brazil 8 1.2 The geography of São Paulo 9 2. DATA AND DESCRIPTIVES 11 2.1 Main data sources 11 2.2 Definition of key variables 12 2.3 Preliminary descriptive statistics 14 3. MAIN FINDINGS 16 3.1 Overall property ownership 16 3.2 Wealth ownership 19 3.3 Implications for property tax distribution 24 4. POLICY IMPLICATIONS AND CONCLUSION 26 REFERENCES 29 APPENDICES 31 A.1 Determining the sex of property owners 32 A.2 Calculation of property taxes in São Paulo 34 A.3 Credit markets for property acquisition in Brazil 36 EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 3 >>> ACKNOWLEDGEMENTS This note was prepared by Thiago Scot (DIME, World Bank), Tatiana Flores (DIME, World Bank), Davi Moura (PUC-Rio), Javier Feinmann (UC Berkeley), and Roberto Hsu Rocha (UC Berkeley). We would like to express our gratitude to The World Bank Global Tax Program (GTP) and DEC Research Support Budget (RSB) for their generous support. We also extend our thanks to Ceren Ozer (Senior Economist and Program Manager of the GTP, World Bank) for her valuable advice. Our peer reviewers, Hitomi Komatsu (MTI, World Bank) and Caio Piza (DIME, World Bank), provided us with insightful comments and suggestions, which we greatly appreciate. We would also like to thank Juan Santini (DIME, World Bank) for his feedback and Miguel Jacob for his generous help in accessing the data. 4 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE >>> EXECUTIVE SUMMARY As income inequality increased in the developed world in the last decades and the labor share in total income declined, academic and policy discussions on capital and wealth taxation have garnered substantial attention (Saez, Zucman, and Piketty 2022; Piketty and Saez 2012; Zucman 2019). The most common form of wealth taxation, going back centuries, is immovable property and land taxes (Dray, Landais, and Stantcheva 2022). Despite their antiquity and positive features,1 property taxes are often the target of severe political opposition and are overall underutilized (Cabral and Hoxby 2012). Taxes on property collect particularly little revenue in low- and middle-income countries— on average 0.3 to 0.6 percent of GDP, while OECD countries collect 2 to 3 percent (Ali, Fjeldstad, and Katera 2017), suggesting significant increases are possible. Despite the challenges to compiling accurate, comprehensive, and up-to-date cadasters, local governments across the developing world have invested significant resources in generating systems to capture and update property registries (Sepulveda and Martinez- Vazquez 2012; Okunogbe 2021).2 Since property taxation is intrinsically linked to property ownership and valuation, understanding ownership patterns across societal groups is key to understanding these taxes’ distributional effects. Several studies in low- and middle-income countries have shown that women face significant barriers to owning property, including cultural norms, legal restrictions, and lack of access to credit (Gaddis, Lahoti, and Swaminathan 2022; Komatsu et al. 2021b). Although they are fading, gender-based legal distinctions in inheritance and asset ownership are a reality: nearly 40 percent of all countries still have some level of restriction on women’s property rights (World Bank 2022). This knowledge note provides new evidence on property ownership and taxation patterns across genders in São Paulo (Brazil), the largest city in the Americas, with 12 million inhabitants. We exploit microdata on all commercial and residential properties to document the share of total property and property wealth owned by women, the geographic distribution of female-owned properties, and the implications of this data for property taxes in the city. 1 Property taxes are potentially much less distorting than income taxes and are often directly connected with provision of local public goods and a stable tax base. 2 Knebelmann (2022) discusses how new digital tools like satellite imagery and geolocated cadasters can improve property detection, registration, and billing processes and, overall, change how local governments enforce property taxation. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 5 We start by documenting large gaps in property ownership in the city: women own 30 percent of properties. Properties solely owned by men represent half of all properties, a 20 percentage point (p.p.) higher share than those solely owned by women. Joint ownership by both genders represents less than 10 percent of all properties, and the remaining properties are owned by private and public firms. These gaps in ownership are even higher (26 p.p.) for commercial as compared to residential properties. Gender disparities are even higher in property wealth ownership. Not only do women own fewer properties overall, but the gap is particularly large for high-value properties: among the top 1 percent of highest-value properties, female ownership is approximately 20 percent. That implies the property wealth gap—the difference in property wealth owned by men and women—is even larger: women own 18 percent of total property wealth compared to 38 percent owned by men. These patterns are replicated geographically, but they are larger in the city’s wealthiest areas. When looking at the property level, we observe that properties solely owned by women are 9 percent less valuable than those solely owned by men. Property characteristics like size, type, and year of construction almost fully explain this gap. We then analyze the implications for property taxation. Since property taxation in São Paulo is increasing in property value, women also pay proportionally less taxes. Despite a flat headline rate of 1 percent, the effective tax rate paid by properties in São Paulo increases in assessed value due to a complex schedule of exemptions and surcharges. We document that women are more likely to solely own fully exempt properties (one- third of their properties are exempt versus 27 percent for men) and less likely to own properties paying more than 1 percent in effective rate. Properties solely owned by women pay 14 percent of all property tax revenue, while those solely owned by men pay 32 percent—the vast majority of the remaining taxes are all paid by properties owned by private firms. Our study contributes new evidence on the ownership of properties and on property wealth across genders for the largest city in the Americas. Coelho et al. (2022) note the scarcity of “data on the distribution of wealth by gender in developing countries.” Gaddis, Lahoti, and Swaminathan (2022) use Demographic and Health Survey (DHS) data from 41 low- and middle-income countries to document patterns of property ownership across genders. In almost all countries, men are 20 to 50 p.p. more likely to own housing than women. Exceptions are documented by Kotikula and Raza (2021) among poor urban dwellers in Bangladesh and by Holden and Tilahun (2020) among farmers in Tigray, Ethiopia, where the gender gap in housing ownership is practically null. The richness of our data also represents an improvement on previous efforts to measure property tax liabilities across genders. Komatsu et al. (2021a) note large discrepancies in tax liabilities for Ethiopian households when comparing self-reported measures with imputed values using land area and tax schedules, stating that this “highlights the importance of updated administrative tax data and land registries to complement survey data.” 6 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE We are also able to provide detailed evidence of gender disparities in ownership in a setting with no de jure restrictions on female ownership. According to the World Bank (2022), Brazil provides ample legal protection for women regarding property ownership: men and women have equal ownership and administrative rights over property in a relationship, sons and daughters have equal inheritance rights, and surviving spouses have equal rights regardless of gender. Nonetheless, these protections were only introduced with the new 2002 Civil Code. Before that, the law of the land was the 1916 Civil Code, which stated, “The husband is the head of the household, a function he exercises with the collaboration of the wife,” and which made husbands responsible for “managing joint assets.” Despite these crucial legal changes, the fact that laws explicitly limited the role of women in owning and managing properties within a couple suggests that norms might still assign different roles to husbands and wives within households. In the concluding section, we discuss in more detail the implications of our findings in light of previous research on de jure versus de facto ownership rights. Some caveats about our findings should be noted. First, our property value and property wealth ownership measures are based on assessed value, not on the market value of properties. Previous studies have not only documented that assessments are often lower than market value, but they have also found them to be regressive (Berry 2021). Potentially more worrying for our findings, Avenancio-León and Howard (2022) show that the gap between assessed and market value in the United States is particularly large for minority groups. The main mechanism behind this gap is that assessments fail to capitalize on local amenities that increase property market prices—since cities are vastly segregated in the United States, this leads to higher gaps in areas where white owners live. While we recognize that the levels of property value we measure likely underestimate true property value, we argue that differences across the owners’ gender are unlikely to be meaningful, since homeowners do not segregate across space based on gender. Second, our data only allows us to make statements about property tax liabilities, not about property tax payments. Important questions related to the levels of compliance across genders, including nonpayment and late payments, will require additional data. Finally, we use administrative data from the municipality’s property cadaster, meaning that our data does not cover informal settlements and unregistered properties. Carvalho Junior (2017) discusses estimates of property cadaster coverage in Brazil and puts coverage in São Paulo at the 66 to 75 percent range. Our results, in that sense, should be interpreted as being representative of these formally registered properties. The rest of this report is organized as follows. Section 2 provides institutional context and setup for the property tax in Brazil. Section 3 details data sources and main data descriptive statistics. Section 4 analyzes property ownership patterns and wealth ownership, discussing some of these patterns’ implications for property tax distribution. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 7 1. Institutional context 1.1 Property taxes in Brazil Property taxes in Brazil (IPTU, from the Portuguese acronym for “urban property and land tax”) are assessed and collected by municipalities, the paid at once but are divided into as many as ten installments over the year. Brazilian municipalities rely heavily on transfers lowest administrative level in the country.3,4 Tax from the federal government to provide a wide rates vary between cities, but mostly fall between range of services, including health and education. 1 and 3 percent of assessed property valuation, In Figure 1a we document that, for the vast majority payable every year. In the municipality of São Paulo, of municipalities in the country, revenue from the headline tax rate for residential properties is 1 property taxes is very small in relative terms: IPTU percent of the assessed value, while the rate for revenue is less than 5 percent of total revenue in other properties (mainly commercial or unused 90 percent of municipalities. Although no single land) is 1.5 percent. As we discuss further below, factor explains the mixed revenue performance several adjustments introduce “progressivity” to of IPTU across municipalities or its overall low the flat rate, such that higher-value properties face performance as a revenue source, São Paulo higher rates. Often, property taxes need not be 3 See Carvalho Junior (2017) for a thorough discussion of history, main features, and debates about property taxation in Brazil. 4 We use the term “property taxes” to refer to recurrent, yearly tax liability based on property values. Property transaction taxes, also called stamp duties, also exist in Brazil in the form of ITBI (Tax on Immovable Property Transactions), another municipal tax. 8 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE stands out: in 2019, property taxes represented 18 Figure 1b, property taxes become more relevant: percent of the municipality’s total revenues. When the median municipality collects 25 percent of its excluding transfers from the central government own taxes from IPTU, while São Paulo collects over and focusing on local tax revenue, as described in 35 percent.5 Figure 1: Share of local property taxes in revenue (a) IPTU as share of total revenue (b) IPTU as share of own revenue 2000 300 Number of municipalities Number of municipalities 1500 200 1000 100 500 0 0 0 10 20 30 40 0 25 50 75 100 IPTU as share of total revenue (%) IPTU as share of own revenue (%) Source: Matos dos Santos, Souto Valente Motta, and Estorani de Faria (2020). Note: These figures present histograms of the share of local property taxes in total revenue (panel a) and own-revenue (panel b) across Brazilian municipalities in 2019. 1.2 The geography of São Paulo São Paulo, the most populated municipality in municipality’s core lies the densely occupied central Brazil, has a population estimated at over 12 district, dominated by commercial properties million in 2021 and forms the center of a broader and surrounded by higher-income residential metropolitan region with approximately 20 million areas, particularly to its southwest. Lower-income inhabitants. The map of São Paulo in Figure 2 households are more often found in an outer ring provides some insight into the city’s geography by of regions farther from the central district, in plotting the property density in each of its more particular in the vast eastern regions, which are than 40,000 zip codes. The vast southern and home to 4.5 million people or close to 40 percent of extreme northern areas show little density and are the city’s population. nonurban, environmentally protected areas. At the 5 The most important own-revenue for most municipalities is the tax on services (ISS, imposto sobre servicos), a turnover tax on commercial service provision. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 9 Figure 2: Number of properties by zip code Source: Original calculations for publication. Note: This map presents the number of properties (residential and nonresidential) in each zip code of São Paulo for 2018. Larger regions delimited by thick lines are fiscal zones, the administrative boundaries related to property valuation assessment. 10 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE 2. Data and descriptives 2.1 Main data sources The main data source for this study is administrative data on the universe of property assessments in the municipality of São Paulo, Brazil, between 2008 and whether the property is single residency or multiunit, and valuation assessments, such as assessed price per square meter. 2018. This is an administrative cadaster, updated We complement this dataset with two other data yearly, containing information such as property sources. First, the municipality of São Paulo provides type (commercial or residential), address, zip code, a shapefile of the city that links each property year of construction, and a unique identifier that identifier to a geolocated area, allowing us to map remains constant over time, thus allowing us to all the properties in the database and construct track properties even when they change ownership. aggregates at levels such as zip code and fiscal The annual update to the database covers all zones, the city administrative boundaries linked to key information used to calculate the property’s area valuation. Second, we match properties to their assessed value (valor venal in Portuguese), which owners using the property identifier. Ownership is the base for property taxes. We discuss in more data contains the name and national identifiers of detail below how the tax base is calculated, but owners (CPF),6 which allows us to match them to the information available includes both property other administrative databases. characteristics, such as land and built areas and 6 The Individual Taxpayer Registration (CPF) is the registry maintained by the Federal Revenue of Brazil, where any natural person must register once, regardless of age or nationality. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 11 2.2 Definition of key variables This section discusses in-depth how we construct only the value of total area but also other factors the key variables used throughout the study, starting related to the type of terrain, whether the unit is with how we assign sex to property owners and in a multiapartment building, and the property followed by details on the calculation of property perimeter.10 We then assign to each property an valuation and tax liabilities. assessed value equal to the sum of these three components. It is worth mentioning that the excess Our dataset includes the full names of property land value only applies when large fractions of the owners, which enables us to extract each of total property area are undeveloped; the precise their first names. We then match the owners to thresholds vary by region of the city, but overall a database generated from the 2010 Brazilian we estimate that only 8 percent of total properties Census, which contains more than 100,000 have positive excess area value. first names and, for each name, determine the frequency with which the individuals self-declare Once we obtain each of the three valuation their sex as male and female.7 First names in Brazil components, we can calculate the total tax liability. are mostly gendered, with over 90 percent of We follow the nomenclature used by the municipality unique first names exclusively used by a single sex. and define the total property liability of property i of We then use a simple classifier, assigning an owner type t = {Residential, Commercial} as as being female if more than 50 percent of the Ti,t = τP,t * (ValueBuildingi + ValueOccupiedi) + τL * registered individuals with the same first name are ValueLandi + Adjustmentsi registered in the Census as female.8 We recognize that this method will sometimes misclassify where Ti,t is the total tax liability of property i of individuals whose names are not clearly gendered type t; τP,t is the tax rate applied to the sum of the and that it does not distinguish individuals whose constructed and occupied area for a property of self-identified gender does not conform to names type t; and τL is the tax rate applied to unoccupied mostly used by a different sex. land. In the period we study, the tax on constructed and occupied area τP,t was 1 percent for residential We then compute the assessed value for each properties and 1.5 percent for commercial property, which is the sum of three key quantities: properties, and the tax on unoccupied land was 1.5 the value of the constructed area, the value of the percent. occupied area, and the value of “excess land.” The value of the constructed area is simply calculated Despite the headline flat taxes on property values, in as the assessed value per square meter multiplied practice the final tax liability depends on a complex by the total constructed area, with adjustments schedule of exemptions and surcharges, which we for depreciation depending on building age.9 The summarize in the expression above with the term calculation of the value of the occupied area Adjustmentsi. The main form of exemption is based and excess land is more complex, involving not on assessed value. For 2018, any residential property 7 This data is publicly available at https://github.com/turicas/genero-nomes. 8 We provide more detail on this procedure in Appendix A1. 9 Assessed values per square meter are defined by the city government and are usually updated yearly according to inflation. A more detailed discussion on the method used to define value per square meter is beyond the scope of this note. See Carvalho Junior and De Cesare (2022) for examples. 10 We provide more details on these calculations in Appendix A2, including the detailed tax form received by taxpayers, as shown in Figure A2. 12 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE with an assessed value below R$160,000 was fully we study. All of these adjustments are not directly exempt from property taxes, and properties with available in the public datasets, so we construct values between R$160,000 and R$320,000 faced them based on detailed rules to recover the actual a decreasing exemption. An additional exemption liabilities for each property. The main implication adjustment can also increase or decrease tax of all these adjustments, as we document below, liabilities between -0.03% and +0.05% of assessed is that in practice effective tax rates are increasing value. Finally, a municipal law capped the year-over- in assessed property value, despite a flat headline year increase in tax liability at 10 percent in the period rate.11 11 We note that other, more idiosyncratic exemptions exist. Over time, exemptions were provided for Second World War veterans, low- income elderly individuals, individuals suffering from cancer, parents of foster children, etc. Unfortunately, we do not observe this information in the data and thus are not able to incorporate these exemptions in our analysis. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 13 2.3 Preliminary descriptive statistics We provide preliminary descriptive statistics of that the median property value was R$160,000 our sample in Table 1. We observe approximately and that three-quarters of properties were valued three million properties every year and slightly over at less than R$300,000. Note, however, that the two million unique property owners. The mean distribution includes a long right tail of properties assessed value of all properties was ≈ R$340,000 in valued above R$1 million. The mean tax liability was 2018, but with wide dispersion: Figure 3 presents 12 close to 1 percent of the mean assessed value at a histogram of 2018 assessed values, documenting R$3,900. Table 1: Descriptive statistics 2012 2014 2015 2017 2018 Property characteristics Share residential 0.84 0.85 0.85 0.85 0.85 Mean assessed value (R$2018) 243,190.19 237,029.56 337,759.33 350,788.89 345,898.11 Mean tax liability (R$2018) 2,706.85 2,728.48 2,671.52 3,068.20 3,936.76 Share of exempt properties 0.31 0.28 0.31 0.27 0.27 Property ownership Share owned by women only 0.27 0.28 0.28 0.29 0.29 Share owned by men only 0.51 0.50 0.50 0.49 0.49 Share owned jointly by 0.09 0.09 0.09 0.09 0.09 women and men Share owned by private firms 0.10 0.11 0.11 0.11 0.11 Share owned by public firms 0.02 0.02 0.02 0.02 0.02 Number of properties 3,085,491 3,186,315 3,235,448 3,383,561 3,414,208 Number of zip codes 41,132 41,319 41,399 41,689 42,143 Number of owners 2,160,459 2,253,353 2,285,747 2,385,551 2,420,329 Source: Original calculations for publication. Note: This table presents the descriptive statistics for 2012, 2014, 2015, 2017, and 2018. The share of residential properties is the share of all properties the municipality considers to be of the type Residência/Residencial. Mean assessed value is the sample mean of the computed assessed value for all properties each year, following the municipality’s method of calculating the property value. We take 2018 values using INPC. Mean tax liability is calculated from the mean assessed value, taking into account exemptions and discounts for each different property. Number of owners for each year is the number of unique owners. We consider owners to be unique owners if their names and the four digits of their CPF differ from those of all other owners. 12 Approximatelly USD150,000 in 2018 PPP exchange rate (USD 1 = BRL 2.23). 14 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE One key component in the calculation of tax remaining non-firm-owned properties are classified liabilities is the exemptions: we estimate that each as being owned solely by men, solely by women, or year due almost one-third of all properties were jointly by men and women. We first document the exempt from paying property taxes due to their low prevalence of single-gendered property ownership assessed value.13 over mixed-gender joint ownership, the latter representing 10 percent of properties across the In Table 1 we also provide some preliminary years. The pattern of ownership by gender is quite statistics on the characteristics of owners. Around stable: half of the properties are owned exclusively 10 percent of all properties are owned by firms; by men; 27 to 30 percent of properties are owned as we discuss below in more detail, private firms exclusively by women. own a larger share of commercial properties, and public firms own some residential properties. The Figure 3: Histogram of assessed values in 2018 200,000 150,000 Count 100,000 50,000 0 0 500,000 1,000,000 1,500,000 Assessed value Source: Original calculations for publication. Note: This figure presents the histogram of the computed 2018 assessed values for all properties. The vertical red lines show the 25th, 50th and 75th percentiles. We trim the distribution at the 97.5th percentile, equivalent to R$1.4 million valuation, for visualization purposes. 13 This magnitude is consistent with estimates from Carvalho Junior (2017) for large municipalities in Brazil. Smaller municipalities often exempt less than 10% of all properties. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 15 3. Main findings This section presents the key findings regarding gender-related aspects of property/wealth ownership in the city of São Paulo and the relationship of gender to property taxes. For the remainder of the paper, we focus on data for 2018, the most recent year available in our panel. 3.1 Overall property ownership Women own fewer properties than men, across 40 percent of commercial properties, women solely different types of properties and throughout own 16 percent; private firms own almost a third of the city. In Table 2, we document key differences all of this type of property. in property ownership by gender. First, women This pattern is replicated throughout the city as solely own 30 percent of all properties in the city documented in Figure 4. Across most zones, women versus 50 percent owned exclusively by men. own between 20 percent and 40 percent of all Approximately 9 percent of properties are jointly properties, with a slightly higher ownership share in owned by two or more individuals from both sexes, the central, wealthier part of the city. But overall the while the remaining properties are owned by firms. gender gap in property ownership is rather similar As documented in Table 1, residential properties across the geography of the city and similar in represent approximately 85 percent of all properties. magnitude in both central and peripheral areas and Gender gaps are even larger for the smaller share of for areas with very different income and wealth levels. commercial properties: while men solely own over 16 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE In Table 2 we also document that the average value properties, which are valued at R$223,000. In the of properties owned solely by women and solely following section, we assess in more detail these by men are very different. Male-owned properties gaps in properties’ assessed value, since they have are valued at R$270,000 on average, 20 percent implications for the differences in ownership of higher than the average value of female-owned property wealth across genders. Table 2: Summary statistics on owners 2018 Mixed Men/ Private All Men Women Women Firm Public Firm Number of properties 3,414,208 1,668,753 997,855 308,631 393,857 45,112 Share of all properties (%) 100.00 48.88 29.23 9.04 11.54 1.32 Share of residential 100.00 50.02 31.44 9.13 7.96 1.46 properties (%) Share of commercial 100.00 42.50 16.89 8.56 31.49 0.56 properties (%) Mean value of all 345,898.11 270,139.06 223,848.55 322,429.09 1,012,848.28 185,656.74 properties (R$) Mean value of residential 240,357.42 238,936.00 205,450.22 289,521.31 355,983.64 103,132.56 properties (R$) Mean value of commercial 934,168.80 474,831.74 414,756.28 518,026.35 1,937,876.93 1,377,638.15 properties (R$) Source: Original calculations for publication. Note: This table presents the Number of properties, corresponding to the number of properties owned by each group; Share of all properties is the corresponding shares of ownership. Share of residential properties (%) and Share of commercial properties (%) are the share of the ownership for each group of the total number of residential/commercial properties. Mean value of all properties (R$) is the mean assessed property value per group in Brazilian reais. Likewise, Mean value of residential properties (R$) and Mean value of commercial properties (R$) represent the mean value for residential and commercial properties in Brazilian reais. Designations of Women or Men are considered when only one gender owns the property. If a property is split between men and women, it is considered Mixed property. We group public and private firms following the same algorithm. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 17 Figure 4: Share of female owned properties—all properties Source: Original calculations for publication. Note: This map plots the share of all properties owned solely by women in each fiscal zone of São Paulo in 2018. 18 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE 3.2 Wealth ownership The gender gap in property ownership is larger 9 quantiles for the top 1 percent; and 10 quantiles for high-value properties, implying a larger gap for the top 0.1 percent. We then exclude properties for property wealth ownership. In the previous owned by firms and compute the share solely section we documented the overall levels of owned by women, by men, and women and men property ownership by women. But the documented jointly. Our key takeaway is that, for most of the gaps might be very different depending on the distribution, the share of properties solely owned assessed value of properties—our best proxy for by women is rather stable at approximately 30 to property wealth. We use the valuation of properties 35 percent, as we previously documented. This to understand how this gender gap behaves across share starts to decrease for properties in the top 20 the distribution of property values. percent, such that for the 1 percent of highest-value properties (those with an assessed value above We summarize these findings in Figure 5. We R$2.5 million), female ownership is approximately first order properties by their assessed value and 10 to 15 percentage points (p.p.) lower, at about 20 separate them into 118 quantiles: 99 quantiles for percent. properties in the bottom 99 percent of valuation; Figure 5: Top G percentiles—no firm-owned property 1.0 Share of owners Women 0.75 Mixed Men/Women Men 0.50 0.25 Sizes per percentile 0.00 1-99%: 25875 99.1-99.9%: 2587 Bottom (1.2%) Top (1%,.9%) Top (.1%,.09%) Top .01% Percentiles 99.9 - 100%: 258 Source: Original calculations for publication. Note: This figure presents the share of ownership for the different percentiles. We show 99 columns for the bottom 99 percentiles, 9 columns for the bottom tenth-of-percentiles of the top percentile, and 10 for the 10 one-hundredth-of-percentiles of the top tenth-of-percentile. This only shows properties not owned by firms (public or private). Designations of Women or Men are considered only when one gender owns the property. If property ownership is split between men and women, it is considered mixed. We group public and private firms following the same algorithm. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 19 Figure 6: Difference of share of wealth/ share of ownership—women Source: Original calculations for publication. Note: This figure presents the difference between the share of wealth on properties owned by women and the share of property ownership by women in each fiscal zone. Properties are considered owned by women when only women own the property. If ownership is split between genders, we consider the property to have mixed ownership. 20 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE This decrease in sole female ownership at the top These patterns are replicated throughout the is driven both by an increase in properties owned by city, but gaps are larger in areas with higher- males only, which increase from 55 percent overall valued properties. We previously documented that to 60 percent at the top 1 percent, but also by an the share of properties owned by women is quite increase in properties owned jointly by male and stable across the geography of São Paulo. When we female owners, which increases from 10 percent to consider not only the share of properties, but the 15 to 20 percent at the top. 14 share of property wealth, the result is somewhat Figure 7: Difference of share of wealth/ share of ownership—women 0.8 Difference share wealth owned to ownership 0.4 0.0 -0.4 10.0 12.5 15.0 17.5 20.0 Log(average property value by zip code) Source: Original calculations for publication. Note: This figure presents a scatter plot of the difference between the share of total wealth on properties and the share of ownership, by women, by zip code and the mean assessed value of all properties. The linear fit is a simple regression of the difference of wealth owned and the share of ownership on the average wealth at the zip-code level, controlling for the log number of properties on each zip code. We use the mean value of log of the number of properties to plot the intercept in the figure. 14 In Figure A4 we reproduce these results across percentiles of property assessed value and include the participation of firms as well, both for all firms and separately for residential properties only. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 21 different. In Figure 6 we plot, for each fiscal zone fact. Racial differences in market and assessed in the city, the difference between the share of property value in the United States, for example, property wealth owned by women and the share are explained both by property characteristics of the number of properties owned. If the average and by location factors that drive property value valuation of women- and men-owned properties (Avenancio-León and Howard 2022). In Table 3 we were the same, this indicator would be zero across explore whether the gender gaps we observe are the city. What we observe instead is particularly fully explained by property traits like size and type of large gaps in the central district and adjoining use or by geography as well; since we only observe areas, where the share of property wealth owned assessed values, these differences would be driven by women is often 10 to 30 p.p. lower than the share by differences in assessed value per square meter of property units owned by them. In more distant in different areas of the city. We start by quantifying areas, where property valuations are lower, we the average gap in column (1): properties owned often observe smaller gaps and even positive gaps solely by women are 9 percent less valuable than in some zones, where women own a larger share of those exclusively owned by men. Once we control property wealth than they own property units, that for a series of property characteristics in column is, women in these zones own properties that are (2), this gap almost fully disappears: the point more valuable than men on average. We formalize estimate on the gap is actually positive but only that suggestive evidence in Figure 7, where we by +0.5 percent, an order of magnitude smaller in plot, for each zip code, the difference between the absolute values than the raw gap. These property share of property wealth and the share of property characteristics directly enter the value assessment, units women own and the average of all property but they are unrelated to the city’s assessment of value in each zip code. What we see is a negative which areas are more or less valuable, which is correlation, with a semi-elasticity of -0.02—when incorporated in assessed square-meter values. moving between two zip codes with 1 percent In column (3) we re-estimate the model by adding different average property values, we expect to see zip code fixed-affects, which absorbs almost a 2 percentage point higher gap in female wealth all variation in terms of valuation of space. This ownership in the zip code with more expensive changes our coefficient to -1 percent, but again the properties. magnitude is very small compared to the overall gap. As we discussed above, this is consistent with Property characteristics fully explain the gender the idea that the gender gap in property value is not gap in property values. While we document that explained by gender segregation across space but women own property that is, on average, less by men and women owning properties that differ valuable than property owned by men and that across observable dimensions, such as size, age, these gaps vary across the city, we have so far and use. not discussed what mechanisms explain that 22 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE Table 3: Regression table: Determinants of gender gaps in property value and tax liability Dependent Log (Winsorized property value) Winsorized tax due variables: OLS Poisson (1) (2) (3) (4) (5) (6) Model: OLS OLS OLS Poisson Poisson Poisson Variables -0.0904*** 0.0051*** -0.0102*** -0.2056*** -0.0367*** -0.0484*** Women (0.0012) (0.0006) (0.0003) (0.0026) (0.0036) (0.0015) -0.3030*** 0.0747*** -0.0167*** -1.080*** -0.0092 -0.2948*** Residential (0.0023) (0.0038) (0.0024) (0.0029) (0.0203) (0.0061) Fixed effects Zip code Yes Yes Property traits Yes Yes Yes Yes Fit statistics Observations 2,666,608 2,666,608 2,666,608 2,666,608 2,592,268 2,586,867 R2 0.01323 0.73234 0.91943 0.06664 0.05790 0.63620 Pseudo R2 0.00488 0.48280 0.92259 0.09237 0.63235 0.84033 BIC 7,244,352.3 3,767,391.5 1,179,502.3 1.04 × 1010 -2,147,483,648.8 1,773,962,517.8 Heteroskedasticity-robust standard-errors in parentheses Signif. Codes: ***: 0.01, **: 0.05, *: 0.1 Source: Original calculations for publication. Note: Control for property traits include year of construction and type of use fixed effects, as well as controls for the log of build and land areas. Sample is restricted to properties owned exclusively by men or women. Property value and tax liability values are winsorized at the 1st and 99th percentiles. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 23 3.3 Implications for property tax distribution Despite the headline 1 percent flat rate, the to high-value properties facing higher effective effective property tax rate in São Paulo increases rates. In 2018, the exemption level for residential with assessed value due to exemptions and other properties was close to the median of the property adjustments. In Figure 8 we plot the average and distribution, so the median property below that median effective tax rate, defined as the tax liability level is paying a zero effective rate.15 Starting from divided by assessed value, across the percentiles the median, the effective rate increases linearly of assessed value distribution. If tax liability was toward 1 percent as discounts are phased out and simply calculated as 1 percent of assessed value, then keeps increasing, particularly rapidly at the both lines should be flat at 1 percent, since all top as property prices rise fast—the top 1 percent properties would face the same effective rate. In highest-value properties face an effective rate of reality, the complex rules of exemptions, discounts, 1.8 percent due to surcharges for properties above and surcharges create a tax schedule that leads R$1.2 million.16 Figure 8: Effective tax rate per percentile 2.0 Median Mean 1.5 Effective tax rate 1.0 0.5 0.00 0 25 50 75 100 Percentile Source: Original calculations for publication. Note: This figure presents the mean and median effective rate, i.e., the tax liability divided by the computed value of the property, for all properties and for each percentile of property value. Lower percentiles can be formed of parking spaces, which are never exempt; also, because each owner can only have one property exempt, the municipality chooses the property with the highest discount, therefore some lower percentiles’ values can have higher effective tax rates than properties in the higher percentiles. 15 The high level of effective rates at the bottom 10 percent of the distribution is mainly explained by properties used as parking spots: these have low value but are never exempt. Exemptions also apply only to an owner’s first and most valuable property, so some low- value properties are not exempt due to multiple property ownership. 16 In Figure A3, we present the average assessed value for properties in each percentile. While the median property is assessed at R$160,000, as previously discussed, the average value at the top 1 percent is R$10 million. 24 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE Women face lower effective rates than men properties that pay over 1.1 percent in effective since they own less valuable properties. Taken rates and thus face surcharges compared to the together, the facts that women own less valuable flat 1 percent rate. properties and that effective rates increase with Mirroring our exercise for gender disparities in assessed value would imply that men pay higher property valuation, in Table 3 we also present property taxes in São Paulo than women. In Figure regressions of tax liability on indicators of the 9 below we quantify that difference, assigning gender of the owner and other controls. In column taxpayers to bins of effective rates related to the (4) we show the overall gap is 20 percent, and marginal rates present in the tax schedule. First, we that is reduced by 3 to 5 percent once we include note that 32 percent of properties owned solely by property and area controls.17 These differences women are exempt from property taxes, versus 27 in tax liabilities across genders are important in percent for those solely owned by men. Women are aggregate: the aggregate liability for properties also 2 percentage points more likely to pay effective owned solely by women represent 14 percent of rates between 0 and 0.7 percent, which is the first total taxes, while those owned by men face more bracket of low but positive rates. For higher rates, than double the total amount or 32 percent of total properties solely owned by men are systematically property taxes. overrepresented: men are 3.6 percentage points (10.2 percent versus 6.6 percent) more likely to own Figure 9: Share of effective tax rate by gender 40 38.0% 34,4% Men Women 31,8% 30 27.3% 27,2% 24.6% Share (%) 20 10 8.9% 5,9% 1.3% 0 0,7% Exempt (0, 0.7] (.7, 1.1] (1.1, 1.5] > 1.5 TAX INTERVAL Source: Original calculations for publication. Note: This figure presents the share of properties, by gender, on each bin of effective tax rate, i.e., the tax liability divided by the property value. The sample is restricted to properties owned exclusively by men or women. 17 Since, with exemptions, a large share of properties face zero tax liability, we estimate a zero-inflated Poisson regression. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 25 4. Policy implications and conclusion In this report we use microdata on property cadasters in São Paulo, Brazil, to document gender gaps in property ownership, wealth, and tax liabilities. Our findings of large gaps in a setting with no de jure restrictions on female property ownership as the property owner within a couple of different sex might still disfavor women. Gaps in property ownership might also be the result of biases in other settings, such as in labor markets (where women are paid less for the same job or face barriers to invites several other questions. We briefly introduce entry into certain professions) or in the process here three important ones. of obtaining mortgage loans.18 Deere and Leon (2003) study gender gaps in land ownership across First, what are the causes of these gender gaps Latin America using survey data and discuss as in property ownership and wealth? While there important drivers of these gaps the role of gender have been no legal restrictions on female property norms within families and communities and biases ownership since 2002, current levels of ownership in state programs and the acquisition market. may still be partly explained by the legacy of those We see these as important avenues for further restrictive laws. Furthermore, gender-based norms research to understand the determinants of gaps in may change more slowly than legal provisions, and ownership of urban properties. current norms about who should be registered 18 We summarize the main features of Brazil mortgage loan markets in Appendix A3, including that approximately half of all purchases are made with subsidized credits from federal programs. 26 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE Second, what are the consequences of these Nonetheless, we should not assume this implies gender gaps in property ownership and wealth? One they have the same rights as the registered owner important dimension to note in the Brazilian setting in practice. Doss and Meinzen-Dick (2020) discuss is that, unless couples explicitly choose otherwise, the importance of robustness of property rights, the law establishes that all property acquired meaning “the extent to which they are enforceable during a legally recognized relationship belongs to when under threat.” Delays in court procedures, for both individuals, regardless of who is the owner example, can make the enforcement of conjugal on paper. While we are unable to assert whether 19 rights very costly. Furthermore, prior knowledge each property is owned by individuals who of these rights might also be absent: Bernardino are married, it is reasonable to assume that a (2021) documents in interviews with survivors of substantial share of the gap we observe is due to gender-based violence that often women do not men being the registered property owners within know they have property rights even if they are not a heterosexual couple. By law, if the property listed as owners and that they are “likely to lose was acquired during the relationship, most of their rightful share of property upon separation the women in these cases are also legal owners. and inheritance, while their attempts to claim and 19 Article 1.658 of the Civil Code describes the “Partial community property regime” as the default unless couples choose a different regime, with the two main alternatives being a full separation of all assets between spouses and the full joint-ownership between spouses. See, for example, https://swisscam.com.br/en/publicacao/doing-business-in-brazil/23-aspectos-sobre-o-direito-de-familia-brasileiro/. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 27 exercise property rights can trigger or aggravate attention from policy makers in Brazil: since 2009, gender violence.” These findings suggest that even the flagship housing program for low-income though marriage laws assure property rights for families encouraged registration in the name of the spouses, direct property ownership in cadasters female spouse (Law 11.9777/2009).20 A similar law, provides important assurance of those rights. giving priority to women as formal owners in housing programs, also exists in the State of São Paulo.21 A significant body of literature has shown that Similar policies to encourage property ownership ownership of assets plays a critical role in women’s by women exist elsewhere, but their effectiveness bargaining power, productivity, and safety (World is unclear. Awasthi et al. (2023) document that Bank 2012; Beegle, Frankenberg, and Thomas several cities in India provide discounts on stamp 2001; Alvarado et al. 2021). Studies focusing on duties and recurring property taxes when women granting women land ownership in rural areas have are registered as owners. In interviews with demonstrated positive effects on investment and stakeholders, the perception is that stamp duties child outcomes (D. Ali et al. 2015), while in urban encourage women to be registered as owners areas, insecure land and housing rights, particularly but do little to change the actual management of in slums, pose greater challenges for women, who properties, while discounts to recurring property tend to outnumber men living in these vulnerable taxes are ineffective. D. A. Ali et al. (2016) show conditions. Studies suggest that women living in that in informal settlements in Tanzania, small slums are disproportionately affected by multiple subsidies substantially increase the probability deprivations, including worse educational outcomes that households will register female members as compared to their non-slum-dwelling counterparts owners. Nonetheless, the authors recognize that and to men living in the same conditions (Azcona the impact on welfare outcomes for these women is et al. 2020). Despite suggestive evidence of the conditional on future enforcement of the legal rights disproportionate effects of asset ownership on involved. Clear understanding of the mechanisms women in the cities, data and conclusive evidence and magnitudes of the impact of such policies to on the mechanisms and size of these effects are encourage female ownership will be an important still lacking. step toward closing these gender gaps. Finally, given the diagnosis of existing gender gaps in property ownership and previous evidence on its implications, an important question is what policies can be enacted to encourage increased female property ownership. This topic has received 20 UN Habitat (2013) documents that in the 2009–2010 period, 80 percent of contracts were signed by female-headed households. The law also established that, in the case of divorce, formal ownership of the house would belong to women, but that specific rule has been successfully challenged in court several times as unconstitutional. 21 Originally Law 16.792/2018, incorporated into the broader Law 17.431/2021. 28 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE References Ali, Daniel, Klaus W. Deininger, Markus Goldstein, and Eliana La Ferrara. 2015. “Empowering Women through Land Tenure Regularization: Evidence from the Impact Evaluation of the National Program in Rwanda.” Development Research Group Case Study, World Bank, Washington, DC. http://documents.worldbank.org/curated/en/241921467986301910/Empowering-women-through-land- tenure-regularization-evidence-from-the-impact-evaluation-of-the-national-program-in-Rwanda. Ali, Daniel Ayalew, Matthew Collin, Klaus Deininger, Stefan Dercon, Justin Sandefur, and Andrew Zeitlin. 2016. “Small Price Incentives Increase Women’s Access to Land Titles in Tanzania” [in en]. Journal of Development Economics 123 (November): 107–22. https://www.sciencedirect.com/science/article/pii/S030438781630044X. Ali, Merima, Odd-Helge Fjeldstad, and Lucas Katera. 2017. “Property Taxation in Developing Countries” [in en]. CMI Brief 1. https://www.cmi.no/publications/6167-property-taxation-in-developing-countries. Alvarado, Gina, Boaz Anglade, Zachary B. Catanzarite, Jackeline Contreras, Cheryl R. Doss, J. Y. Suchitra, Abena D. Oduro, Hema Swaminathan, Jennifer Twyman, and Pilar Useche. 2021. ¿Casa propia? La autonomía económica de las mujeres en Ecuador [in spa]. Edited by Carmen Diana Deere. Serie Savia. Quito: Ediciones Abya Yala$pEditorial FLACSO. ISBN: 978-9978-67-577-9 978-9942-09-756-9. Avenancio-León, Carlos F., and Troup Howard. 2022. “The Assessment Gap: Racial Inequalities in Property Taxation.” Quarterly Journal of Economics 137, no. 3 (August): 1383–434. Awasthi, Rajul, Katie Pyle, Namita Aggarwal, and Parvina Rakhimova. 2023. “Gender-Based Discounts on Taxes Related to Property: Role in Encouraging Female Ownership—A Case Study of Indian States and Cities” [in en]. Policy Research Working Papers, World Bank, Washington, DC. January 2023. http://elibrary.worldbank.org/doi/book/10.1596/1813-9450-10287. Azcona, Ginette, A. Bhatt, S. Duerto Valero, and T. Priya. 2020. “Harsh Realities: Marginalized Women in Cities of the Developing World.” UN Habitat Spotlight on SDG11, UN Women, New York. https://data.unwomen.org/publications/spotlight-sdg11-harsh-realities-marginalized-women-cities-developing-world. Beegle, Kathleen, Elizabeth Frankenberg, and Duncan Thomas. 2001. “Bargaining Power Within Couples and Use of Prenatal and Delivery Care in Indonesia” [in en]. Studies in Family Planning 32 (2): 130–46. https://onlinelibrary.wiley 4465.2001.00130.x; https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1728-4465.2001.00130.x. Bernardino, Raquel Ludermir. 2021. “Housing for Survival: Insecurity of Tenure, Property Loss and Domestic Violence Against Women in Recife.” Master’s thesis, Universidade Federal de Pernambuco, Recife, Brazil. Berry, Christopher R. 2021. “Reassessing the Property Tax” [in en]. SSRN Scholarly Paper, Rochester, NY. March 2021. https://papers.ssrn.com/abstract=3800536. Cabral, Marika, and Caroline Hoxby. 2012. “The Hated Property Tax: Salience, Tax Rates, and Tax Revolts.” NER Working Paper, National Bureau of Economic Research, Cambridge, MA. November 2012. https://www.nber.org/papers/w18514. Carvalho Junior, Bruno Pedro Humberto. 2017. “Property Tax Performance and Potential in Brazil.” PhD diss., University of Pretoria, South Africa. Carvalho Junior, P. H. B., and C. De Cesare. 2022. “Who Pays the Property Tax? Evidence from Brazil” [in en]. Land Use Policy 120 (September): 106235. Coelho, Maria Delgado, Aieshwarya Davis, Alexander D. Klemm, and Carolina Osorio Buitron. 2022. “Gendered Taxes: The Interaction of Tax Policy with Gender Equality.” IMF Working Papers, International Monetary Fund, Washington, DC. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 29 Deere, Carmen Diana, and Magdalena Leon. 2003. “The Gender Asset Gap: Land in Latin America” [in en]. World Development 31, no. 6 (June): 925–947. https://linkinghub.elsevier.com/retrieve/pii/S0305750X03000469. Doss, Cheryl, and Ruth Meinzen-Dick. 2020. “Land Tenure Security for Women: A Conceptual Framework” [in en]. Land Use Policy 99 (December): 105080. https://www.sciencedirect.com/science/article/pii/S0264837719310853. Dray, Sacha, Camille Landais, and Stefanie Stantcheva. 2022. “Wealth and Property Taxation in the United States” [in en]. Working Paper 31080, National Bureau of Economic Research, Cambridge, MA. March 2023. DOI 10.3386/w31080. Fioravante, Dea, and Bernardo Alves Furtado. 2018. “Crédito Imobiliário” [in pt-BR]. In Financiamento do desenvolvimento no Brasil, edited by João Alberto De Negri, Bruno César Pino Oliveira de Araújo, and Ricardo Bacelette. Instituto de Pesquisa Econômica Aplicada (IPEA). https://repositorio. ipea.gov.br/handle/11058/8820. Gaddis, Isis, Rahul Lahoti, and Hema Swaminathan. 2022. “Women’s Legal Rights and Gender Gaps in Property Ownership in Developing Countries.” Wiley Online Library, Population and Development Review 48 (2): 331–77. Holden, Stein T., and Mesfin Tilahun. 2020. “Farm Size and Gender Distribution of Land: Evidence from Ethiopian Land Registry Data” [in en]. World Development 130 (June): 104926. https://linkinghub.elsevier.com/retrieve/pii/S0305750X20300528. Knebelmann, Justine. 2022. “Digitalisation of Property Taxation in Developing Countries” [in en]. ODI Report. London: ODI. Komatsu, Hitomi, Alemayehu A. Ambel, Gayatri Koolwal, and Manex Bule Yonis. 2021a. “Gender and Tax Incidence of Rural Land Use Fee and Agricultural Income Tax in Ethiopia” [in en]. Policy Research Working Papers, World Bank, Washington, DC. June 2021. http://elibrary.worldbank.org/doi/book/ 10.1596/1813-9450-9715. Komatsu, Hitomi, Alemayehu A. Ambel, Gayatri Koolwal, and Manex Bule Yonis. 2021b. “Gender Implications of Rural Land Use Fee and Agricultural Income Tax in Ethiopia” [in en]. Brief, World Bank, Washington, DC. June 2021. https://openknowledge.worldbank.org/handle/10986/36029. Kotikula, Aphichoke, and Wameq Azfar Raza. 2021. “Housing Ownership Gender Differences in Dhaka, Bangladesh” [in en]. Land Use Policy: Land Governance and Tenure Security at Scale 110 (November): 104983. https://www.sciencedirect.com/science/article/pii/S0264837719307720. Matos dos Santos, Claudio Hamilton, Ana Carolina Souto Valente Motta, and Monise Estorani de Faria. 2020. “Estimativas anuais da arrecadação tributária e das receitas totais dos municípios brasileiros entre 2003 e 2019.” Carta de Conjuntura 48–3o Trimestre. https://www.ipea.gov.br/cartadeconjuntura/index.php/2020/07/estimativas-anuais-da-arrecadacao-tributaria-e-das-receitas-totais-dos- municipios-brasileiros-entre-2003-e-2019/. Okunogbe, Oyebola. 2021. “Becoming Legible to the State: The Role of Detection and Enforcement Capacity in Tax Compliance” [in en]. Policy Research Working Papers, World Bank, Washington, DC. November 2021. https://doi.org/10.1596/1813-9450-9852. Piketty, Thomas, and Emmanuel Saez. 2012. A Theory of Optimal Capital Taxation. Working Paper, National Bureau of Economic Research, Cambridge, MA. April 2021. https://www.nber.org/papers/w17989. Saez, Emmanuel, Gabriel Zucman, and Thomas Piketty. 2022. “Rethinking Capital and Wealth Taxation.” Draft. https://eml.berkeley.edu/~saez/PikettySaezZucman2022RKT_v5.pdf. Sepulveda, Cristian F., and Jorge Martinez-Vazquez. 2012. “Explaining Property Tax Collections in Developing Countries: The Case of Latin America.” ECON Publications, Dept. of Economics, ScholarWorks @ Georgia State University, Georgia State University, Atlanta, GA. UN Habitat. 2013. “Scaling-Up Affordable Housing Supply in Brazil: The ‘My House, My Life’ Programme.” UN Habitat, Nairobi, Kenya. https://unhabitat.org/sites/default/files/download-manager-files/Scaling-up%20Affordable%20Housing%20Supply%20in%20Brazil.pdf. World Bank. 2012. World Development Report 2012: Gender Equality and Development. Washington, DC: World Bank. World Bank. 2022. Women, Business and the Law 2022. Washington, DC: World Bank. https://openknowledge.worldban k.org/handle/10986/36945. Zucman, Gabriel. 2019. “Global Wealth Inequality.” Annual Review of Economics 11 (1): 109–38. https://doi.org/10.1146/annurev- economics-080218-025852. 30 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 31 A.1 Determining the sex of property owners In this section we describe our method of determining the sex of each property owner in our database. Our first step is to extract the first word of the owners’ names, which is the (first) given name in Brazil. In Table A1 we list ten of the most common owners’ names: five we assign as female (Maria, Ana, Marcia, Vera, and Sandra) and five we assign as male (Jose, Antonio, Joao, Luiz, and Carlos). For example, of approximately four million unique owners, we observe that almost 200,000 (6 percent of total owners) are named Maria and over 160,000 (5 percent) are named Jose. Table A1: Most common names by gender Name Obs. Share Ratio appearing female Assigned gender MARIA 186,003 0.057 0.997 F ANA 32,345 0.010 0.997 F MARCIA 14,078 0.004 0.997 F VERA 12,147 0.004 0.998 F SANDRA 11,558 0.004 0.998 F JOSE 166,783 0.051 0.004 M ANTONIO 91,492 0.028 0.003 M JOAO 61,307 0.019 0.004 M LUIZ 48,877 0.015 0.004 M CARLOS 43,669 0.013 0.004 M Source: Original calculations for publication. In a second step, we match each of the names to a database containing all first names encountered in the 2010 Brazilian Census. For each name, the database includes information on the share of individuals with that first name that declare their sex as male or female. This allows us to compute the share of individuals with a given first name identified as female. In column (4), we present these shares for each name. For each of the top five female names, over 99.5 percent of individuals with that name identify as female. Conversely, for first names like Jose, Antonio, and Joao, less than 0.5 percent of individuals identify as females. Our last step is then a rule-based decision to assign sex to first names (and therefore to owners with that first name): if over 50 percent of individuals on the Census with that name identify as female, we assign their sex as female; otherwise, we assign their sex as male. All sex-based statistics presented in this report derive from this process. We recognize this is not a perfect process, but we argue the sex-assignment errors that certainly exist in our data (male owners we identify as female and vice versa) are likely to be small and to have little effect on aggregate statistics. Figure A1 presents a histogram, across all first names in our property-ownership database, of the share of individuals in the Census identifying as female. The distribution is highly bimodal: similar to the most common names we list above, for over 80 percent of first names, 99 percent or more of individuals are identified as one sex. Not only are the vast majority of names overwhelmingly used only by one sex, but that is particularly true for popular names: first names for which 99 percent or more individuals identify as one sex represent 93 percent of all individuals in our database. 32 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE Figure A1: Density of frequency of name assigned as female 30 20 Density 10 0 0 25 50 75 100 Frequency assigned as female Source: Original calculations for publication. Note: This figure presents a histogram of the share of individuals identified as female in the 2010 Census, across all first names of property owners in São Paulo in 2018. It documents that for over 40 percent of first names, the share of females is above 99 percent, while for another 40 percent the share of females is below 1 percent; that is, these first names are overwhelmingly associated with individuals from one sex. While the vast majority of owners have first names that are clearly gendered, and therefore unlikely to be misclassified, a small share of owners have names that in the Census data belong to individuals declaring themselves male or female in similar proportion. In Table A2 we exemplify some of these names, presenting the most popular first names for which the ratio of female individuals in the Census is between 45 and 55 percent—meaning if these names are classified as female, our guess at the correct gender will be as close possible to random. The key finding from that table, consistent with previous results, is that these nongendered names are quite uncommon. The two most popular among them are “Darcy,” which we classify as male because 49.3 percent of Census respondents with that name are female, and Juraci, which we classify as female because 54 percent of respondents are female. Taken together, individuals with these top 10 names represent slightly more than 3,500 property owners in our sample, or only 0.1 percent of the total number of owners. Taken together with the previous evidence presented, we argue that this suggests our results are unlikely to be biased by our inability to perfectly assign sex to every owner in the database. Table A2: Nongendered names Name Obs. Share Ratio appearing female Assigned gender JURACI 845 0.0003 0.546 F YOUNG 218 0.0001 0.503 F EDIR 150 0.0000 0.504 F EDY 120 0.0000 0.544 F TSAI 118 0.0000 0.507 F DARCY 1,249 0.0004 0.493 M LIN 407 0.0001 0.471 M ELY 286 0.0001 0.481 M LAIR 188 0.0001 0.455 M DIOMAR 132 0.0000 0.460 M Source: Original calculations for publication. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 33 A.2 Calculation of property taxes in São Paulo Figure A2 shows the prefilled tax form presented yearly to property owners in São Paulo, containing the main fields used to calculate tax liabilities. The overall property tax is composed of two parts: a property tax on construction and one on land. The tax base for the former is the sum of assessed value of a constructed area and the assessed value of land incorporated by the building (which varies according to areas of the city and also takes into account factors like the perimeter of the area, the type of terrain, and whether the property provides single-family or multifamily housing). The tax base for land is “excess area,” a measure unconstructed area. The sum of these two components determines whether a property is eligible for an exemption or discount. For example, in 2018, residential properties with a value below R$160,000.00 were exempt from the IPTU, while those with a value between the exemption threshold and R$320,000.00 received a linear discount. Similarly, nonresidential properties were exempt if their value was less than R$90,000.00, and they received a similar discount, measured by the difference between their value and R$180,000.00. Finally, the land tax is only calculated if the total area exceeds the constructed space and the land site incorporated by the building; the land tax is measured by the following formula Min(ηSZU · Occupied area,Total Area), which is not subject to exemptions or discounts. After taking into account any exemptions and discounts, the IPTU tax rate is applied progressively based on the values of the properties within each threshold. For residential properties, the base tax rate is 1 percent, while for nonresidential properties, it is 1.5 percent. The discounts and increases are then applied accordingly to determine the final tax amount. Table A3 shows the specific discounts and increases for each threshold. It is important to note that the final value of the IPTU tax also depends on the tax levied the previous year. For instance, in 2018, limits were placed on the increase of nominal tax values. Residential properties were not permitted to increase their nominal tax value by more than 10 percent, while nonresidential properties had a limit of 15 percent. Table A3: Tax discounts/increases by property value in 2018 Valor Venal Residential Nonresidential (-) ≤R$150,000.00 -0.3 p.p. -0.4 p.p. 0 R$150,000.00 < Val ≤ R$300,000.00 -0.1 p.p. -0.2 p.p. 300 R$300,000.00 < Val ≤ R$600,000.00 0.1 p.p. 0 p.p. 900 R$600,000.00 < Val ≤ R$1,200,000.00 0.3 p.p. 0.2 p.p. 2,100 ≥ R$ 1,200,000.00 0.5 p.p. 0.4 p.p. 4,500 Source: Secretaria Municipal da Fazenda São Paulo 34 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE Figure A2: Tax form for property tax in São Paulo EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 35 A.3 Credit markets for property acquisition in Brazil Mortgage lending in Brazil is dominated by government-regulated programs with concessional terms. According to data from the Central Bank, “regulated” credit represents approximately 90 percent of total mortgage credit flow every year. The main program under that umbrella is the Housing Financial System (SFH), originally created in 1964. This program offers credit at reduced interest rates exclusively for residential purchases and is funded in large part by mandatory contributions from formal workers (FGTS) (Fioravante and Alves Furtado 2018). SFH loans are capped at 80 percent of the property’s assessed value and cannot be used to purchase properties above R$1.5 million. The second source of mortgage financing is the Real Estate Financing System (SFI), where interest rates are market-based and loans can be used for either commercial or residential properties; SFI places no cap on property value. Mortgage-financed transactions represent approximately half of all property transactions in São Paulo. Unlike mortgage operations, which are closely tracked by the Central Bank, data on the universe of property transactions is scarcer, making it more difficult to estimate precisely the share of cash-only versus mortgage- financed transactions. The Regional Board of Realtors of the state of São Paulo publishes monthly surveys on the nature of property transactions and estimates that approximately 50 percent are financed.* That magnitude is consistent with microdata from the property transaction tax (ITBI) available for the municipality of São Paulo between 2019 and 2022: 40 percent of the properties paying ITBI during this the period were financed by the SFH system. Considering that we do not observe in that data the smaller share of properties not financed by SFI, and that properties in some SFH programs are exempt from ITBI and therefore missing from our data, the share of actual residential transactions with a mortgage is likely closer to 50 percent. A.4 Additional Figures and Tables Figure A3: Mean property value per percentile 10,000,000 Mean property value (R$) 1,000,000 100,000 10,000 0 25 50 75 100 Percentile Source: Original calculations for publication. Note: This figure presents the mean assessed property value for all properties and for each percentile of property value. The Y-axis is presented in logarithmic scale. * Reports are available at https://www.crecisp.gov.br/comunicacao/pesquisasmercado. 36 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE Figure A4: Share of ownership per percentile of property value (a) All properties 1.0 Women Share of owners Men 0.75 Mixed Men/Women Private Firm 0.50 Public Firm 0.25 Percentile size 0.00 68284 properties 0 25 50 75 100 Percentile of total property value (b) Residential properties only 1.0 Women Share of owners Men 0.75 Mixed Men/Women Private Firm 0.50 Public Firm 0.25 Percentile size 0.00 57896 properties 0 25 50 75 100 Percentile of total property value Source: Original calculations for publication. Note: This figure presents the share of ownership per percentile of all properties in panel (a) and of residential properties in panel (b). Designations of Women or Men are considered only when one gender owns the property. If a property is split between men and women, ownership is considered mixed. EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE <<< 37 38 <<< EQUITABLE GROWTH, FINANCE & INSTITUTIONS NOTE