The Role of Housing Finance in Mexico's Vacancy

2 downloads 0 Views 820KB Size Report
periphery of Mexican cities were still empty and journalists honed in on the plight of the people living in these ... market imperfections, and low holding costs due to low property taxes. ...... deductions, as well as business tax. Property taxes are ...
The Role of Housing Finance in Mexico’s Vacancy Crisis

Paavo Monkkonen Assistant Professor of Urban Planning UCLA Luskin School of Public Affairs 3250 Public Affairs Building, Box 951656 Los Angeles, CA 90095-1656 E-mail: [email protected] Telephone: (310) 482-7733

Abstract In the wake of the housing market crash in the United States in the late 2000s, images of abandoned homes on the urban periphery of American cities dominated international media coverage. This narrative of peri-urban over-extension was used by media documenting the housing crisis in Mexico, despite the profound differences in context, namely the role of the government in housing finance. This paper disentangles the issue of Mexican housing vacancy from surface similarities with the US through an examination of vacancy rates within cities in Mexico, and tests of four hypotheses about their determinants using data from the 2010 Census of Population and Housing. Results confirm that violence related to the drug war, international migration, and housing finance are associated with vacancy. However, more housing finance is strongly related to higher vacancy in the central city but not in the urban periphery. In spite of the existence of vacancy in newly built houses, the expansion of credit for new housing in Mexico has been most significant for the role it has played in hollowing out the central city. The paper concludes with a review of policies to address the vacancy crisis in Mexico and a framework of best practices.

1

A. Introduction When the United States housing market crashed in the late 2000s, images of abandoned foreclosed homes became the dominant symbol of the financial catastrophe. Photos of large tract homes standing empty seemed to capture the excess of a private market run amuck. The images were often accompanied by poignant stories of embattled homeowners trying to keep communities intact in the face of neighborhood disintegration, and other news stories captured the ghostliness of brand new housing developments stopped mid-construction before a single resident had moved in. Given the international scope of such coverage, it was perhaps not surprising that when Mexico’s exurban housing also began to show high rates of vacancy in 2013, news coverage followed this well-established narrative (Economist 2013). Newspaper articles in the United States in particular focused on how newly built communities on the periphery of Mexican cities were still empty and journalists honed in on the plight of the people living in these abandoned communities (Eulich 2013, Levin et. al. 2013, Guthrie 2013). The similarities between the American and Mexican housing markets are not false, but relying on the American housing crisis as an explanatory model ultimately paints a misleading picture of the issues confronting the Mexican housing market. The presence in both countries of speculative home construction outside of city boundaries and similarly high rates of housing vacancy should not obscure the fact that housing problems in Mexico and the United States occurred within very different economic and policy contexts. Perhaps the most critical difference is that the high vacancy rates in the United States were caused by a lack of government involvement and regulation (Immergluck 2011), yet in Mexico it was the government’s significant presence in the housing finance system that proved problematic. This paper analyzes vacancy rates across large urban areas in Mexico in order to better understand the causes of this vacancy. The analysis indicates that a large share of vacant houses in Mexico are not foreclosures nor are they abandoned due to an inability to pay a mortgage. Nonetheless, it is linked to public investment in new housing, in addition to other factors such as violence and international migration. There is compelling evidence that government-backed housing finance is strongly related to higher vacancy rates in central cities but not in the urban periphery, where the majority of newly built and mortgage-financed houses are located. Thus, the government’s expansion of credit for new housing ultimately had the direst effects on central city areas, contributing to their hollowing out. The second part of this paper assesses proposals to address the issue of housing vacancy and offers a policy framework for reform. Given the heterogeneity of causes for housing vacancy in Mexico, addressing the issue will require multiple interventions by different actors, including federal, state, and local levels of government as well as INFONAVIT, the government run housing fund. The government has an important role to play in housing by allowing asset markets to function, but instead of permitting these asset markets to direct development, as occurred during Mexico’s recent failed peri-urban expansion, national housing agencies should focus on facilitating individual choice.

2

B. Background on housing vacancy Every housing market, even those that are very active, has a share of housing sitting vacant at any given point in time. This occurs for primarily two reasons, the first of which has to do with construction and upgrading. Newly built housing is not occupied immediately and owners often do not live in housing units that are being remodeled or upgraded. The second structural cause of vacancy that occurs is linked to unavoidable friction during the process of buying, selling and renting. Sellers of houses often move out during the sale process or before new owners move in, rental properties sit vacant after a tenant moves out and the owner searches for a new one, and sometimes property is purchased solely as an investment. Vacancy in excess of the structural vacancy rate that creates important externalities (Rosen and Smith, 1983). It reduces the economic productivity of urban areas and leads to excess investment in infrastructure to service unnecessary urban expansion. Moreover, recent decades have revealed the negative environmental impacts associated with urban sprawl, a phenomenon that is exacerbated by vacant houses in central cities that would otherwise be occupied by tax-paying households playing a part in neighborhood communities. Moreover, there are pernicious community level impacts of vacancy, such as crime, a lack of social cohesion, and decreased property values. Some scholars have even argued that the negative environmental impacts of excess housing vacancy overwhelm any mitigating efforts of sustainable building regulations (Sun et al., 2011). The average vacancy rate internationally was roughly 10 percent in 2010 (Torres, 2012). Developing countries often have higher vacancy rates than wealthy countries (Struyk, 1988), for a number of reasons. Rosen and Smith (1983) describe a model of the rental housing market in which the vacancy rate is determined by the cost of holding inventory, search costs, demand for space in given locations, and the costs of changing contracts, and within this context developing countries operate in a market of high transactions costs, contract enforcement problems, credit market imperfections, and low holding costs due to low property taxes. These issues compound to result in vacancy rates that are higher than countries with both lower transaction costs and higher holding costs. Mexico, despite not being a poor country, reported a vacancy rate of over 14 percent in the 2010 Census. This is 4 percentage points higher than the international average, and points to significant shortcomings in the housing market. By comparison, in the United States rental housing unit vacancy has generally varied from five to ten percent since 1960, and has been much lower still for ownership housing (2 percent) (US Census, 2013). Given that the vacancy rate in Mexico is higher than the structural rate for a country of its income level, efforts to reduce housing vacancy will yield benefits to all Mexicans in terms of increased access to housing, improved urban efficiency, and greater economic productivity. Yet in order to effectively reduce the vacancy rate, a better understanding about the multiple causes of the phenomenon is needed. Mechanisms by which the housing finance system in Mexico would lead to high vacancy rates exist on the demand side and supply side. The demand component has long been influenced by the fact that all formal, salaried employees are members of Instituto del Fondo Nacional de la Vivienda para los Trabajadores (INFONAVIT), the national housing provident fund (for more 3

on INFONAVIT see Monkkonen (2011a)). As members, they are required to contribute five percent of their salary to the fund, which meant that until recent changes in rules salaried employees had a strong incentive to take out a loan and purchase a house. Buying a home was the main way to for employees to benefit from their mandatory contributions, and thus they had incentive to do so even if they did not intend to move. Additionally, loans from INFONAVIT are issued with below market interest rates and for this reason it is common for members to purchase a house as an investment, perhaps without fully considering the financial implications. Yet there are also causes of vacancy on the supply side. Developers of housing in Mexico have taken advantage of the expansion of INFONAVIT to build subdivisions that might not have services or retail stores until years after they are completed, making their habitation a challenge (Monkkonen, 2011a). Absentee owners and speculation in real estate are common reasons given anecdotally for vacant housing, reasons which stem from a combination of low holding costs, limited investment opportunities, and difficulties in renting on the other. Real estate exists in two types of markets; the market for assets and the market for physical space (DiPasquale and Wheaton, 1996). In the former, it competes with stocks, bonds, and other investment vehicles and is national or even global in scope, whereas in the latter it is extremely local as the demand for physical space in a city depends principally on that city’s labor market. Additional possible causes include housing belonging to or being saved for families working outside the country, and, in rural or other low-demand areas, some units are held for use as storage or have simply been withdrawn from the housing stock, because of a lack of demand for space in that area (Struyk, 1988). In addition to the supply and demand elements, identifying where the problem is concentrated is also an essential component of understanding the root causes of housing vacancy in Mexico and working towards solutions. For instance, although vacant houses in new peri-urban developments are the most notorious type of housing abandonment in Mexico, this type of vacancy is not necessarily the most prevalent. A report on territorial development of PueblaTlaxcala found that vacancy rates in the city center reach over 40 percent (OECD, 2013: 134), and the phenomenon of inner-city housing vacancy, which was previously hinted at by Ward (2001), signals a dearth of attention towards the problems generated by both irregular urbanization from decades prior and a lack of attention to the redevelopment and upgrading of these inner-suburbs. The deteriorated urban environment contributes to the vacancy problem as it lowers the demand for real estate in neighborhoods in relatively central parts of the city (Monkkonen, 2008), and a more detailed examination of Mexican vacancy patterns needs to account for such trends. There are several important questions about vacant housing for Mexican policymakers and it is important to tell the story of Mexico’s housing crisis without relying on a narrative that comes from the United States. To address the lack of knowledge on this topic, I assess the spatial variation in vacant housing both within and across the 100 largest urban areas 1 in Mexico using the year 2010 Census of Population and Housing. I do this in two stages. First, I examine where 1

In this report city and metropolitan area are used interchangeably to refer to urban areas that in some cases include multiple municipalities. I define the 100 largest cities (or metropolitan areas) using urban population numbers from the year 2000. This definition includes all 59 metropolitan areas defined CONAPO/INEGI/SEDESOL (2012), as well as 19 of the conurbations and 20 urban centers. 4

vacant housing is located within cities, and what kinds of neighborhoods have high vacancy rates. Second, I use basic statistical models to assess the kinds of cities that have high vacancy rates overall, in their urban periphery, and in their center. These models will allow me to test a number of hypotheses about the determinants of high vacancy rates and the relative importance of these factors.

C. Where are the vacant houses in Mexico? Roughly 21 million of the 35 million housing units in Mexico in 2010 were located in these 100 cities. Somewhat surprisingly, the 100 largest cities contain an almost exactly identical proportion of vacant housing units (14 percent of stock) relative to the vacancy rate for the country as a whole. In order to better understand the causes of vacancy, I first measure vacancy rates within cities. Vacancy rates are calculated for the central city area and the urban periphery part of the city using the number of vacant housing units by census tract. 2 First, I calculate the distance of every census tract from the city center 3. The 25 percent of census tracts closest to the center of the city are classified as inner-city and the 25 percent furthest from the center are classified as peri-urban. Figure 1 shows the city of Aguascalientes, with census tracts categorized in this manner.
Table 1 reports the average share of vacant housing in the inner-city and peri-urban parts of cities in two different yet highly relevant cuts. The first (columns two and three) is the average vacancy rate for census tracts in the inner-city and urban periphery. The second (columns four and five) is the share of the city’s vacant housing that is located in the inner-city and urban periphery. This distinction is important because although on average the vacancy rate is higher in the peri-urban parts of the city, inner-city areas have on average a much larger share of these cities’ vacant housing. This is due to the fact that there are more houses in the central parts of cities and census tracts have a higher housing unit density. The standard deviations reported in Table 1 show that there is comparably higher variation in the peri-urban vacancy rates than those in central cities. Although the supposition that the vacancy housing phenomenon is predominately a problem in the urban periphery is in some ways correct, this is not true everywhere. In 64 of the 100 largest cities in Mexico, vacancy rates are higher in the peri-urban parts of the city than they are in the city overall. However, in 10 of these 64 cities the difference is relatively small (less than 10 percent), and in cities where peri-urban vacancy rates are higher than the city’s overall vacancy rate, the peri-urban rates are are on average 30 percent higher. 2

In Mexico, the census tract equivalent is referred to as a ‘basic geo-statistical area’, or AGEB for its initials in Spanish. 3 The city center is defined by the author using a combination of criteria; the geographic center for cities that are mostly circular, the historic center in many other cases such as port cities or cities along the USMexico border. The former definition generally matches the latter. 5

These city statistics show that when looking at the actual numbers of houses rather than rates of vacancy, inner-city housing vacancy is a problem of greater magnitude than peri-urban vacancy. Roughly one third of vacant housing in the 100 largest cities is located in the inner-city. And vacancy rates in the urban periphery are lower than those of the city overall in a third of large cities. In eight cities, vacancy rates are higher in the inner-city. Another way to examine the spatial distribution of housing vacancy is the extent to which vacant houses are clustered in certain census tracts as compared to being evenly spread across the city. Thresholds have been found to be very important in research on neighborhood effects (Galster, 2012). Neighborhood characteristics such as problems with physical infrastructure, public services, or concentrations of poverty often do not have linear impacts on quality of life and social outcomes; rather, negative impacts appear only after there is more than a threshold amount. To examine patterns of clustering, I identify tracts where more than one quarter of housing units are vacant. Roughly 30 percent of vacant units are located in tracts with high levels of vacancy on average in the 100 largest cities, indicating that the phenomenon of vacant housing is, fortunately, relatively dispersed across cities. I then evaluate where these high-vacancy tracts are located within cities. On average, a much greater share of tracts are high vacancy in the periurban areas of cities – 40 percent – as compared to the inner-city – 17 percent. The fact that vacancy is more concentrated in neighborhoods in the urban periphery might explain some of the media focus on this particular type of vacancy, in spite of the fact that there are actually fewer vacant units there. Before assessing the determinants of high vacancies across cities, I explore descriptive correlations at the geographic level of the census tract. What kinds of neighborhoods have high vacancy rates? What kinds of neighborhoods are likely to have a cluster of vacant housing? Variables available for consideration are the population density of the tract, its distance to the center of the city, the average household size in the tract, the share of houses with no drainage, the share of houses with piped water, the share of people with health insurance from the Instituto Mexicano de Seguro Social (IMSS) 4, and the share of people over 18 with a high school degree (as a proxy for socioeconomic status due to the fact that data on income is not available at the tract level in the 2010 census). This analysis is useful to gain perspective on dynamics of vacancy within the city, although as will be discussed in the following section, most determinants of vacancy are factors that operate at the city level.
Table 2 reports the results of two regressions of vacancy on census tract characteristics. Both include fixed effects for the different cities in which tracts are located. The first column is an ordinary least squares (OLS) model and the second column reports results from a logistic regression. The dependent variable in the logistic regression takes a value of one if more than 25 percent of the houses in the tract are vacant and the model then predicts the probability of a tract 4

The IMSS is the main insurance plan for private sector employees and is a proxy not only for formal employment but also for access to housing loans through INFONAVIT. 6

being categorized as high vacancy. In this case, all coefficients have the same sign as the previous model, but it has less explanatory power. Neighborhoods (or census tracts) have less vacancy if they have a higher population density, and the same is true if they have larger households. This is expected and reflects the fact that there is more demand for space in those particular parts of the city. Similarly, there is less vacancy (as a share of housing units) in central parts of this city, which also reflects the greater demand for space in more accessible locations. The presence of higher levels of public infrastructure – measured by a combination of drainage, access to piped water, and electricity – is associated with more vacancy but the magnitude of the impact is very small compared to other factors. In terms of socioeconomic variables, the share of a neighborhood’s population with a high school degree is associated with lower levels of vacancy. This is not surprising, as wealthier places are more likely to have a more active real estate market. There is also a greater prevalence of vacancy in places where individuals are more likely to have access to health insurance from the IMSS. Due to the fact that access to IMSS generally is accompanied by membership to INFONAVIT, the association lends support to the notion that new housing developments supported by INFONAVIT mortgages have high vacancy rates.

D. Determinants of housing vacancy in Mexico Understanding the neighborhood make-up of vacancy is important but housing markets operate at the metropolitan level. Thus, I analyze factors associated with high levels of housing vacancy at the city level in this section. I do this not only for vacancy rates in the city overall, but also in the spatially disaggregated manner described above, using inner city and peri-urban vacancy rates. What can we expect to cause high levels of housing vacancy in a city? As described in the introduction, the major causes can be framed as excess supply or too little demand in the market for space of a given area, a greater level of demand for real estate as an asset than as a physical commodity, or high transactions costs in property rights. In the case of Mexico, there is, surprisingly, only one academic article on this topic (Sánchez and Salazar, 2011). The authors propose four explanatory hypotheses for high vacancy rates that are specific to the Mexican context; the economic recession that began in 2008, migration to the United States, violence associated with the ‘war on drugs’, and the housing finance policy of INFONAVIT. I measure these factors using data on changes in value added at the city level from the Economic Census of Mexico (1999 and 2009), data on international migration 5 from the 2010 Census of Population and Housing, data on drug-related homicides as recorded by the Crime Indicator Database for the Justice in Mexico Project at the Trans-Border Institute of the University of San Diego (Molzan et al., 2012), and publically available data from CONAVI 6 on the number of housing loans for new housing issued in that city between 2000 and 2010. For this 5

The actual variable measures the share of households in a city that have at least one member currently living abroad. This fails to capture total out-migration from a given city but is a good proxy as it is expected to be very highly correlated. 6 Available at: http://www.conavi.gob.mx/informacion-de-financiamientos-para-vivienda (last accessed, May 2014). 7

last variable, I divide the number of loans issued by the number of houses in the city in the year 2000 in order to measure the penetration of housing finance in a given place. In addition to these four hypothesized determinants, I also consider other variables as controls and possible determinants. The size of the city (number of housing units), its population growth rate, population density, and housing market characteristics such as the share that is self-built and the average household size 7. Other variables considered but ultimately deemed less useful due to the fact that they are composite measures, were an index of marginality and the ratio of urban expansion to population growth from 1980 to 2010 - as reported in SEDESOL (2012). An assessment of the relationship between the ratio of urban expansion to population growth from 1980 to 2010 (SEDESOL, 2012) and vacancy rates in the urban periphery, however, yields an important and somewhat surprising finding relevant to the policy discussion around urban sprawl in Mexico. The bivariate correlation is negative but barely significant. 8 Cities that are thought to have a greater amount of sprawl in fact have lower vacancy rates in their peri-urban areas specifically. This result likely has more to do with the measure- ratio of urban expansion to population growth- which is very much a product of differential rates of growth before the year 2000. The average ratio between 1980 and 2010 across the 90 largest cities is about four; however, the same measure taken between the year 2000 and 2010 is about 1.1. This means that the more rapid period of urban expansion was sometime in the 1980s or 1990s. An assumption that seems to underlie much of the popular coverage of housing policy in Mexico and to some degree the policy proposals by the Secretariat of Agricultural, Territorial, and Urban Development (SEDATU), is that Mexican cities are experiencing a phenomenon akin to urban sprawl in the United States. Yet this is not the case. In fact, suburban housing developments produced through INFONAVIT financing are of a much higher density than suburbs in the United States. More importantly, INFONAVIT financed developments are of a much higher density than the majority of peri-urban growth in Mexico, which in most cities is still dominated by low-density, incrementally built neighborhoods (Monkkonen, 2011b). For this reason, it is not surprising that cities with more INFONAVIT lending did not also have a more ‘sprawling’ pattern of urban growth in recent decades.
Table 3 reports summary characteristics of the variables included in the statistical analysis of vacancy rates. Table 4 reports the results of regressions of the measures of vacancy on the variables hypothesized to impact these rates. Three models are run; one for the measure of vacancy overall, vacancy in the peri-urban part of the city, and vacancy in the central city. Standardized coefficients, which are the same regression using variables that have been standardized to have a variance of one, are also presented in order to directly compare the importance of different variables. I do this because variables are measured differently; for 7

Household size serves to some extent as a proxy for the price of housing. Unfortunately, no reliable data on housing prices or rents are available in Mexico to our knowledge, the price index from the Sociedad Hipotecaria Federal (SHF) tracks changes not absolute differences between cities. 8 The correlation between the same ratio between 2000 and 2010 and vacancy rates is not statistically significant. 8

example, the average income in a city is measured in the natural log of pesos whereas economic growth is a percent change. Thus, it can appear that one is more important if a coefficient is larger when in fact this not the case. For example, the coefficient on the log of murders is 0.006 as compared to that for percent of households with migrants, which is 0.4, when in fact, the variable measuring narco-violence is relatively more important, as reflected by the standardized coefficient of 0.27 versus 0.26 for migration.
The regressions directly support three of the four hypotheses proposed by Sánchez and Salazar (2011) to explain vacancy. The prevalence of international migration from a city, measured by a proxy variable from the population census, is the most consistently significant variable across the three regressions, and has the strongest relationship with high vacancy rates in the urban periphery. There is more migration from poorer, smaller cities but once these factors are controlled for, it is migration that stands out as more closely associated with vacancy. This is likely due both to families’ abandoning houses as well as houses being built for or kept empty for the eventual return of those working abroad. The prevalence of murders associated with the drug war in Mexico – a measure of narco-violence – is positively associated with overall vacancy rates. There are a large number of anecdotal reports of houses in peripheral areas being used for activities related to the cartels and this would greatly reduce the attractiveness of a neighborhood to most families. Yet, the variable with the strongest relationship to high vacancy rates both city-wide and within central areas is the prevalence of housing finance from public agencies 9, as indicated by the standardized coefficients of 0.49 and 0.47. Although this is an expected relationship, it is stronger than anticipated. Moreover, given the widespread discussion of new, suburban houses lying vacant, it is surprising that housing finance is strongly associated with high central city vacancy rates but not significantly associated with peri-urban vacancy rates. How can this be when the vast majority of houses purchased with finance are located in the urban periphery? A likely explanation is that the large amount of peri-urban housing construction facilitates families moving out of central city housing, and the much less active market for used housing does not fill vacancies when they move out. This is a pattern somewhat akin to the suburbanization process in the United States in the second half of the 20th century. The fourth hypothesis proposed by Sánchez and Salazar (2011) – the economic recession – is not significantly associated with variation in vacancy rates at the city level. However, this does not mean that this hypothesis is rejected overall. The models presented here test the relationship between various factors and the variation in vacancy rates across cities. Since the recession hit the entire country it might not explain any variation in vacancy across cities but still be a contributing factor to vacancy nationally. Moreover, several factors that are significant, such as migration and violence, are closely related to economic conditions. Control variables like city size, population density, income levels, and economic growth are demonstrate a mild correlation with vacancy in a bivariate analysis, yet once other variables are 9

This result is robust to using only loans from INFONAVIT; the prevalence of lending from INFONAVIT in a given city is highly correlated (0.7) to that of lending from other public agencies. 9

included they lose statistically significant explanatory power in most cases. Household size shows a strong negative correlation with vacancy, indicating housing market tightness is associated with lower levels of vacancy. Cities with larger household sizes have less central city and peri-urban vacancy.

E. What is being done to address the problem? In this section, I outline policy efforts on the part of various Mexican housing and urban institutions. In the Programa Sectorial de Desarrollo Agrario, Territorial y Urbano, 2013-2018, only one action item explicitly addresses the vacant housing phenomenon (Strategy 3.4), and it is to deepen the understanding of vacancy in order to identify possible solutions. Nonetheless, many of the policy proposals related to urban management and investment in urban areas by SEDATU have the potential to address some of the root causes of housing abandonment. For example, the Programa Sectorial de Desarrollo Agrario, Territorial y Urbano, 2013-2018, includes several strategies to improve urban planning and connectivity of housing developments to services such as health centers, schools, and recreational facilities. These are important efforts for quality of life and would make new peri-urban housing projects more viable, but as documented in the analysis above, the majority of vacant units are not actually located in the urban periphery where access is a problem. Many of SEDATU’s proposed policies that seek to promote infill development also address housing vacancy. They would have indirect impacts, but potentially with great impact. This is especially true with respect to pushing housing loans to central parts of urban areas in combination with efforts to promote the purchase of used housing and stimulate the secondary housing market and housing improvement. Efforts to improve urban environments in new housing developments should also reduce the rate of vacancy in these places. Finally, the modernization of cadastres and public registries of property will facilitate local understanding of residential vacancy, but not directly reduce it. Given the heterogeneity of causes of housing vacancy, the question of where responsibility to address it lies should be addressed. INFONAVIT is well aware of the problem of housing vacancy and has taken several measures to address it. Their focus, however, is understandably on houses for which they financed the purchase, which is a fairly limited to a subset of all vacant housing in the country that are referred to as abandoned housing rather than vacant. Older, centrally located housing that is vacant falls within the purview of local governments, state housing agencies, and SEDATU, rather than INFONAVIT. Unfortunately, it does not seem that these entities are facing the problem with policies of a sufficient magnitude. The causes of abandonment and vacancy are different. Studies by INFONAVIT, based on data gathered through phone interviews and qualitative studies, identified one of the major causes of housing abandonment and foreclosure to be the location of houses (INFONAVIT, 2012). However, these studies by INFONAVIT also reveal that many more houses are vacant because individuals took loans simply because they were available to them, not because they wanted to move. INFONAVIT notes that the majority of owners of currently unoccupied houses do intend

10

to move eventually, and that they were even satisfied with their decision to purchase in spite of not currently living in the house. INFONAVIT has proposed two areas of reform: one set to prevent the continuation and increase in housing abandonment, and the other to deal with the vacant housing that is already part of its portfolio. They propose to prevent growing housing abandonment through three main mechanisms 10: improved access to information for members about purchasing to help them make the right decision, a more diverse set of options for members who are eligible to borrow money, and improved criteria for developers to build housing that meets the needs of members. Financial education is an extremely important public service, especially given the potential for developers and real estate agents to influence decisions of less educated individuals; however, if a clear financial incentive is pushing people to exercise their option to a subsidized house, they will take it. Increasing the options of INFONAVIT members once they are eligible for a loan will be a much more effective measure. Programs like home improvement loans or loans to purchase used housing should have a significant impact. The report by INFONAVIT (2012) contains much discussion of how to build housing that better meets members’ needs, including the goal of better matching the supply of loans to housing demand. This area is crucially important, but it is unclear what steps are being taken to achieve it. A proposal to stimulate ‘vertical’ housing in good locations by subsidizing construction finance actually runs counter to this goal. It changes the criteria for developers to ones that are ‘better’ from an urban planning perspective, but continues the model of financing the building of new housing based on criteria set by INFONAVIT rather than through market demand. Additionally, INFONAVIT has proposed various measures to ensure the urban environments of newly purchased houses do not deteriorate and to motivate families to move into currently vacant houses. One example is an insurance policy that guarantees the maintenance of various aspects of services and infrastructure in the local environment. Another initiative that has particular potential is the collection of property taxes and fees for maintenance of community facilities through monthly mortgage payments. This kind of policy could perhaps be more widely applied in Mexican cities; for example, linking property taxes to utility bills. During the second half of 2013, INFONAVIT introduced a pilot program to rehabilitate 20,000 abandoned homes. The Sale and Brokerage of Affordable Housing program (or VIVE for its initials in Spanish) facilitates the remodeling and sale of abandoned houses to INFONAVIT members at a discounted price. Private sector operators do the rehabilitation and marketing to resell houses, whereas INFONAVIT appraises the houses before and after the work. Federal agencies are not the only actors in housing and urban policy in Mexico, and I interviewed six state housing agencies (Baja California, Durango, the state of Mexico, Guanajuato, Hidalgo, Nuevo Leon and Puebla) to assess their understanding of the vacant housing problem, their opinion on its importance, and any policies they are implementing to address it. Not surprisingly, all six state housing agencies acknowledge the vacancy issue as a 10

Other ideas are proposed by INFONAVIT (2012), but are less important. For example, there is a proposal to have members live in a house for a period of time before actually purchasing it. Given the high costs of search and moving this does not seem to be an efficient solution. 11

crisis and consider it among the five or ten most pressing issues in the state. They present various explanations for the phenomenon, the majority of which reflect their understanding the problem as connected to new housing developments built under INFONAVIT lending. Several state agencies attributed vacancy to the location of new developments, the small size of houses, inability to make mortgage payments, and a lack of public safety. In general, attitudes towards policy responses reflect the association of housing vacancy with new housing developments. Various states mentioned ‘better’ rules for permitting new developments and ‘better’ urban planning. Yet, only two of the six states surveyed have programs for a policy response to the problem; Baja California and Guanajuato. Baja California’s Proyecto No. 62 is an effort to obtain ownership of over 10,000 foreclosed homes and provide a path to ownership for people in the informal economy. In the state strategic housing plan, Guanajuato includes an objective of rehabilitating vacant houses. Both of these policies are on hold until federal funding can be obtained to support them.

F. What should be done? There are four groups of polices recognized as best practices to address housing abandonment and vacancy. Although in Mexico violence and public safety were found to be significantly associated with variation in vacancy rates, these are causes that are outside the scope of housing and urban policy to some extent and not addressed explicitly here. The four groups of problems/policies are: •

• • •

Policies that would improve the urban environment, such as improving public services, upgrading deteriorated physical conditions, and/or making locations more accessible. These neighborhood improvement strategies are efforts to deal with the limited demand for space of a given area of the city. Land banking to return existing abandoned houses to use. Efforts to reduce speculative investment and the underuse of property in certain locations. Efforts to address problems with the mortgage allocation system.

Low demand in places with a deteriorated urban environment Policies to improve the urban environment can stimulate demand and address the limited demand for space of a given area of the city. This is a large area of urban policy and many tools are covered in the context of infill development later in this report. A high vacancy rate can be a crucial indicator for where to focus policy attention and public investment in a city. Given the nature of urban economies and the difficulties with redevelopment, it is generally cheaper to improve newly built parts of urban areas whereas greater benefit can be obtained by reviving older, centrally located neighborhoods. The latter are by definition more proximate to jobs and services, and residents can benefit from this access. Investment in physical infrastructure and public services in the inner-city can be a catalyst for private investment in housing. Ward (1993) argues that inner cities in Latin America generally retain a vibrant economy of small-scale artisan activities that limits the potential for large-scale 12

urban redevelopment and reinvestment. In light of this fact, he suggests policy should aim to improve on existing uses for the working-class population rather than attract new land uses. This is a challenge; however, as densification generally must consist of new construction, which will be at a higher quality than existing buildings and therefore, some degree of gentrification will likely occur. New policy approaches to address vacant housing have been developed in regions with shrinking cities, in many cases representing a marked shift away from the growth-centered planning paradigm that dominates the discipline (Weichmann and Pallagst, 2012). In Eastern Germany, for example, vacant housing has been a problem during the decades since reunification and the government has undertaken a strategic consolidation of neighborhoods and demolition, or ‘backwards-construction’ of the least commercially viable and worst located housing stock (Glock and Häussermann, 2004). A similar strategy is currently being proposed in the US city of Detroit, which has lost a large share of its population over the last several decades. These strategies, however, should be considered carefully for Mexico. Housing vacancy in most urban areas in Mexico will eventually be occupied as cities continue to grow. However, in regions with low population growth, which might include some small towns and rural areas, the destruction of vacant housing should be considered if it is associated with problems such as crime, or if the provision of services to a dispersed population is excessively costly. Land banks to deal with existing abandoned houses Municipal land banks have proven to be an effective strategy to deal with vacant and abandoned houses, especially those that have been foreclosed upon. Broadly, land banks are public agencies that acquire and prepare real estate for development or redevelopment, yet their actual function can vary a great deal. They have been most successfully employed in the North American context, in cities of the Midwest and Northeast to acquire tax delinquent, vacant and/or blighted properties in order to limit their negative externalities, stabilize property values, and assemble parcels for redevelopment (Alexander, 2008; Sage, 2009). This approach could be put to use in Mexico but would depend to some extent on the reform of current property tax practices. Addressing the issue of absentee owners and underused property One of the major causes of housing vacancy and underutilized real estate assets in urban areas in all of Latin America is a speculative investment. Investment in residential real estate is very common in developing countries due to relatively low holding costs (property taxes) and the often limited set of other investment opportunities. Additionally, because of limited financing, individuals often purchase land with the intention of building after saving for a number of years. Struyk (1988) analyzed the high vacancy rate in Jordan and argued for a two-pronged set of policies; on the one hand, policies should lower the rate of return on empty real estate relative to other investments, and on the other hand they should make renting out investment property more attractive. How can these be accomplished? The former can be accomplished primarily by increasing holding costs for real estate, i.e. property taxes. This is not only a good idea to incentivize property owners to develop land to its highest and best use, but it can also provide more funds to make investments in physical infrastructure and increase demand for space. The latter can be accomplished by policies that seek to stimulate the rental market. 13

The problems of low property tax rates and a dearth of collection are well-documented in Mexico and most of Latin America (Smolka and Furtado 2001). Low property tax rates and low collection is a not only a problem for local fiscal health, it also contributes to property vacancy and underutilization of space. Additionally, higher tax rates can be applied to vacant land and underused property. Feldstein (1977) argues that removing the tax on structures and exclusively taxing land values will increase investment in structures. In the United States, there is evidence that split rate taxation has led to a greater intensity of land use and higher densities (Banzhaf and Lavery, 2010), but in the cities of Baja California, which implemented a form of land value taxation in the early 1990s, the impacts are less clear (Monkkonen, 2013). It is difficult to identify impacts there because of low tax and collection rates (roughly 50 percent); uncollected taxes will not give any actor an incentive to change behavior. Stimulating and formalizing the rental sector would make renting out investment property more attractive to land owners as well as reduce vacancy. An international survey of policies to stimulate the rental market points out that Mexico suffers from two common problems; strong tenant protections and unfriendly taxation (Peppercorn and Taffin, 2013). Although laws protecting tenants’ rights were somewhat relaxed in the 1990s in Mexico, very long delays in resolving disputes between tenants and landlords- which can take up to several years- are a strong disincentive to renting out property. Investors in rental property pay income tax with some deductions, as well as business tax. Property taxes are deductible but interest on loans to build rental housing is not. As a result, a very small share of rental income is actually reported and taxed. Peppercorn and Taffin (2013) also note that attracting large-scale investors to the rental housing sector – through the above policies as well as the professionalization of property management and expansion of financing for multifamily buildings – would assist with goals of promoting compact cities by increasing the supply of multifamily buildings. Continuing to reform the mortgage allocation system of INFONAVIT This final set of policies are receiving the most attention, and these deal with the problems built into the mortgage allocation system of INFONAVIT. Most of these reform efforts (presented above) are laudable and if successful will address many of the problems in newly built housing developments. However, as Chiquier and Lea (2009) argue, there are several basic problems with housing provident funds. The core internal conflict is between their multiple functions: lender, pension fund, and subsidy provider. Savers subsidize borrowers because they are in effect lending them funds at below market interest rates, making the system somewhat regressive. Some of the new policy proposals, such as promoting the use of loans to purchase vacation homes, magnify this issue. Additionally, INFONAVIT’s dominance of the mortgage market hinders private lenders’ success, although co-financing programs alleviate this problem to some extent by taking advantage of the customer service INFOVANIT’s infrastructure and exposure. In spite of laudable proposals and actions to reduce the vacancy problem, the mortgage allocation system of INFONAVIT has two fundamental problems that must be addressed; i) the set of built-in incentives pushing members to take out a loan and ii) the close connection between developers and lenders for new housing that was set up initially to stimulate the supply of lowcost housing. Incentivizing members to take out loans seems to have led many households to 14

purchase a house they did not intend to occupy immediately. In a study by INFONAVIT, 13 percent of those who purchased a house through INFONAVIT reported doing so because they did not want to lose their opportunity to get a loan, while 18 percent reported having purchased a home as an investment (INFONAVIT, 2012). Increasing lending for home improvements and purchasing used houses will decrease the incentive to buy a new home solely as an investment. Construction lending associated with INFONAVIT mortgages should be reduced, and there should be a shift away from pre-construction guarantees of mortgage financing for new developments. This will encourage developers to build housing based on demand rather than criteria set by INFONAVIT or SEDATU. If mortgages and buyers are guaranteed, developers will not respond to the needs of purchasers and will be less likely to build housing that people want to live in. Creating a more demand based approach to the allocation and use of mortgages from INFONAVIT is one way to reduce future vacancy. Moreover, encouraging the use of mortgages to buy and upgrade ‘used’ housing will fill existing vacant properties.

G. Conclusion This article analyzes vacancy rates across large urban areas in Mexico and assesses proposals to address the issue of housing vacancy. The analysis shows that housing vacancy in Mexico is a multifaceted phenomenon, and although much attention has been paid to empty and abandoned new houses in peri-urban areas, and rightly so, there are actually many more empty houses in central parts of cities. Variation in vacancy rates across cities is found to be closely associated with violence and international migration, but especially with public investment in new housing. The analysis uncovers an unexpected result: cities with higher levels of mortgage lending have higher levels of vacancy in the city center but not the urban periphery. This suggests that because of the strong bias in lending towards new housing, populations have been drawn out of city centers, reminiscent of trends in the United States in the second half of the 20th century. Given the heterogeneity of causes for housing vacancy in Mexico, addressing the issue will require multiple interventions by different actors, including the three levels of government and INFONAVIT. INFONAVIT should continue to address the problem by rehabilitating and reselling existing inventory of abandoned and foreclosed homes. They should also continue to reform the mortgage allocation system that led to the abandonment in the first place, as discussed above, by making financing more flexible so that borrowers purchase housing they want to live in and are able to afford, and developers are more responsive to demand. The Mexican federal government, through SEDATU and other agencies, can play a major but relatively indirect role in the vacant housing crisis. Their efforts should be concentrated on directing and expanding the efforts of local governments rather than centralizing activity. Investment in efforts to improve the urban environment by expanding public services such as transportation networks and upgrading physical conditions should be undertaken by local governments with subsidies from the Federal government and through the increased collection of property taxes. Local governments already have the power to take such action yet are not doing so, and therefore the federal government should devise incentives in this area. Similarly, SEDATU should encourage local governments to test out policies to reduce speculative 15

investment in real estate that can lead to the underuse of property, of which higher and more consistently collected property taxes are an important first step. The fact that the United States and Mexico had similar rates of housing vacancy in 2010 is remarkable considering the very different economic and policy contexts in each country. In the United States, private companies were allowed to lend for housing in unsustainable ways to individuals who were not be able to afford their loans in the long run. In Mexico, the causes of vacancy are somewhat more complicated, but housing finance also played a major role. However, it was the government’s major presence in the housing finance system that is the problem, not its absence, and addressing the vacancy issue in Mexico underscores the importance of tailoring housing finance policy to the distinct conditions of individual countries. Mexico has taken steps to re-orient its loan policy away from peri-urban development and back towards central cities, but more systemic policy efforts are needed to truly make the housing market more responsive to the people of Mexico.

16

References Alexander, F. 2008. Land Banking as Metropolitan Policy. Brookings Institution. Banzhaf, H.S & Lavery, N. (2010) Can the land tax help curb urban sprawl? Evidence from growth patterns in Pennsylvania. Journal of Urban Economics 67, 169-179. Chiquier, L., and M. Lea, eds. (2009) Housing Finance in Emerging Markets. Washington, DC: World Bank. Cresce, A.R. (2012) Evaluation of Gross Vacancy Rates From the 2010 Census Versus Current Surveys: Early Findings from Comparisons with the 2010 Census and the 2010 ACS 1-Year Estimates. Social, Economic and Housing Statistics Division, U.S. Census Bureau, Working Paper Number 2012-07. DiPasquale, Denise and William Wheaton. (1996) Urban Economics and Real Estate Markets. Prentice Hall, New Jersey. Economist. Dropping a Brick (2013). [WWW document] URL http://www.economist.com/news /business/21578680-changing-government-policies-haveplunged-housebuilders-crisis-dropping-brick (accessed 21 August 2014) Eulich, W. (2013) In Mexico, Low-Income Homeowners Watch Their Dreams Crumble. Christian Science Monitor (Boston) [WWW document] URL http://www.csmonitor.com/World/Americas/2013/0626/In-Mexico-low-income-homeownerswatch-their-dreams-crumble (accessed 21 August 2014) Feldstein, M. (1977) The Surprising Incidence of a Tax on Pure Rent: A New Answer to an Old Question. Journal of Political Economy 85, 349-360 Galster, G. (2012) The Mechanism(s) of Neighbourhood Effects: Theory, Evidence, and Policy Implications. Neighbourhood Effects Research: New Perspectives, 23-56 Glock, B. and Häussermann, H. (2004) New trends in urban development and public policy in eastern Germany: dealing with the vacant housing problem at the local level. International Journal of Urban and Regional Research 28, 919–929. Guthrie, A. (2013) Mexico Housing Bust Bruises Investors, Buyers. Wall Street Journal (New York) [WWW document] URL http://online.wsj.com/news/articles/SB10001424127887323336104578503583127152320 (accessed 21 August 2014) Immergluck, D. (2011) The Local Wreckage of Global Capital: The Subprime Crisis, Federal Policy and High-Foreclosure Neighborhoods. US. International Journal of Urban and Regional Research 35. 130–146.

17

INFONAVIT. (2012) Vivienda deshabitada y abandonada: Dimensionamiento, causas, soluciones, medidas correctivas y preventivas instrumentadas. Levin, J. and B. Bain (2013) Mexico Housing Hits US Investors as Plan Collapses. Bloomberg News (New York) [WWW document] URL http://www.bloomberg.com/news/2013-1206/mexico-housing-hits-u-s-investors-as-plan-collapses.html (accessed 21 August 2014) Molzahn, C., Rios, V., and Shirk, D.A. (2012) Drug violence in Mexico: Data and analysis through 2011. Justice in Mexico Project. Trans-Border Institute. Joan B. Kroc School of Peace Studies. University of San Diego. San Diego, CA:. Monkkonen, P. (2008) Using Online Satellite Imagery as a Research Tool: Mapping Changing Patterns of Urbanization in Mexico. Journal of Planning Education and Research 2008 28, 225236. Monkkonen, P. (2011a) The Housing Transition in Mexico: Expanding Access to Housing Finance. Urban Affairs Review 47(5), 672-695. Monkkonen, P. (2011b) Do Mexican Cities Sprawl? Housing Finance Reform and Changing Patterns of Urban Growth. Urban Geography 32(3): 406-423. Monkkonen, P. (2013) The Impact of Land Value Taxation on Urban Development: The Experience of Mexicali and Baja California Norte. Mimeo. OECD. (2013) OECD Territorial Reviews: Puebla-Tlaxcala. Peppercorn, I.G., and C. Taffin. (2013) Rental Housing: Lessons from International Experience and Policies for Emerging Markets. Directions in Development. World Bank, Washington, DC. Rosen, K.T. and L.B. Smith. (1983) The Price-Adjustment Process for Rental Housing and the Natural Vacancy Rate. The American Economic Review 73(4), 779-786. Sage Computing Inc. (2009) Revitalizing Foreclosed Properties with Land Banks. Report to the United States Department of Housing and Urban Development. Sánchez, L. and C. Salazar. (2011) Lo que dicen las viviendas deshabitadas sobre el censo de población 2010 [What does vacant housing tell us about the population census of 2010]. Coyuntura Demográfica 1, 66-72. SEDESOL. (2012) La Expansion de las Ciudades en Mexico. 1980-2010. Smolka, M.O. & F. Furtado. (2001) Lessons from the Latin American Experience with Value Capture. Land Lines 13, 4. Lincoln Institute of Land Policy, Cambridge, MA. Struyk, R. (1988) Understanding High Housing Vacancy Rates in a Developing Country: Jordan. The Journal of Developing Areas 22(3), 373-380. 18

Torres, Y. (2012) Mexico is above the world average in vacant housing. The Economist July 26. URL http://eleconomista.com.mx/finanzas-publicas/2012/07/26/mexico-arriba-promediomundial-casas-deshabitadas (accessed February 1, 2013). US Census. (2013) Housing Vacancies and Homeownership section of the Current Population Survey, Table 1. United States Census Bureau. Sun, W., S. Xu and L. Li. (2011) Reducing Vacant Houses Is more Important than the Implementation of Green Buildings. Advanced Materials Research 280, 250-254. Ward, P. M. (1993) The Latin American inner city: differences of degree or of kind? Environment and Planning A 25(8), 1131 – 1160. Ward, P. M. (2001) The rehabilitation of consoli­dated irregular settlements in Latin American cities: towards a third generation of public policy housing analysis and development. Paper pre­sented at the ESF/N-AERUS Workshop: Coping with Informality and Illegality in Human Set­tlements in Developing Countries, Leuven and Brussels, May. Wiechmann, T. and Pallagast, K. M. ( 2012) Urban shrinkage in Germany and the USA: A Comparison of Transformation Patterns and Local Strategies. International Journal of Urban and Regional Research, 36, 261–280.

19

Tables Table 1. Vacancy rates and shares of vacant housing in inner-city and peri-urban areas Vacancy rate Variable Percent Standard Deviation

Share of city’s vacant housing

Inner-city

Peri-urban

Inner-city

Peri-urban

12.8

16.3

32.8

19.8

3.0

6.0

7.7

7.6

20

Table 2. Summary of Selected Characteristics of 28,783 Census Tracts in 100 Mexican Cities Variable

Mean

Median

0.19 0.20 7.72 11.75

0.16 NA 5.75 7.91

0.13 0.40 7.01 11.18

Household Size Infrastructure Index a

3.98 0.14

3.91 0.01

3.27 0.30

Percent of population with IMSS b Percent over 15 with high school Percent population born in another state

0.38 0.23 0.25

0.39 0.23 0.20

0.18 0.09 0.18

Percent of housing vacant Percent of tracts with >25% vacancy Population Density (000s per hectare) Distance to CBD (km)

Std. Dev.

Percent population aged 15-64 0.65 0.65 0.05 Notes: a The Infrastructure Index is a sum of the percent of houses without access to piped water, drainage, or electricity. b The share of people with health insurance from the Instituto Mexicano de Seguro Social (IMSS), which is the largest insurance plan for private sector employees and denotes formal employment and membership in INFONAVIT. Source: INEGI, 2010.

21

Table 3. Regression Results: Census Tract Level Vacancy Rates on Tract Characteristics

Variables Population density (log) Distance to CBD (log) Household size (log) Infrastructure Index a Percent of population with IMSS b Percent over 15 with high school Percent pop. born in another state Percent population aged 15-64

OLS (DV=% Vacant) Coeff. Std. Coeff. 0.019 0.149 [0.001] -0.263 -0.307 [0.007] 0.018 0.043 [0.003] 0.170 0.229 [0.008] 0.180 0.118 [0.014] 0.194 0.267 [0.007] -0.284 -0.111 [0.025] 0.076 0.134 [0.004] 28,783 0.38 85.91

Logistic (DV=Vacancy>25%) 0.559 [0.010] 1.598 [0.041] 0.006 [0.001] 2.108 [0.132] 13.760 [2.045] 40.996 [9.971] 95.296 [15.498] 0.002 [0.001] 28,783

Observations Adjusted R2 F-statistic LR chi2(107) 7726.60 Log likelihood -10689.19 Pseudo R2 0.27 Notes: Standard errors are in brackets. All coefficients are statistically significant at the 0.01 level. City-level fixed effects are included. a The Infrastructure Index is a sum of the percent of houses without access to piped water, drainage, or electricity. b The share of people with health insurance from the Instituto Mexicano de Seguro Social (IMSS), which is the largest insurance plan for private sector employees and denotes formal employment and membership in INFONAVIT.

22

Table 3. Summary of Selected Characteristics of 100 Largest Mexican Cities Variable

Mean

Median

Houses (thousands), 2010 Population growth (%), 2000-2010 Population density a, 2010 Economic growth b, 2000-2010 Mean household income c, 2010 Murder rate d, 2006-2009 HHs with int’l migrants (%), 2010 Loans e 2000-2010 / houses 2000 Share self-built, 2010

180.95 0.46 38.64 0.95 9.66 1.45 0.03 0.35 0.24

67.63 0.40 38.66 0.97 9.41 1.29 0.03 0.31 0.23

543.90 0.29 14.00 0.47 2.06 1.67 0.02 0.23 0.07

2.45

2.43

0.15

Mean household size, 2010

Std. Deviation

Sources: Census of Population and Housing 2000 and 2010, Economic Census 1999 and 2009, Crime Indicator Database for Justice in Mexico, CONAVI, and author’s calculations with Urban Cartography from INEGI. Notes: a People per hectare. b The percent change in income. c Thousands of pesos d The murder rate is measured at the state level, the number of murders per 100,000 people. e Only public sector loans are considered.

23

Table 4. Regression Results: City Level Vacancy Rates on City Characteristics

Variable Houses (log), 2010 Pop. growth (%), ‘00-‘10 Pop. density, 2010 Econ. growth a (%), ‘99-‘09 Mean HH income (log), ‘10 Murder rate b, 2006-2009 Int’l migrants c, 2010 Loans/houses d 2000-2010 Share self-built, 2010 Average HH size, 2010 Constant Number of observations Adjusted R-squared F-statistic

All vacancy Std. Coef. Coef. -0.002 [0.004] 0.022* [0.013] 0.009 [0.007] -0.001 [0.007] -0.011 [0.020] 0.006** [0.003] 0.400*** [0.137] 0.071*** [0.020]

-0.053

-0.086*

-0.179

[0.046] -0.085 [0.058] 0.269 [0.178] 100

0.192 0.105 -0.015 -0.068 0.273 0.263 0.489

-0.154

Central-city vacancy Std. Coef. Coef. -0.001 [0.004] -0.008 [0.013] 0.001 [0.008] -0.002 [0.007] -0.022 [0.019] 0.000 [0.003] 0.385*** [0.139] 0.062*** [0.021]

-0.021

0.031

0.071

[0.061] -0.096* [0.055] 0.383 [0.179] 100

-0.082 0.018 -0.035 -0.158 0.010 0.281 0.473

-0.194

Peri-urban vacancy

Coef. 0.007 [0.007] 0.022 [0.030] 0.018 [0.014] -0.003 [0.016] 0.025 [0.033] 0.007 [0.005] 1.037*** [0.342] 0.045 [0.038] -0.104 [0.116] -0.260** [0.129] -0.015 [0.334]

Std. Coef. 0.120 0.104 0.120 -0.021 0.086 0.203 0.375 0.169 -0.119 -0.259

100

0.38

0.18

0.28

11.53

3.74

5.68

Notes: White robust standard errors reported in brackets. *, **, *** indicate statistical significance at the 0.01, 0.05 and 0.01 levels. a The percent change in value added. b Murder rate is measured at the state level, the number of murders per 100,000 people. c The percent of households where a member is living abroad. d The number of loans issued by public sector agencies from 2000-2010 is divided by the number of housing units in the year 2000.

24

Figures

Figure 1. Census tracts classified by distance to city center, Aguascalientes, Aguascalientes

25

Figure 2. Share of housing vacant by census tract, Ciudad Juarez, Chihuahua Source: Author with INEGI, 2010

26

Figure 3. Share of housing vacant by census tract, Pachuca, Hidalgo Source: Author with INEGI, 2010

27

Figure 4. Share of housing vacant by census tract, Leon, Guanajuato Source: Author with INEGI, 2010

28