Housing Finance in Latin America and the Caribbean: What is holding it back? 8096)":,;)"+"*$,)",2%+<)/,

Transcription

1 "#$%&'($%)*+",-$.$/01($"#,2+"3,4,5$6$+%*7,-$1+%#($"#, Latin American Research Network Latin American and Caribbean Research Network Project, Housing Finance in Latin America and the Caribbean: What is holding it back? Research Proposal 8096)":,;)"+"*$,)",2%+<)/, "# $%&'()*+,-,./'012/&34%&'0-**()%4'5*(6 "##$%&'()%&*%+$',-.""/'"0'1*234$225'6$&$)%/'7438$)23+9'"0':3"'&$';%4$3)"' 7'(8'(%)'9,:;5%/((/< "##$%&'()%&*%+$',-.""/'"0'1*234$225'6$&$)%/'7438$)23+9'"0':3"'&$';%4$3)"',

2 1. Introduction The Brazilian economy has undergone profound macroeconomic transformations over the past 15 years, which began with the successful introduction of the Real Plan that assured the end of an era of high inflation, after a long period in which the monthly inflation rates reached exceed 80%. At the same time, the introduction of the Fiscal Responsibility Law and of the floating exchange rate regime guaranteed, respectively, a downward trend for the ratio public debt over the GDP and a substantial improvement in external accounts, rendering the country a net international creditor. The combination of low inflation rates, falling public debt and reduced external vulnerability led to a substantial reduction in interest rates. The development stage of Brazilian states is diverse, from the heavily industrialized states in the South and South East to the impoverished states in the Northeast and in the North. Income distribution is more homogeneous in the South and disparities are greater in the wealthier Southeastern states of São Paulo and Rio de Janeiro. The concentration of federal employees in the Federal District gives it the highest income per head in the federation. The objective of this study is to describe and analyze the current conditions of housing finance in Brazil, highlighting its limitations and discussing the main recent policies and the possibilities of change. The study will present an analysis of the determinants of the levels of housing credit to GDP at the national and state levels. The analysis by state will consider the large regional differences that characterize the Brazilian economy and allow some exogenous variables that are heterogeneous at the state level, such as income, development stage, and the quality of the judiciary, to be considered. The paper will conclude with an assessment of current policies and recommendations for improvement. The housing financial system in Brazil is currently comprised of two basic structures: the Housing Finance System (SFH - Sistema Financeiro da Habitação) and the Real Estate Financial System (SFI - Sistema Financeiro Imobiliário). Law 4380 established the SFH in 1964 with mandatory funding from the savings accounts of families and the resources in the FGTS, a fund formed from mandatory contributions by employers levied on payroll. The FGTS works as a safety net for workers in some situations, such as when one is jobless, and its resources are also used to fund urban development. Sixty-five percent of savings accounts balances have =

3 to be direct to SFH housing financing and the rest may fund the mortgage portfolio at market rates. Law 9514 created the SFI in 1997 to complement the SFH using market interest rates. The SFI brought two major innovations: (1) the securitization of mortgages, facilitating and speeding up the negotiation of real estate loans by means of market mechanisms, with the expectations about the growth of a broad and dynamic secondary market for bonds tied to mortgages, and (2) less legal uncertainty in contracts, by introducing the concept of a lien to improve the collateral value of the property, providing greater security to the creditor because the ownership of the property remained with creditor until the full payment of the house. The prior system protected debtors and the collateral value of the housing property was low. A number of additional institutional improvements in the system have been implemented thereafter. Nevertheless, the main source of funds for home loans is still the savings accounts, in contrast to mortgage-backed securities in many wealthier countries. Despite the advances described above, Brazil is still a country with a high housing deficit. In 2005, the estimates pointed to a deficit of 7.8 million homes, representing 14.7% of all households in the country, this deficit is concentrated, in absolute terms, in the states of Rio de Janeiro and São Paulo (the most densely populated and more developed) and in the lower income population. Currently, the estimates show a deficit of around 5.6 million single-family homes. Several factors account for the present conditions, among them, the low levels of per capita income, the high interest rates charged and unsatisfactory regulation due to the persistent judicial uncertainty. For these reasons, the volume of total credit in the Brazilian economy is still low at 45% of GDP, being only 2.9% of GDP for housing loans. However, the outlook for the housing sector is promising. The construction industry employs 1.9 million people formally (5% of total formal employment in Brazil) and, in 2009, the industry has already created 210,000 jobs (18% of all new formal jobs in Brazil). It is estimated that the growth potential of real estate as a whole can triple in size in the next two decades. The continued of macroeconomic stability will ease the growth of credit at longer terms (essential for real estate loans), lowering risk premiums, and ensuring greater availability of resources for this sector. >

4 Data A detailed data template is offered in the appendix to this proposal. Several kinds of time series are available at the state, regional, and national levels. The more detailed time series are shorter. The data will allow the analysis of panels for each state in time. We specify the data that we have initially identified as necessary for our analysis below. In the course of the project, it is possible that we employ additional time series data depending on what our literature review suggests and on what is available. Bank credit data is available on a monthly basis with a breakdown by state from National level data about bank credit is available since 1994, including housing credit and other types of bank credit may be easily obtained from the Brazilian Central Bank, whose database is available freely on the Internet and provides the means for downloading time series data into text or spreadsheet formats. We will analyze data only from 1994, or later, because of major economic transformation after the Real Plan in 1994, including a different currency and the stabilization of inflation. Economic data (such as GDP, GDP per head, inflation rates, interest rates, etc) are easily obtained from several of the databases we have free access to, such as those provided by the Brazilian Central Bank, the IBGE (National Institute of Geography and Statistics), Economatica, the IFS database, and others. GDP data is also available per state on an annual basis. This type of data will allow the computation of several credit variables relative to state and national GDP. GDP data is available on an annual basis at the state level and on a quarterly basis for the national level. Thus, credit relative to GDP data will be available annually for state level panels and quarterly for national level time series analysis. Demographic data, such as population, population density, and other urban development indicators may be easily obtained from the IBGE (National Institute of Geography and Statistics) database. The need for their use as exogenous or control variables may emerge from our literature review. The inefficiency of the judiciary is important for the development of credit markets. Domestic measures that gauged inefficiency were employed by Pinheiro and Cabral (1998) and Pinheiro (2003) and are available at the state level for 1996 and We believe that these measures may not be obtained for more recent years, but that is yet to be determined. However, it is possible to check the stability of such inefficiency index by means of the "Legal Rights" and "Credit Information" indices?

5 from the World Bank's Doing Business reports or Worldwide Governance Indicators. If these variables are stable for Brazil, then using the inefficiency index computed for 1997 could be acceptable. For example, the "Property Rights" index computed for Brazil by the The Heritage Foundation for its "Index of Economic Freedom" has been stable at 50 (out of 100 points) since 1995, indicating the there was no general improvement in the protection offered by the judiciary. The inefficiency of the judiciary index is made of three components: cost, delays, and fairness. It has been computed from surveys carried out in 1996 and 1997 with 602 and 279 business professionals and owners from the privately and state owned firms from 22 out of the 27 Brazilian federative units (states and the federal district). We will consider using such measure of the quality of the judiciary in each state in our empirical exercises. Alternatively, Pinheiro (2003) identified states that have a potentially better judiciary. Dummies may be used for a few states in place of the inefficiency index. We provide a topic summary of the date that may be employed below. For more detail, please exam the templates provided in the appendix of this proposal. Time series data on aggregate measures of domestic bank credit for housing and commercial property financing as well as detailed time series data disaggregated by state and type of housing credit: acquisition, building, and remodeling. There are monthly data about volume, number of financial institutions, type of borrower (developers, individuals, and cooperatives), interest rate regime (National Financial System interest, a kind of non-market interest, and market interest), and the type of property financed (housing or commercial). Time series data on aggregate volume of domestic bank credit for all other types of credit, including rural credit, as well as time series total volume data disaggregated by state. Time series credit volume as a percentage of GDP data are provided by the Brazilian Central Bank and may be calculated for its components from GDP data provided by the IBGE. GDP per capita is easily obtained on an annual basis from IBGE, with a state level breakdown. Time series data on a quarterly basis about the national GDP and its growth is available from IBGE and the percentage of the GDP of the state generated by the

6 sector as well as the GDP of the state in monetary units may be obtained from IBGE on a annual basis. Time series data about the level of interest rates of several types of bank credit as well as the interbank rate used as a reference by the market in Brazil may be easily obtained from the Brazilian Central Bank and from the Anbima (the Brazilian Association of Financial and Capital Markets Institutions) databases. The volatility of the interbank interest rate may be easily obtained for monthly or quarterly periods from the daily rate available from the same sources. A time series rate of SFH rate, which is a non-market rate established by the government for most housing financing is also available. Time series inflation data are available for several indices. Time series data of the amounts issued and outstanding of property related collateralized bonds is available from 1995 and the sources are the Brazilian Securities Exchange Commission (CVM), the Stock, Futures, and Merchandise Exchange (BM&FBOVESPA), and the National Debentures System (SND). The collateral of these bonds may not involve housing related financing at all. Methodology The strategy to collect information to produce the third section of the project on public policy initiatives adopted for the development of the national housing finance system will be a search on them main economic news vehicles (such as newspapers, public banks, government websites, ABECIP, the Brazilian Association of Savings and Real Estate Loans Institutions, and others). The Brazilian government offers a very well organized website for all laws and regulations passed in the country. Business newspapers subscribed by the authors institutions offer a very good online news retrieval system that allows the identification of major actions regarding housing and its financing. One of the authors is associated to a group that publishes a monthly economic newsletter and that follows all the major initiatives by the government, such as those related to housing and housing financing. ABECIP offers several resources and statistics about the housing market. Other important economic agents also provide such information and research, such as the National Economic and Social Development Bank (BNDES) and Caixa Econômica Federal (CEF), the main loan agent in the Brazilian mortgage debt market. A

7 For the fourth section of the project we intend to use econometric modeling. The main references, so far, for the development of such models are Pinheiro and Cabral (1998), Warnock and Warnock (2008), and Geyer (2009). The last two studies perform a multi-country analysis of the determinants of the of housing credit in a large sample of developed and emerging countries for the period and for the period, respectively. From the models employed in those studies, we propose two models below. A national level time series model and a state level panel model are possible with the data identified as available so far. In general, a model could be constructed relating the variable representing housing credit to some of the main potential explanatory variables that have been used in the literature (for which we shall provide an appropriate economic justification) or identified in the previous sections of this proposal. Equation 1 portrays a panel model for the state level analysis. HF /GDP s,t = J s + 2 A s,t + 3,t I s,t + 4 G s,t + 5 X + " s,t Eq. 1 HF/GDP s,t is house financing by state gross domestic product (GDP) for state s and year t. More detailed measures of housing finance at the state level are the volume of credit disaggregated according to purpose (acquisition or building) and type of interest rate (SHF interest system or market interest rate). The detailed information may provide additional insights about the determinants of housing finance in Brazil. The other explanatory variables in Equation 1 are available on annual basis at the state level, except for J s that represents the quality of the judiciary in the state. We do not expect to obtain this variable for every year, thus it will be a state level variable that may be represented by an index of the judiciary inefficiency employed by Pinheiro and Cabral (1998) or by dummies that single out a few states perceived to exhibit a less inefficient judiciary, such as in Pinheiro (2003). A s,t is the percentage of the state GDP corresponding to the rural sector and denotes its the stage of development. I s,t is the income per head in the state and G s,t its economic (GDP) growth rate. X represents a vector of state level control variables that may be identified as the project develops. Many housing credit time series are available on a monthly basis but we may not have a full set of explanatory variables with a monthly or quarterly frequency to implement a model like Equation 1 only with an yearly frequency. B

8 Equation 2 represents a possible model for the national level analysis. The macroeconomic environment may obviously influence housing financing. We plan to include the following variables that are available on a quarterly basis: economic (GDP) growth (G t ), the level of the domestic inter-bank interest rates (i t ), the volatility of the inter-bank interest rate (v t ), the inflation rate (P t ), and the level of non-housing bank credit relative to GDP (C t ). In alternate specifications, one could consider the GDP per head and its volatility and growth rate as well as the volatility of inflation, instead of the volatility of interest rates. The proper economic justification of such variables will be developed in the project. HF /GDP t = G t + 2 C t + 3,t i t + 4,t v t + 5 P t + " t Eq. 2 More detailed measures of housing finance at the national level are the volume of credit disaggregated according to purpose (new housing acquisition, used housing acquisition, building, materials, and remodeling), to borrowers (developers, individuals, cooperatives), the type of interest rate (SHF interest system or market interest rate), and the interactions between these variables. Data on the volume of issuance of collateralized debt obligations that employ collateral related to the real estate market is available. However, the collateral is usually related to commercial and industrial property, such as lease back contracts, and may not apply to the project. These obligations are called Certificados de Recebíveis Imobiliários (CRI) and we will investigate how far they apply to the housing market by examining the prospectuses of the main issues. Finally, regarding the statistical methods, the appropriate techniques for panel regressions regarding the choice of fixed effects and random effects will be verified with the Hausman test. The p-values of the coefficients in the panel regressions will be calculated using clustered standard errors as in Arellano (1987) and Rogers (1993). In the time series analysis, adjustments, such as AR"#" will be verified too. Any OLS regressions will be performed with robust standard errors. " Literature" Our initial survey of the housing financing literature in Brazil revealed a limited number of articles about housing and housing finance, most of then descriptive. We concentrate here on the most recent ones. We have mentioned that the studies by C

9 Pinheiro and Cabral (1998) and Pinheiro (2003) about the judiciary, its inefficiencies, costs, delays and its impact on the credit markets in each Brazilian state will be important for the development of the project. Pinheiro (2003) surveyed 741 judges in 11 Brazilian states and the Federal District and suggests dummies for those states that supposedly have a more developed judiciary. We should also mention that the main references about the housing deficit in Brazil are the studies by Fundação João Pinheiro (2006) funded by the UN and the IADB. Regarding housing finance, Costa (2004) offers a brief description of the Brazilian housing financing systems and performs an analysis of the time series properties of total and housing bank credits. No analysis of the determinants of the levels of bank credit is provided but a co-integration analysis of total credits and housing credits is performed. The housing credit time series is exogenous relative to the total credit levels series. The author concludes that the housing credit market behaves independently of the total credit market because of its specific policies and exclusive mandatory ways of funding. Moreover, the author believes that the policies induce banks to concede housing credit up to the mandatory legal boundaries but that they do not have incentives to expand from that. Finally, some suggestions for the improvement of the system are offered, such as the easing to recover the collateral and more flexibility in the terms of credit. Costa and Lundberg (2004) analyze the mandatory rural and housing financing by banks and reach similar conclusions. There are a number of relatively recent descriptive works about the housing financing system. Vedrossi (2007) describes the Mexican housing financing system and contrasts it with the Brazilian. FGV Projetos (2007) also provides a report briefly describing the Brazilian housing finance system, contrasts it with those in the US, Mexico, Chile, South Korea, and the European Union and offers a proposal for a revised housing finance system. Ferreira (2003) provides a historical account of housing financing subsidies through case studies and concludes that these subsidies were perverse and led to wealth transfers from the poor and from those that did not own homes to the wealthier and to homeowners, respectively. Other descriptive studies include Rossbach (2005) and Santos (1999). Carneiro and Goldfajn (2000) offer a model for mortgage securitization in Brazil and for the design of these securities, such as the inflation indexation to be used, and discuss their risks and potential policies to develop the market. Leal and Carvalhal-da- Silva (2009) analyze the Brazilian bond market, including the securitizations related D

10 to commercial property and the determinants of the issuance of securitized obligations by Brazilian listed companies. Wilson (2009) speculates that the development of the securitized mortgage obligations market in Brazil will be delayed by the international crisis of 2008 and that lower domestic interest rates will help, but his results were not conclusive about the impact of interest rates on the development of such market. References: Arellano, M. Computing robust standard errors for within-groups estimators. Oxford Bulletin of Economics and Statistic 49, p , Carneiro, Dionísio; Goldfajn, Ilan. A securitização de hipotecas no Brasil. Texto para Discussão 426, PUC-Rio: Rio de Janeiro, Available at Access on 12/2/2009. Costa, Ana Carla. Mercado de Crédito: uma Análise Econométrica dos Volumes de Crédito Total e Habitacional no Brasil. Trabalhos para Discussão n. 87, Banco Central do Brasil: Brasília, Available at Access on 12/2/2009. Costa, Ana Carla; Lundberg, Eduardo. Direcionamentos de Crédito no Brasil: uma Avaliação das Aplicações Obrigatórias em Crédito Rural e Habitacional. In: Banco Central do Brasil, Economia Bancária e Crédito - Avaliação de 5 Anos do Projeto Juros e Spread Bancário, p , Banco Central do Brasil: Brasília, Available at Access on 12/3/2009. Ferreira, Thaís. A concessão de subsídios por meio do sistema financeiro de habitação. Master's Thesis, Departamento de Economia, PUC-Rio, Available at dice.html. Access on 12/3/2009. FGV Projetos. O Crédito Imobiliário no Brasil - Caracterização e Desafios. FGV Projetos - Fundação Getulio Vargas: São Paulo, Available at Access on 12/2/2009. EF

11 Fundação João Pinheiro. Déficit Habitacional no Brasil Fundação João Pinheiro: Belo Horizonte, Available at CodDocumento=55667#. Access on 12/3/2009. Geyer, Roberta C. Análise Empírica do Volume de Crédito Imobiliário. Master's Thesis, EESP (Escola de Economia de São Paulo) da Fundação Getulio Vargas, Available at 20Geyer.pdf?sequence=1. Access on 12/3/2009. Leal, R. P. C.; Carvalhal da Silva, A. L. The Development of the Brazilian Bond Market. In: Borensztein, E., Cowan, K., Eichengreen, B., Panizza, U. Bond Markets in Latin America - On the Verge of a Big Bang?, Cambridge, MA: MIT Press, p , Pinheiro, Armando. Judiciário, Reforma e Economia: a Visão dos Magistrados. Texto para Discussão 966, IPEA: Rio de Janeiro, Available at SSRN: Access on 12/3/2009. Pinheiro, Armando; Cabral, Célia. Mercado de Crédito no Brasil: O papel do judiciário e de outras instituições. Ensaios BNDES, 9, BNDES: Rio de Janeiro, Disponível vos/conhecimento/ensaio/ensaio9.pdf. Access on 12/3/2009. Rogers, W. Regression standard errors in clustered samples. Stata Technical Bulletin 13, p , Rossbach, Ana Cláudia. O Financiamento Habitacional no Brasil. Master's Thesis, Departamento de Economia, PUC-SP, Available at 331/Publico/Financiamento%20Habitacional%20no%20Brasil.pdf. Access on 12/3/2009. Santos, Cláudio. Políticas Federais de Habitação no Brasil. IPEA - Instituto de Pesquisa Econômica Aplicada Discussion Paper 654. Available at Access on 12/2/2009. EE

12 Vedrossi, Alessandro. Mexican Housing Financing and its Comparison with the Brazilian Market. VII Seminar of the Latin American Real Estate Society, São Paulo, Brazil, Oct , Available at Access on 12/2/2009. Warnock, Veronica; Warnock, Francis. Markets and housing finance. Journal of Housing Economics, v. 17, p , Wilson, Peter. Influência das Taxas de Juros e do Canal de Crédito na Formação de um Mercado Secundário de Hipotecas no Brasil. Master's Thesis, EESP (Escola de Economia de São Paulo) da Fundação Getulio Vargas, Available at Cortes%20Marsden%20Wilson.pdf?sequence=1. Access on 12/3/2009. E=

13 Appendix: Excel Data Templates National Level Data Available Monthly Since 1994

14 National Level Data Available Monthly Since 2002 "#

15 "#$%&'"()*'+,$-$+.)*"#+/01)*)2(13'"' 4353'"'6-'(+'7+$89'%"$%+/ "$

16 :$;()*'+,$-$+3'"'6-'(+'7+$.)*"#+/<(*1$=>>= "%

17 "&

18 <"'"$,$-$+.)*"#+/01)*)2(13'"' "'

19 <"'"$,$-$+C**$D(1($*1/)D"#$E9F(1('%/C*F$GD)%HIIJ'*FHIIK "(

20 )*