Social Housing Finance in Colombia
Research Proposal
Fedesarrollo
- Motivation
Housing deficit in Colombia remains high, despite the recovery of the sector after the crisis of the late nineties. According to the last census, the quantitative housing shortage is 12.4% and there are 23.8% of households dwell in inadequate units, which adds up to an effective housing deficit of 36.2%[1]. The literature suggests that formal housing in Colombia is constrained, among other factors, by land markets failures, lengthy and costly procedures to obtain building permits and low access to credit.
As a matter of fact, housing finance in Colombia is small as compared with Latin American standards. Mortgage loans ranged from 3.0% to 3.5% of GDP during the last five years, while the regional average is around 5%[2]. After the financial crisis of the late nineties, which lasted from 1998 to 2001, the Colombian mortgage market recovery has been weak. While the ratio of total loans increased from 23.8% of GDP in 2001 to 46.2% in 2008, mortgage credit dropped from 6.2%of GDP in 2001 to 3.5% in 2008[3].
Mortgage lending in Colombia only funds a third of total housing, while the rest is funded by informal lenders or self-funded. Formal housing loans are concentrated on the formally employed segment, leaving out 70% of social housing demand[4] which is comprised of households that earn their living from informal activities[5]. In fact, the problem of labor informality, which reaches almost 65% in Colombia, is behind the housing deficit and the lack of credit, since standard financial instruments don’t suit the particular needs of this population.
The Colombian literature has identified different factors that constraint the access to credit, especially by low-income people, among which two are particularly relevant. First, even though poor families usually manage to accumulate a significant amount of capital over the years by self-building, they may not have access to credit because loan providers perceive a high risk of default from borrowers with low and volatile income(Galindo, 2005).In the same direction, Galindo and Hofstetter (2006) estimated the determinants of mortgage-interest rates in Colombia and found that at the microeconomic level the main cause of high level of the interest rates is the credit risk assumed by lenders.Second, the supply of social housing credit is also constrained by the lack of collateral due to deficiencies in deed registration and the high costs and length of recovering the collateral (Cardenas y Badel, 2003).
To promote the provision of social housing in Colombia, the government has implemented a number of instruments including, mainly tax exemptions, provision of guarantees, direct subsidies and rediscount credits. However, the scope and effectiveness of these instruments have been limited. As mentioned above housing access to the poorest segments of the population is still significantly low. The program of subsidies for social housing is the most important, providing direct, one-time subsidies to home buyers who demonstrate having savings for 10% of the total cost of the house.It is also expected thatthe subsidy facilitatesaccess to credit. The provision of guarantees to social housing credit by the National Guarantees Fund (NFG),which aims to address the obstacles of lack of collateral and its recovery, is another instrument of significant incidence on Colombian housing finance - approximately 65% of the social housing loans are backed by these guarantees.
This study will focus on the access to housing credit by the low-income population. The main objective is to determine the impact of public policies addressed to stimulate social housing (input) on the access to housing credit (output) by the poor. Emphasis will be placed on the effect of housing subsidies on the access of low-income households to mortgage,as well as on the role played by the NGF in solving the lack of guarantees for social housing credits. Additionally, we will evaluate the risk involved in social housing credits, through the estimation of the determinants of housing credit default probability.
- Literature Review
The housing sector in Colombia has been widely studied, with papers focusing on the housing market, the mortgage market and the social housing.
Recently, Clavijo et al (2005) studied the Colombian housing market development and find evidence suggesting that household’s disposable income, new housing prices and real interest rates on mortgage credit are important determinants of housing demand. Arbelaez (2004) also estimated the demand for housing and finds that credit, real interest rates, employment and labor income play an important role in determining housing demand.
There are also relevant recent studies regarding the housing financial market in Colombia. For instance, Cardenas and Badel (2003) suggest that the financial crisis of the late nineties in Colombia was a consequence of the drop in housing prices and a significant increase in the value indebted by households, which deteriorated the loan to value ratio in the market. Murcia (2007) used the Quality of Life Survey of 2003 to analyze the socio-economic determinants of access to mortgage loans and credit cards and they find that the probability of having a mortgage loan increases by 11.7% for households in the higher quintile of the income distribution. This probability also increases if the household has a housing subsidy or if it is located in urban areas. The main limitation of this study is that the estimations only include households applying for one of the credit instruments, which induces auto-selection biases.
Among the studies focused on social housing finance, Cuellar (2006) presents a thorough analysis of the regulatory framework evolution and its incidence on housing finance development. Rocha et al. (2006) study the main barriers of access to credit by low-income people, estimating the determinants of supply and demand of social housing credit. They find that the probability of getting a mortgage loan increases when households have been granted a subsidy, have high and stable income, and hold programmed saving accounts.Programmed saving accounts are even more important in the econometric estimations than working in the formal sector. However, the authors suggest that credit supply to informal workers is constrained by the lack of appropriate mechanisms of information sharing,and recommend promoting programmed savings among this population as a factor that signals income stability to credit suppliers. The results of this study are estimated from cross-section data which makes it difficult to analyze the dynamics of housing credit and its determinants.
The National Planning Department of Colombia (2007) evaluates the Urban Social Housing Subsidies Program. The evidence shows that assets ownership, education level, and access to information are determinants of the program participation. According to this study, the program has positive and significant impact on the house and neighborhood physical conditions, as well as on the beneficiary households’ expenditure and savings. However, it does not estimate the effect of the subsidy on the access to housing finance.
Some studies evaluate policy instruments that seek to facilitate the supply of social housing credit. Marulanda et al (2006) measure the fiscal cost of the guarantees provided by the National Fund of Guarantees to back social mortgage securities and housing credits, as well as the fiscal cost of tax exemptions designed to promote social housing supply. The authors provide an analysis of the performance of these instruments over time and with respect to set targets. Silva (2007) provides a financial and operative analysis of the social housing rediscount credit provided by FINDETER and formulates recommendations to strengthen the instrument and ensure its financial sustainability.
In contrast with previous evaluations of the subsidy, which rely on time series or cross-section, this study offers an impact evaluation using micro-data information for a panel of individuals from 2008 to 2009. The advantage of this methodology is that it allows estimating the impact of subsidies on changes in credit access by the estimation of a difference-in-difference specification. In addition, one of the most important instruments for promoting social housing credit, the National Fund of Guarantees, has been studied form a fiscal perspective but not from the impact on social housing credit perspective. In this study we will evaluate the role of the guarantees in easing the credit and the characteristics of the beneficiaries using information at individual-level.
- OUTCOMES
The outcome analysis covers the characteristics and evolution of the Colombian housing market and those of the housing finance system. We will include a general description of both markets, focusing on social housing.
- The Housing Market
The characterization of the housing market will include, among others:
-Cycle and trend of the housing sector GDP, built area, building costs and housing prices
-Regional distribution of housing GDP
-Housing prices by income segment
-Cycle and trends of land prices
-Rental market functioning
-Impact of the construction activity on employment
-Analysis of the role of institutions and regulations of the housing market
Focusing on low-income housing we will use Fedesarrollo´s Longitudinal Social Survey (see annex for a complete detail of the Survey) to provide an overview of the housing system in Colombia, emphasizing the mechanisms used by the poorest households to satisfy their housing needs and the conditions of the housing units where they live.
We will characterize the occupancy conditions (rental, occupancy with no title deed, formal ownership), the expenses of the family on housing as a percentage of their monthly income, the physical conditions of the house and its access to public utilities, the number of families living in the same housing unit, household relocations over the last four years and the reasons for moving to a different unit, as well as changes in the general conditions of the dwelling unit.
This section will include a review of the existing diagnosis of the failures on land market regulation and their incidence on social housing supply and the main public policy instruments adopted to overcome this constraint. For example, we will study the evolution of the regulation on housing macro-projects and land use, with emphasis on its modifications over the last year (2009).
- The Housing Finance Market
The characterization of housing finance market will include, among others:
-Cycle and trend of the housing loans relative to total loans, and as a percentage of GDP
-Interest rate volatility and level
-Housing loans by income-segment
-Participation in the financial market of rediscount funds, commercial banks, public banks, cooperatives (other organizations such as cajas de compensación familiar), microfinance entities, and other intermediaries
-Credit supply by income segment and geographical location
-Evolution of mortgage backed securities markets
-Analysis of the role of institutions and regulations of the housing finance market
Based on information from Fedesarrollo’s Longitudinal Social Survey, we will evaluate the credit housing access, using the following indicators:
-Distribution of beneficiaries per income quintile
-Percentage of benefited households earning most of their income from informal vs. formal activities
-Socio-economic characteristics of the household’s head receiving the subsidy (education attainment, income level, gender, geographical location, number of members in the household)
-Quality of housing of beneficiaries receiving credit and its evolution over the last two years (building materials, access to public utilities).
- INCOMES
This section will describe the main public policy instruments adopted to promote housing finance development and their evolution over the last ten years. Especial attention will be placed on the description of regulations that impact social housing finance including:
-Social Housing Subsidies
-Programmed Saving Accounts
-Rediscount Banks, with emphasis on FINDETER and the different measures adopted to strengthen the financial resources of the bank
-National Guarantees Fund
- Guarantees to social housing credits
- Guarantees to securities on social housing
-Voluntary agreements between the Government and the commercial banking sector to provide a percentage of (0,5%) gross loans to social housing financing
-National Saving Fund (Fondo Nacional del Ahorro)
-Tax exemptions on revenues from social housing loans, mortgage securities and housing leasing
-Inflation coverage from UVR fluctuations above the inflation target
-Interest rate subsidy implemented on April 2009
A compilation of qualitative information and time series will be analyzed in order to determine the scope of the instruments over the last five years.
Based on information from Fedesarrollo’s Longitudinal Social Survey, we will evaluate the social housing subsidies targeting, using the following indicators:
- Distribution of beneficiaries per income quintile
- Percentage of benefited households earning most of their income from informal vs. formal activities
- Socio-economic characteristics of the household’s head receiving the subsidy (education attainment, income level, gender, geographical location, number of members in the household)
- Quality of housing of beneficiaries receiving the subsidy and its evolution over the last two years (building materials, access to public utilities)
- LINKS between incomes and outcomes
This section focuses on establishing the incidence of the social housing program of subsidies and the provision of financial guarantees on the development of social housing finance.
- Evaluation of the impact of subsidies
- Impact of subsidies on the access to credit
The purpose of this section is to identify the impact of social housing subsidies on access to housing credit. We are particularly interested in identifying the population segments obtaining a higher benefit from these subsidies, by including information on their income level and their working conditions (formal/informal). Another feature we will focus on is the incidence of bancarization on access to credit. We will control for socio-economic conditions such as household head educational attainment, gender and household geographical location.
We will estimate a difference-in-differences model using the panel data available for the years 2008 and 2009 of Fedesarrollo’s Longitudinal Social Survey, which provides information of households with subsidies and without subsidies, as well as the evolution of the access to credit for both groups over time. The specification will be:
Where:
Yit denotes the outcome for household i in period t, i.e. it is a dichotomous variable with value Yit = 1 if a household has a housing credit in period t, and Yit = 0 otherwise.
Subsidyit is a dummy variable taking the value 1 if the household has a housing subsidy in period t and 0 otherwise; Tt is a dummy variable taking the value 1 in the post-treatment period (after the first year of been benefitted by the subsidy) and 0 in the pre-treatment period (in the first period of been benefitted by the subsidy).
Xit is a vector of exogenous household´s characteristics including household head’slabor conditions (formal/informal) gender, age, and educational attainment. Xit also includesthe number of members of the household, the income level, and indicators for household bancarization and payment habits.
The indicator of bancarization will include the household’s access to financial products and services including: current and saving accounts, investments (certificates of term deposits), credit cards, microcredit and insurances.
Following Murcia (2007) we will also construct an indicator of payment habits. This indicator will include information regarding the following variables:
- Lags in housing debt payments during the last 4 months
- Lags in utilities payments and others associated with housing maintenance during the last 4 months
- Lags in tax payments
This indicator is calculated according to the following expression:
Where:
mi is a dummy variable taking the value 1 if the household reports a lag in the payment of housing debts in the last 12 months and 0 otherwise; m is the mean of this variable reported by the households who did not pay this amount, and sm is the standard deviation. The coefficient f1 is the first estimated principal component. The variables ut and tax correspond to the utilities and tax payments, respectively. The weighted sum of these components corresponds to the index of payment habits which will allow us to assign a credit score to each household.
Wit corresponds to region-level control variables such as Gross Domestic Product (GDP), Consumer Price Index (CPI), Securitized housing loans-to-mortgage debt ratio, and Social-Interest Housing Loan Guarantees-to-Mortgage Debt ratio. Securitization is one of the methods of financing the purchase of new or used housing and it is an alternative to mitigate maturity mismatch and interest rate risks. Therefore, controlling for this variable will allow us to consider the effect of this instrument to promote the access to housing credit in Colombia.The Social-Interest Housing Loan Guarantees program is administered by the Fondo Nacional de Garantías (National fund of guarantees, FNG) who backs loans whose destination is to finance the acquisition of Social-Interest Housing. The availability of this program facilitates access to credit to low-income households who do not have collateral to apply for a loan.
The error term itis composed of individual, family and community unobserved characteristics and a stochastic disturbance term, and ciis an unobserved fixed effect.
The interaction Tt * Subsidyit is a dummy variable which takes the value of 1 only for the treatment group in the post-treatment period.
In the case of binary outcomes (such as whether or not a household has access to a mortgage-backed loan), we use a probit model specification. In non-linear models, the Difference-in-Differences estimator of the impact of subsidy on access to housing finance is the estimate of, the coefficient of theinteraction term of the treatment and time dummy (Puhani, 2008), i.e. Tt.and Subsidyit