Female-Headed Households and Homeownership in Latin America

The gender of the household head has often been treated as an exogenous determinant of housing tenure. We argue that several determinants of homeownership also affect household headship and that failing to explicitly account for this endogeneity leads to inconsistent results. Using individual level data for Chile, Honduras and Nicaragua we show that although on average women have lower probability of being homeowners, those women that head their families (single, separated or divorced) have larger probabilities of attaining homeownership. Thus household level analysis should control for the endogeneity of household headship in order to properly address the gender effect on housing tenure. We estimate a bivariate probit model and find evidence that all else equal female headed families have lower probability of owning their home in Latin American countries. Without the endogeneity control this result was not present in eight countries. * The authors wish to thank Alexis Avcharian for excellent research assistance and the MECOVI program for help in the access to the household surveys data. This paper benefited from comments from Marina Bassi, Arturo Galindo, Eduardo Lora, Andreas Moro, Hugo Ñopo, Claudia Piras, Claudio Santibañez, Victoria Rodriguez. The usual disclaimer applies. 2 As stated in IADB (2004) “Poverty is both cause and effect of poor housing conditions. Lack of effective demand resulting from the low income of households is the underlying cause that prevents the private provision of houses.... Conversely, improving housing conditions can have a major influence on poverty alleviation through improvements in the living standards of low income families, and on poverty reduction via increased employment opportunities.” Therefore, understanding the determinants of housing tenure and potential gender discrimination is important for poverty reduction policies. The study of the determinants of housing tenure and the concerns with possible discrimination has been on the research agenda even before appropriate econometric techniques were commonly used. Li (1977) is the first paper that goes beyond linear models and estimates a logit model to the determinants of homeownership but does not consider the gender of the household head. Several types of variables have received most of the attention of the researchers: income and wealth, life cycle status, location and neighborhood attributes and a variety of socioeconomic indicators. In particular, much attention has been given to the racial or ethnic origin of the father. There is substantial evidence of racial discrimination in the access to mortgage credit and homeownership. The gender economic discrimination literature has also spent lots of efforts to study the existence of discrimination on dimensions like salaries, promotions, etc. One common strategy is to include an explanatory variable indicating the presence of women and to conclude that if the estimated coefficient is significantly different from zero, females or female headed households receive a discriminatory (positive or negative) treatment. It is therefore striking the absence of comments on gender differentials in the studies of the determinants of homeownership. The reason is that most studies find more favorable outcomes for female headed families or do not find significant results at all. Given the outcomes, in other contexts, of the gender 1 Van Leuvensteijn and Koning (2004) and Gandelman and Gandelman (2004) find that female headed households have higher probabilities of owning their household in the Netherlands and Uruguay respectively. Chiuri and Jappelli (2003) and Arimah (1997) do not find gender differences in fourteen OECD countries and Nigeria respectively. Manrique and Ojah (2003) found that men headed households are more likely to own their household but female headed households tend to have higher household expenditure in Spain. 3 discrimination literature these results are surprising. We argue that the determinants of women household headship and those of homeownership are correlated and therefore the specification used in most studies has an endogeneity problem that leads to inconsistent and often counterintuitive results. If female household headship is not exogenous to the tenure choice, then, even in the presence of lower probabilities of homeownership, a naive view of the data may reflect that women headed households have higher probabilities of owning their home. For instance, that have lower income, more children, etc. will probably not divorce their husbands even if they want to. There is a selection bias in which women headed families tend to have better socioeconomic indicators than what they would have if female headship were a completely random process. Thus, the gender of the household head can not be treated as other truly exogenous characteristics like race and ethnic origin. To the best of our knowledge this is the first paper that focuses on the factors affecting homeownership and household headship jointly by explicitly providing an econometric solution to the endogeneity issues that arise by the joint determination of both variables. Our results for seventeen Latin-American countries show that the biases are important and that female headed families have a substantially lower probability of attaining homeownership.