Life cycle and housing decisions: a comparison by age cohorts

Abstract

The use of decomposition methodologies when the involved variable is continuous is not common in the literature. This article uses this methodology, together with other decomposition methodologies, to explain how age can influence on housing decisions. In particular, we use Spanish data to study whether the age of the householder plays a significant role in influencing household decisions with respect to housing tenure and demand. From the comparison of housing decisions between different groups of households classified by the age of the householders, we conclude that age plays the primary role in explaining the gap between households regarding tenure choice, while it shares its importance with other covariates of the model in the housing demand decision.

Keywords:

• decomposition analysis
• housing demand
• housing tenure choice
• life cycle pattern-age profile

JEL Classification:

• D12
• J71
• R21

View correction statement:

Erratum

Acknowledgements

The authors thank Instituto Valenciano de Investigaciones Económicas for their helpful collaboration in the provision of statistical information. We further acknowledge the financial support from the Generalitat Valenciana (grant GR00-80). Víctor E. Barrios also thanks the Spanish Ministry of Science and Innovation (grant ECO2009-13669) for financial support. We thanks two anonymous referees for their helpful comments.

Notes

1. Home ownership can guarantee a minimum level of consumption for the elderly in their stage of lower income, when they can least afford to pay high rents for a house. Furthermore, they can use this good as a reverse mortgage, although this last option is only available on dwellings with special characteristics.

2. The choice of 35 years as a threshold of separation between the two groups is considered a reasonable assumption for the Spanish real estate market (see Bover, 1993, Colom et al., 2002, Taltavull, 2007).

3. See Ñopo (2008).

4. See Fairlie (2005) and Bauer and Sinning (2008).

5. As Oaxaca (1973) states, ‘the magnitude of the estimated effects of discrimination crucially depends upon the choice of control variables for the outcome regression’. If it were possible to control for virtually all sources of variation in outcome variable, then the unexplained gap would constitute a small portion of the outcome gap, and we could eliminate discrimination as a significant factor of mean outcome differences.

6. See Oaxaca (1973).

7. Equation 4 is a special case of Equation 5 where vector β* is defined as (β_0B, β_1B).

8. Following Jann (2008), we include a dummy variable (group indicator) as an additional covariate in estimating the linear regression model with the pooled sample, but it is not used to calculate the decomposition. ‘It avoid that it may inappropriately transfer some of the unexplained part of the differential into the explained component’.

9. See Neuman and Oaxaca (2004).

10. Bauer and Sinning (2008) generalize the Blinder–Oaxaca decomposition method to other nonlinear models with discrete and limited dependent variables. A recent revision of these decomposition methodologies can also found in Fortin et al. (2011).

11. Sonstelie and Portney (1980) and Linneman (1980) agree that imputed rent rather than house value should be used in studies of housing demand.

12. Friedman (1957) states that individuals base their consumption decisions on a longer term view of an income measure, perhaps a notion of lifetime wealth or a notion of wealth over a reasonably long horizon. This long-term income or permanent income is reinforced in the case of housing through the mortgage market.

13. See Goodman and Kawai (1982), Manrique and Ojah (2003) and Rodríguez (2004).

14. The results of the estimation of the income equation by OLS can be obtained from the authors upon request.

15. See Wald’s joint test of significance in the case of categorical variables with more than two categories.

16. In this regard, authors such as Darby (1972), Dynarski and Sheffrin (1984) and Bourassa (1995) conclude that although decisions about housing tenure are more likely based on expectations regarding permanent income, transitory income can also have an important influence on housing decisions. Other authors, such as Goodman (1988), argue that transitory income might not be significant in the tenure choice decision as purchasing a house typically entails substantial transaction costs that might not be covered by transitory income.

17. This interaction test analyses whether the vector of coefficients of the model differs systematically between both groups of households. For that purpose, the null hypothesis that the coefficients of both a dummy variable representative of age (0 if age is under 36 and 1 otherwise) and all of its interactions with the rest of the predictors of the model are simultaneously equal to zero is tested.

18. With respect to this question, Ioannides and Rosenthal (1994) suggest that the inclusion of a correction term in the demand function, which proceeds from a previous binomial probit model with the same explanatory variables, can generate problems of multicollinearity, making insignificant the correction selectivity term.

19. Results of the two-step method of Heckman can be obtained from the authors upon request.

20. See Section II.