Multidimensional housing deprivation indices with application to Spain

Abstract

The main aim of this article is to define a multidimensional housing deprivation index and identify the main determining characteristics of this phenomenon, using Spain as reference. A latent variable model is used in order to overcome some of the traditional difficulties encountered in multidimensional deprivation studies. The construction of a latent structure model has allowed a set of partial housing deprivation indices to be grouped together under a single index. It has also enabled each individual to be assigned to a different class depending on the level and type of deprivation. Results show that the vector of observed variables (having hot running water, heating, a leaky roof, damp walls or floor, rot in window frames and floors and overcrowding) and the correlations among such variables can be explained by a single latent variable. There are also specific characteristics that differentiate the population affected by housing deprivation.

Notes

¹ The lack of commodities caused by individual preferences should not be considered as a deprivation situation.

² Kamanou (2000) used an alternative version of the main components analysis technique based on a standardized uniform transformation of the set of discrete variables comprising the household wealth index. This approach allows one to take into account differences in the variance of the variables used to construct the index.

³ An alternative would consist of assessing the consistency of the deprivation indicators by estimating the Cronbach alpha coefficient. Nevertheless, the use of such methods suffers from some important constraints (Moisio, 2001).

⁴ We use the EM algorithm proposed by Bartholomew and Knott (1999).

⁵ Ayala et al . (2005) estimated to what extent living in poor housing conditions could determine individuals’ health status using the 1998 ECHP data.

⁶ The modified OECD scale is applied (taking a single-person household as a reference and giving a weighting of 0.5 to the rest of the adults and 0.3 to children under 14 years old).

⁷ Health status is defined based on a self-assessment made by the individuals themselves: very bad, bad, regular, good and very good.

⁸ Another possible way of defining overcrowding could be to consider the number of rooms available corrected by an equivalence scale (Chiappero, 2000).

⁹ Other studies have also found that the results of the overcrowding variable are different from the other housing indicators (Marsh et al ., 1999).

¹⁰ We consider income as one of the covariates explaining housing deprivation despite this variable was one of the criteria used to select housing indicators. This procedure allows us to identify income's influence on housing deprivation once we have controlled for other factors. Furthermore, it is also interesting to test whether or not belonging to different income deciles cause different effects on the probability of belonging to a specific class of housing deprivation.

¹¹ Households with the highest income levels (ninth decile) can serve as an example. They suffer a relative risk of belonging to the second class that is 57% lower than that of the households in the first decile. This percentage falls to 49% for the third class and rises to 81% for multiple deprivation.

¹² Several studies point out the fact that the lack of the spouse influence on the stability in their income level (Cantó, 2002).