There are numerous issues that have emerged as the study has developed and, as far as possible, these have been covered as the background and findings have been written. However, particular mention should be made of two particular issues that appear central to the whole discussion of gender sensitivity in shelter- related interventions.
Categorizing households from a gender perspective
Examining households in general from a gender perspective presents substantial methodological problems as there are relatively few households composed entirely of male or female individuals. Household headship provides a somewhat less complex option and it is for this reason that the head of household was used not only as the informant in this study but also as the study's main unit of analysis. However, classifying households by headship is by no means straight forward and the categorization of headship used in this study does not provide a totally satisfactory alternative to the traditional division of households into male-headed (whether or not there are two adults living as joint partners) and female-headed.
While the categorization used for analytical purposes in this study allows a differentiation between single adult heads, of both sexes, and those living in a joint situation, it does not take into account factors such as the longevity of the head's status as a single or Joint head, the impact of more than two generations being present in the household and the manner in which non-residential partnership may influence household dynamics. These omissions were, in some ways, inevitable because the survey that produced the initial database had not been designed to focus specifically on gender-based differences in household behaviour. However, they are omissions that should be rectified in future research.
Interactions between gender-related variables
It can be difficult to provide an adequate framework for analysis that Is gender-sensitive, because gender differences in specific variables do not necessarily emerge clearly and starkly on a discrete basis. The gender differences that emerge in an area such as shelter are far more likely to be the result of a complex interweaving of small differences in a large number of variables. For example, the differences in land tenure interact with the differences in asset levels and savings patterns to create a complex shelter environment that, in its totality, varies for different types of households in a manner that can, ultimately, be seen to be gender-linked. If a single variable alone is considered, gender-related differences may not be so apparent at first glance.
In the long term, and with more extensive data, it should be possible to develop an index which reflects a range of variables that appear to be gender-related. However, this was beyond the scope of the present study which only used indexation in order to determine variations in dependency rates between different kinds of household.
The index in question, the earner's index. was based on the following formula:
The higher the score on this index, the higher the dependency burden of the household.