employment, ownership and access to durable goods, purchase of household durables,
over the past 12 months, access to production assets, nutrition and fertility status, sources
of income and transfers.* The diary sheet was the principal instrument used for collecting
expenditures on a detailed set of food and nonfood items during the month preceding the
interview. For food items, although both quantity and value were listed for each item in
the diary sheet, only the values of items were recorded in the two surveys, Thus, it is not
possible to derive unit costs of individual food items. This, together with the lack of
price information in the village questionnaire, has some important implications for the
calculation of poverty lines (see below). Most importantly, it limits the extent to which
cost of living differentials can be spatially disaggregated. In this study, however, we
computed the regional costs of living indices based on the monthly average prices that
were available for 5 cities in 1997-98, Thus, we could to some extent take account of
regional costs of living differences that exist in the country.
3. Measure of welfare
We chose per capita consumption as the measure of individual welfare. While
both income and consumption are plausible measures of welfare, the latter is often
considered preferable for both theoretical and practical considerations, especially for
developing countries (Deaton, 1997). For Lag PDR in any case, LECS I did not collect
information on incomes, and hence consumption was the only feasible option for this,
study. We also use a per capita normalization which assumes that household needs
increase proportionally with household size. This rules out scale economies in
consumption and treats all members as adult equivalent. We use the per capita
normalization not because either of these is a realistic assumption, but because there do
not exist any ctedible estimates of the scale economies or adult equivalence parameters.
Nor is it possible to uniquely identify these parameters from observed demand behavior
(Deaton, 1997). ‘Thus, rather than use more or less ad hoe values of these parameters, we
prefer to use the per capita normalization that is both simple and transparent. This
“Data on incomes were not collected in LECS I (1992-93). Further information on the surveys can be
found in Lundgren and Safstrom (1991) and Backstrom and Safstrom (1997).