22
Problems with Measuring Intrahousehold
Resource Allocations
There are numerous methodological difficul-
ties with trying to test hypotheses concerning
intrahousehold allocation issues, or the magni-
tude of their importance. For instance, as men-
tioned earlier, it is difficult to measure indi-
vidual nutrient intakes for children who are
receiving breastmilk or are eating from a com-
mon family pot.
Measuring how much control different fam-
ily members have over income, in order to test
hypotheses of whether women spend their in-
come differently than men, is even more com-
plicated. One problem is how to determine, in
households where there is more than one par-
ent, how much relative control over income
each parent has. For instance, when a woman
goes to the store to make a purchase, is she
acting on her own preferences or, rather, acting
on instructions, explicit or implicit, from her
husband or someone else. As Gittelsohn (1992)
observes, household allocative behaviors: (1)
frequently occur "behind close doors" making
them difficult to observe; (2) are often sensitive
in nature making them difficult topics to sur-
vey; and (3) are made up of many little activi-
ties, making them difficult for respondents to
recall.
One means of avoiding these problems has
been to simply compare single (female) parent
households with those in which a male parent is
present. The latter households are often implic-
itly assumed to be "male-headed" households
in which the fathers make most or all of the
spending decisions. This assumption may be
dubious unless supported by sociological evi-
dence. Other analyses have tried, despite the
methodological hazards, to go further and dif-
ferentiate, in two (or more) parent households,
between income controlled by men and that
controlled by women. This may be done, for
example, by assuming that subsistence income
from crops on land cultivated by women is
controlled by women, whereas, say, cash crop
income is controlled by men. Or nonfarm in-
come income may be assumed to be controlled
by the parent earning the wage. In some cases,
these approaches may be reasonable, in other
cases not. A problem is that little evidence ex-
ists regarding the validity of such approaches in
various contexts.
But even where the amount of household
income under a particular parent's control can
be reasonably established (e.g., one-parent
households), a second problem is how to avoid
biases resulting from the difficulty, mentioned
earlier, of empirically distinguishing between
reallocations of expenditures caused by differ-
ing allocations of time, and those caused by
differences in preferences amongst household
members (Hoddinott and Haddad 1991). That
is, results which suggest that increases in in-
come controlled by women lead to greater in-
creases in food expenditures than do equal in-
creases in income controlled by men, might be
due to increased purchases by women of higher-
priced foods which require less preparation time
(since the opportunity cost of their time is in-
creased), rather than due to inherent prefer-
ences for spending more money on food. In
fact, the effect of increasing opportunity costs
of women's time could have adverse conse-
quences on the nutritional quality of food such
that calorie and nutrient intakes may be even
less, despite greater food expenditures (Franklin
and Harrell 1985).
One approach to this methodological prob-
lem has been to count only unearned income.
*
Because it is independent of current household
labor decisions, this measure may be useful for
abstracting from the price effects that wages
would represent (i.e., the opportunity cost of
time). However, this approach suffers from at
least four drawbacks. First, unearned income
rarely accounts for a significant share of total
household income (Hoddinott and Haddad 1991;
*
Unearned income refers to such income sources as
pensions, gifts, or earnings on assets, which do not
result directly from one's own labor.