undernourished with the calorie cut-off of 2,100 (the proportion rises from about 58% to
Following the extant literature establishing the important role of food prices in induc-
ing substitutions between different sources of calories, we explore with the NSS data
how sensitive calorie shares are to food prices and whether a calorie threshold exists
and, in case it does, whether estimates of undernourished change substantially.32 For an
assessment of relative importance of expenditure and price effects, their elasticities were
computed. To avoid repetition, our comments are selective and brief. Expenditure elas-
ticities for the poor in rural India remained stable at -0.14 over the period 1993-2004.
By contrast, the elasticity (in absolute value) reduced in urban India (from over -0.20
to -0.17). Among the affluent, their (absolute) values fell more in urban India than in
rural India. In both rural and urban India and in both 1993 and 2004, the (absolute)
expenditure elasticities were lower for the poor.
Confining to food-price elasticities, an important finding is that although they vary
between food commodities, across different subsamples, and over time, they are sub-
stantially lower (in absolute value) than expenditure elasticities, implying that expen-
diture has a much larger influence on calorie shares. This, of course, does not negate the
effects of food prices.
The calorie thresholds obtained from Lowess are lower than those obtained from
robust regressions, whereas expenditure cut-offs are considerably higher. As a result,
the estimates of undernourished vary greatly. Consider, for example, the estimates for
rural India. In 1993, the Lowess estimate of the undernourished is 27% of the population,
whereas the robust regression estimate is 0.3%.33 In 2004, the Lowess estimate was about
three times greater than the regression estimate. Besides, although the rate of reduction
with regression is negligible, it contrasts with a sharp reduction in Lowess estimates.
Similarly, for urban India, the estimate of undernourished from Lowess was about twice
that obtained from robust regression for 1993 and even greater for 2004.
In sum, our analysis confirms that calorie and associated income thresholds are
influenced by several other factors—especially food prices—that are omitted. Since
even the poor substitute in response to changes in food prices, the thresholds change.
Thus, purely notional adjustments to estimate the extent of undernutrition will not do,
because their bases are uncertain.
Poverty-Nutrition Trap (PNT)
An assertion by Deaton and Dreze (2009) that poverty and undernutrition are unre-
lated is contentious.
The efficiency-wage hypothesis postulates that, in developing countries, particularly
at low levels of nutrition, workers are physically incapable of doing hard manual labor.
Hence their productivity is low, which then implies that they get low wages, have low
purchasing power and, therefore, low levels of nutrition, completing a vicious cycle of