Enabling poor rural people
to overcome poverty



This section combines some of the single-factor indicators of HFS analysed above in an attempt to characterize household vulnerability to food insecurity in a more comprehensive manner. This simple exercise is intended to lay the groundwork for further research on HFS indicators, while still providing project staff and policy-makers with useful insights on the HFS debate in two geographical areas of IFAD intervention.

In order to construct the indicator, the household must be characterized on the basis of the following features: (i) food production/food market dependency; (ii) income; (iii) asset ownership; (iv)income diversification; and (v) crop diversification. The proxies used for each of these components are, respectively, the household’s corn market dependency ratio14, total household income, liquid asset stocks15, the number of income sources and the number of crops grown. The first variable is included in the indicator to reflect the source of household food supply. The second indicates the household’s ability to access food through earned revenues. The third reflects the household’s ability to cope with short-term food shortages, while the last two variables are indicative of the household’s strategy in reducing the risk of entitlement failure. Each individual variable is ranked from worse off to better off, and the observations are grouped into terciles16. It is assumed that belonging to the low tercile contributes nothing to the HFS measure; the medium tercile, 1 point; and the high tercile, 2 points. The partial scores are added for each individual household.17 Total scores of up to 3 points are presumed to reflect extreme vulnerability, while from 4 to 7 indicates medium vulnerability and from 8 to 12, low vulnerability.18

The results of this simple exercise are reported in Table 17 and Figure 1.19 Households in Suchiquer show, on average, the highest overall vulnerability to food insecurity, with a striking 60% of households falling into the category of extreme vulnerability and less than 9% characterized by low vulnerability. Tituque follows, with almost 30% of its households in the extremely vulnerable group.20 Tituque and the NTX communities present less egalitarian distribution in terms of the indicator, with approximately one third of households in each tercile.

In almost all communities, participation in IFAD projects is positively correlated with lower levels of vulnerability to food insecurity. The exception is Suchiquer, the community with the greatest exposure to food insecurity, where project beneficiaries appear unable to improve their ability to cope with food inadequacy, at least for those determinants included in the present indicator. Based on the data at hand, it is still unclear whether the project can be given credit for the improvements in all the other communities, or whether the relationship is merely a reflection of pre-existing heterogeneities between the two groups.

In Table 16, the figures associating the adoption of NTX with household vulnerability to food insecurity reveal that fewer NTX adopters indicate extreme levels of food vulnerability compared with non-adopters. Once again the extent of the causality is uncertain. Whether the adoption of NTX can be given credit for lessening household vulnerability, or whether superior household endowments led to the choice in the first place, remains unclear. Nevertheless, the findings point to the fact that, based on our assumptions, NTX households are currently less vulnerable to food entitlement failure and better endowed to cope with future problems emerging from the production of NTXs.

Finally, Table 18 and Figure 3 present data on the relationship between household vulnerability and women’s socio-economic status. Regarding the latter, the amount of income earned by women from petty trading, as well as the share of the household’s petty trading income controlled by women, have been used as proxies. The figures appear to support the existence of a weak but positive relationship between women’s income and HFS. The findings, however, constitute grounds for further investigation, since the correlation may be the result of spurious factors and not women’s income alone.

To this end, Table 19 reports the results of a regression analysis in an attempt to simultaneously relate our indicator to some exogenous features characterizing the households surveyed. The coefficient on total cultivated land is positive and significant, indicating a positive relationship between household wealth and food security. A look at the differences in income sources can substantially add to the assessment of HFS. The regression results reveal a negative relationship between the share of off-farm income and food security. The causality of the relationship is uncertain, since it may be the case that food-insecure households are more likely to seek work outside their farm or community. Alternatively, the negative coefficient could be the result of the direct impact of off-farm labour on the household’s well-being as a result of prolonged absence from the home or issues related to income control (off-farm labour is almost exclusively a male activity). On the other hand, households deriving higher shares of their income from NTX production appear to be better off from a food security standpoint. The coefficient on women’s income is positive but not strongly significant – possibly as a result of the relatively low returns and insufficient income women have been able to earn so far from petty trading (the mean sample is less than Q 500 per year).

In contrast with what is stated in the literature (Haddad et al. 1994), the larger families in the sample appear relatively better off in terms of the HFS indicator, although the marginal effect is quite low. However, households demonstrating higher dependency ratios21 are more vulnerable from a food security standpoint, indicating the importance of household composition for the purpose of food security targeting. Also, households with an older female household head appear less prone to food insecurity. This variable is likely to capture the effect of unobservable endowments, skills and experience that are associated with women’s age. Furthermore, although the coefficient is not strongly significant, households in which the women are better educated appear to be less food insecure.

As shown earlier in the univariate analysis, project participation appears to have a positive correlation with food security. However, it is not possible – based on this data – to make any inference on the causality of the relationship. Finally, significant community-specific heterogeneities are captured in the set of community dummies. Holding all the above features constant, households in the Cuchumatanes project are better off than households in Tituque, the community numeraire. On the other hand, the negative coefficient for Suchiquer (the pure subsistence system) once again confirms the significantly poorer conditions and more deficient endowment of households in this community, even when compared with a not so well off community such as Tituque.

In summary, the preliminary findings endorse the use of composite indicators in locating the most vulnerable groups from a food security standpoint. They also illustrate the important role played by IFAD projects in reducing household vulnerability to food insecurity across the communities surveyed. Furthermore, the univariate analysis initially hinted at the existence of a weak but positive relationship between women’s income and HFS. The regression analysis confirmed the feeble nature of this correlation, as a result perhaps of women’s income levels being insufficiently high for this to trickle down to improved food security. However, further investigation on this relationship is needed in order to obtain a clear assessment of the role of women’s income in the HFS strategy.


14/ As has been repeatedly mentioned, the author is well aware that higher dependency on food markets may not always translate into higher food insecurity, particularly in the presence of high uncertainty in production. However, the claim in constructing the composite indicator is that extremely high food dependency - combined with low income, inadequate diversification and poor assets - is likely to be associated with extreme vulnerability to food insecurity.

15/ Liquid assets include ownership of chickens and pigs. As these assets are considered to be the most 'liquid', they are therefore the most relevant to the household coping strategy dealing with short-term food shortages.

16/ For the total income variable, the low tercile includes households with income up to Q 2 340, the medium tercile, incomes between Q 2 341 and Q 5 180, and the high tercile, incomes above Q 5 180. For corn market dependency, the low tercile includes values of the dependency ratio between 57 and 100%, the medium tercile, between 17 and 56.9%, and the high tercile values between 0 and 16.9%. The values for the income diversification terciles are one source for the low tercile, two sources for the middle, and more than two sources for the high tercile. For crop diversification, the low tercile includes all households growing up to two crops, the middle tercile, three crops and the high tercile, more than three crops. The threshold values for animal stocks are up to 5 chickens and 0 pigs for the low terciles, between 6 and 13 chickens and 0 pigs for the middle terciles and more than 13 chickens and at least 1 pig for the high terciles.

17/ The relationships between this household-specific measure and the single-factor indicators that compose it are mapped out in Figure 2.

18/ The indicator constructed is by no means an absolute indicator of vulnerability. The only scope of the preliminary exercise is to locate those households that exhibit simultaneously all those critical features that are likely to be negatively associated with food security - namely, low income, low staple production, low diversification and scarce assets. As such, the indicator can be used with some confidence to identify the most vulnerable groups within a region (relative to the remaining population) from a food security standpoint. Furthermore, attempts to relate this type of relative measure to some benchmark indicator should be pursued in further research.

19/ In Figure 1, the first bar in each production system (marked with B) refers to beneficiary households, the middle one (marked with NB) to non-beneficiaries, and the third and last one (marked with ALL) to the total.

20/ Tituque had scored substantially worse in a computation of an alternative set of indicators that did not take into account crop diversification.

21/ The household dependency ratio was computed as the number of household members below the age of 13 and above the age of 65 over the total family size.