Vietnam’s Evolving Poverty Index Map: Patterns and Implications for Policy

In Sect. 6 we note that caution must be exercised in comparing poverty between 1999 and 2009 as the underlying household survey data are not strictly comparable.

We do not present the full regression results to avoid the lengthy paper. However, readers can contact us to obtain these results. We conduct two exercises to shed light on the question how well the models do in predicting per capita expenditure. Firstly, within the 2010 VHLSS, we randomly split the six regions into two samples, and estimating the regional models in the first sample, we predicted expenditure in the second sample and compared the results to the actual per capita expenditure. We find that the observed poverty rates are in all regions’ subsamples similar to the predicted ones. Secondly, we also examine the sensitivity of the expenditure poverty rate of districts and provinces to different expenditure models. We estimate two expenditure models: one with a large number of explanatory variables and another with a smaller number of explanatory variables. Both models give very similar estimates of poverty indexes at the district and province level. For interpretation in this paper, we will use the estimates from the large model, which give lower standard errors of welfare estimates.

Which is different—usually lower—from the expenditure poverty line set by the General Statistics Office and The World Bank that is used by the international research community to study poverty.

Actually, comparing this no-knowledge benchmark against the successfulness of the geographic targeting approach that Vietnam exercised a decade ago as assessed by Baulch and Minot (2002), it is not that extreme. They found that only 20 % of the poor were reached. If the total population gets distributed an equal part of the budget, 100 % the poor would be reached by definition. The only thing is that the amount received is probably so little in that scenario that it is questionable how many people would get lifted out of poverty.

We focus on the squared poverty gap because of its appealing properties from both a conceptual and technical point of view. The basic approach explored here would also work for other poverty measures, particularly FGT measures with values of parameter α greater than 1. However, with the headcount measure (the FGT measure with α = 0) welfare ‘optimization’ is not well defined and the approach taken here is thus less obviously applicable (see for example Ray 1998, pp 254–255).

Following Foster et al. (1984) the FGT class of poverty measures take the following form:

\(FGT(\alpha ) = \left( {\frac{1}{{\sum {w_{i} } }}} \right)\sum {w_{i} (1 - (x_{i} /z))^{\alpha } }\)

where x_i is per capita expenditure for those individuals with weight w_i who are below the poverty line and zero for those above, z is the poverty line and \(\sum {w_{i} }\) is total population size. \(\alpha\) takes a value of 0 for the Headcount Index, 1 for the Poverty Gap and 2 for the Squared Poverty Gap. For further discussion, see Ravallion (1994).

See for example the World Bank Policy Research Report, “Localizing Development: Does Participation Work?” (World Bank 2012).

Mont and Nguyen (2011) show a strong correlation between disability and poverty in Vietnam.

Other interpolation schemes are possible. For instance, if the poverty gap is given at table values zk/n an even simpler computation presents itself. Often the poverty mapping software will give percentiles of the expenditure distribution. These can also be used for interpolation, but the formulas are more cumbersome, since the percentiles are not equally spaced.