This study uses a relatively new method called “small area estimation” to estimate various measures of poverty and inequality for the provinces, districts and villages of the Lao People’s Democratic Republic (Lao PDR). The method was applied by combining information from the 2002-03 Lao Expenditure and Consumption Survey and the 2005 Population and Housing Census.
The results indicate that the poverty rate (P0) in the Lao PDR is greatest in the remote areas of the east and southeast along the Vietnamese border. Poverty rates are intermediate in the lowland area of the Mekong River basin in the west. The lowest poverty rates are found in Vientiane and other cities. These estimates are reasonably accurate for the provinces and districts, but the village-level estimates must be used with caution since many are not very precise. Comparing these results with previous estimates of poverty, we find a fairly good agreement among the different studies.
Mapping the density of poverty (the total number of poor people in a given area) reveals that, although the poverty rates (the percentage of a population living below a specific poverty line) are highest in the remote upland areas, these are sparsely populated areas, so most of the poor live in the Mekong River valley, in Vientiane, and in Savannakhet.
In the Lao PDR inequality in per capita expenditure is relatively low by international standards. It is greatest in urban areas and in parts of the northern upland areas and lowest in the south and central highlands, and on the Boloven Plateau.
District-level poverty is very closely associated with district-level average per capita expenditure. In other words, inequality does not explain much of the variation in poverty across districts. This study also explores how the spatial patterns of poverty depend on various geographic factors using a global spatial regression model (in which coefficients are constant across space) and a local model (in which coefficients vary across space). In the global model, geographic determinants, including agro-climatic variables and market access, are able to explain the variation in village-level rural poverty to a large extent. Poverty is higher in villages with a rough terrain, higher seasonality in rainfall and located farther from towns and major rivers. By contrast, poverty rates are lower in areas with more flat land, with higher annual rainfall and a greater annual temperature range. These agroclimatic and market access variables are not as successful in explaining urban poverty.
The local regression model reveals that terrain roughness is associated with higher poverty throughout the Lao PDR, but more strongly so in areas where poverty rates are comparatively low and agricultural production is most commercialised and mechanised. The availability of flat land, on the other hand, is most closely related to lower poverty rates in remote upland areas where flat land tends to be particularly scarce. Access to markets measured as travel time to towns has the strongest positive association with poverty in areas where poverty rates are lowest, and agricultural production is most intensive. Overall, the relationship between agroclimatic variables and poverty varies significantly from one area of the Lao PDR to another.
Many anti-poverty programs in the Lao PDR are geographically targeted. The results from this study indicate that it may be possible to improve the targeting of these programs by making use of more precise estimates of poverty at the district and village level.
The ability of market access and agro-climatic variables to explain a large part of differences in rural poverty rates indicate that poverty in the remote areas is linked to low agricultural potential and lack of market access. This illustrates the importance of improving market access. The fact that poverty is closely related to low agricultural potential suggests that efforts to restrict migration out of disadvantaged regions may not be a good strategy for reducing rural poverty.
Finally, the study notes that the small area estimation method is not very useful for annual poverty mapping because it relies on census data, but the inclusion of a small set of questions on specific housing characteristics in the agricultural census would make a more frequent updating of detailed rural poverty maps possible. Furthermore, it could be used to show detailed spatial patterns in other variables of interest to policy makers, such as income diversification, agricultural market surplus and vulnerability. Lastly, it can be used to estimate poverty rates among vulnerable populations too small to be studied with household survey data, such as the disabled, small ethnic minorities or other population segments.