Poverty estimates: Robust methodology needed

Multidimensional Poverty Index is a good step, but it seems little attention was paid on the need of adding some critical variables and improving the methodology in a particular country context for minimising the limitations of global measures

The National Planning Commission recently unveiled the new poverty estimates based on the multidimensional approach in collaboration with the Oxford Poverty and Human Development Initiative (OPHI) at the University of Oxford. This comes following earlier criticism of the consumption- and income-based poverty estimates, which are calculated solely on the basis of monetary value measurement techniques hence they barely reflected the multiple factors perpetuating poverty in countries like Nepal. Therefore, initiation and institutionalisation of the multidimensional approach is a welcome development, as it, in principle, recognises various forms of deprivations — both vertical and horizontal — as the root causes of poverty.

Equally importantly, the development discourse envisaged by the new constitution and Sustainable Development Goals (SDGs) set for 2030 entails the need of such an approach for directly linking policies and programmes with deprivation related multiple factors contributing to poverty.

The new poverty numbers thus derived are based on the computed Multidimensional Poverty Index (MPI) using multiple indicators related to health, education and living standards. The new estimates are based on the Multiple Indicator Cluster Survey (MICS) of 2014 by the Central Bureau of Statistics. Based on the MPIs, the new poverty level has been estimated to be 28.6 per cent on the average for 2014. This is marginally higher than consumption-based poverty at 23.8 per cent for the same year.

More broadly, under-nutrition and lack of completion of five years of schooling have been shown as the major causes of such a poverty level in Nepal. In such exercise, poverty has been derived separately for all provinces as well which, apparently, are in the expected line with highest poverty in Province 6 at 51.2 per cent followed by 47.9 per cent in Province 2 and 33.6 per cent in Province 7. Similarly, variation in rural and urban areas has been high with only 7 per cent urban population in poverty compared to 33 per cent rural population in poverty. The urban poverty in this particular exercise shows an opposite trend from the consumption-based poverty pattern, indicating a rising poverty tendency in the urban areas in recent years. Probably, more urban leaning indicators included in the living standard index has had more pervasive poverty reduction effect.

Notwithstanding such a good initiative, it, however, seems little attention was paid on the need of adding some critical variables and improving the methodology in a particular country context for minimising the limitations of global measures. Needless to add, such a necessity has been persistently advocated by researchers to make such estimates more plausible or useful from policy and programme standpoint. For poverty analysis under multidimensional framework, critical variables omitted in the global exercise must have been included. Such variables include access to land assets, food security, productive employment and social security to represent vulnerability. Similarly, access to connectivity and market as well as financial services is equally crucial in the market economy. Instead, as in the global method, relatively urban biased indicators have been included in the living standard index which perhaps is the main reason for deriving of very low poverty in the urban areas.

From the methodological standpoint also, there are a number of problems inherent in the global methodology which have been replicated in the country exercise. First, the weight for the three groups — health, education and living standard — is arbitrary, impacting the overall estimate. Similarly, in the weights, both output and input indicators are lumped together. Here comes the prices-linked issue too with probable effect on arbitrarily fixed weights. For these reasons, some suggest a dashboard approach to overcome the methodological pitfalls, while others are still reluctant to use MPIs in their analysis.

In the Nepali context, perhaps there is a credibility problem of the estimates as well. The first estimates following the same methodology based on the Demographic and Health Survey of 2006 had shown multidimensional poverty in Nepal at 65 per cent. Thereafter, the new estimates made based on the Demographic and Health Survey of 2011 had shown that such poverty in Nepal is 44.2 per cent. Now based on a different survey, which is MICS, poverty has been derived to be 28.6 per cent. This means, during the first five-year period of 2006-2011, poverty reduced by 20.8 per cent with an annual decline by 4.16 percent. In more recent three-year period of 2011-2014, poverty decelerated by 15.6 per cent with a fall of 5.2 per cent annually. Among others, how the additional data required other than available from the MICS was made available is also not explicitly mentioned.  Apparently, such a sharp fall is not corroborated by the major performance indicators of the economy.

The methodological aspects, data coverage and comparability might have affected the estimates, which is not unusual. The two types of surveys substantiate this. Nonetheless, given the importance of MPIs toward internalising them in planning and policy making, concerted efforts are required to make the methodology more robust.

Khanal is an economist