Predictability of Poverty Rates in 2011

Basan Shrestha, Research, Monitoring and Evaluation Expert

basan_shrestha@yahoo.com, basanshrestha70@gmail.com

Nepal Living Standards Survey (NLSS) 2011 disclosed that on average 25.2 percent people lived below the national poverty line. It was an improvement from the average rate of 30.8 percent in 2004. However, there were variations that some Village Development Committees (VDCs) and municipalities (MPs) decreased the rates and others increased. Given the poverty rates in 2004, how predictable were the rates in 2011? The answer to this question could give some insights for targeting the rates.

The NLSS 2011 defines the poverty rate as the percent of people below the national line of Nepalese rupees 19,261 per capita annual consumption. The Small Area Estimation of Poverty reports in 2013 and 2006, using data from the NLSS 2011 and 2004 and the Population Census 2011 and 2001 estimated the rates of 976 sub-districts (Ilaka) which constituted 3,926 VDCs and MPs located in 75 districts of all five regions including – 1,215 VDCs/ MPs of 19 districts in centre, 907 of 16 districts in east, 876 of 16 districts in west, 539 of 15 districts in mid-west and 389 of nine districts in far-west. The rates of VDCs/ MPs in those Ilakas were compared to estimate the overall predictability and the regional difference in the rates of 2011 given the rates of 2004.

Overall, the poverty rates in 2004 poorly predicted the rates in 2011 indicating that the rate in 2004 alone was not a good predictor of the rate in 2011. There could be many other factors that explained variation in the poverty rates of 2011. Thus, the prediction of the poverty rates in 2011 only based on the rates in 2004 would be poor.

Overall Predictability

The poverty rates at VDC/ MP level in 2004 and 2011 were fairly normally distributed as mean and median were closer. Rates in 2004 had a moderate positive correlation with the rate in 2011 (Karl Pearson’s correlation coefficient=0.47; maximum value is 1, which means total positive correlation). The statistically significant correlation coefficient indicating with 95 percent confidence that the high rate in 2004 was likely to be high in 2011.

In a regression analysis taking the rate in 2004 as the explanatory variable and the rate in 2011 as the response variable, the positive regression coefficient of 0.47 was significant indicating that for sure each percentage point increase in the poverty rate of 2004, the rate of 2011 would increase by 0.47 percentage point. The small coefficient of determination equal to 0.22 indicated that the rate of 2004 was the poor predictor determining only 22 percent of the variation in the rate of 2011.

The poor predictability could be because in some VDCs/ MPs the rates decreased, increased or stagnated in 2011. 77 percent (3039 of 3,926) VDC/ MPs of 68 districts in all five regions decreased the rates in 2011 than in 2004. Three VDCs (Ahale, Mahabharat and Vedetar) of Dhankuta district in east had the highest decline in the rate by 50.5 percentage points from 65.7 percent in 2004 to 15.2 percent in 2011. Unlike, 23 percent VDCs/ MPs (886 of 3,926) of 36 districts in all five regions increased the rate in 2011 than in 2004. Two VDCs (Chhonhup and Lomanthang) in Mustang district of west had the highest increase in the poverty rate by 36.7 percentage points from 28.8 percent in 2004 to 65.5 percent in 2011. Hetauda MP in Makawanpur district of centre stagnated the rate of 6.6 percent in both 2004 and 2011.

Regional Difference

Centre had the highest positive correlation between the poverty rates in 2004 and 2011 (Karl Pearson’s correlation coefficient=0.67), followed by moderate to lower positive correlation coefficients of 0.52, 0.18, 0.11 and 0.08 respectively in west, east, mid-west and far-west. The correlation coefficients were statistically significant in first four regions and insignificant at the far-western region. It was pretty sure if the rate of a VDC or MP was high in 2004 and that would remain high in 2011 in centre, west, east and mid-west. But, one could not be sure about the relationship between the rates of 2004 and 2011 in far-west.

In a regression analysis, the positive regression coefficients of 0.57, 0.54, 0.14 and 0.13 respectively in west, centre, mid-west and east were significant indicating with 95 percent confidence that for each percentage point increase in the poverty rate of 2004, the rates of 2011 were likely to increase by 0.57, 0.54, 0.14 and 0.13 percentage points respectively in those regions.

Centre had the highest coefficient of determination equal to 0.45 that was significant indicating that the poverty rate of 2004 determined moderately, 45 percent of the variation in the rate of 2011. West, east, mid-west and far-west had the lower coefficients of determination equal to 0.28, 0.03, 0.1 and almost nill respectively showing low predictability. The coefficients in the first three regions were significant and that in the far-western region was insignificant.

The moderate to lower predictability was because 93, 83, 79, 73 and 38 percent VDCs/ MPs decreased the rates in 2011 respectively in west, mid-west, centre, east and far-west and the remaining proportions increased. The range, difference between maximum and minimum poverty rates, in the regions also changed from 2004 to 2011. Eastern VDCs/ MPs turned homogenous since the range decreased most from 63 percentage points in 2004 to 49.9 in 2011. Central VDCs/ MPs continued to be heterogeneous although the range slipped from 80.9 to 72.3 percentage points. Western VDCs/ MPs also continued to be heterogeneous as the range slipped slightly from 70.8 to 70.3 percentage points. Far-western VDCs/ MPs became heterogeneous as the range increased heavily from 37.4 to 51.1 percentage points. Mid-western VDCs/ MPs also turned heterogeneous as the range increased from 54 to 62.3 percentage points.

Conclusively, the prediction of the poverty rates in 2011 only using the rates in 2004 would not suffice. There could be many factors contributing to the rates. This could give insight to targeting the Sustainable Development Goal to eradicate poverty across all corners in coming 15 years commencing 2016.