Executive summary

Annex 1.3: Are the Changes in Poverty and Inequality Significantly Significant?

Because poverty and inequality indices calculated from household survey data are based on only a sample of the population, both the point estimates at a given point in time and estimated changes over time are subject to sampling error. ⁴⁴ This section presents confidence intervals and tests for the statistical significance of changes in the welfare measures. Confidence intervals were calculated using bootstrap resampling methods.

Table A.1.3.1 shows the results of the test of statistical significance for the changes in the Gini coefficient between 1997 and 2003. Tables A.1.3.2 and A.1.3.3 display the results for the three FGT poverty measures, calculated using both the moderate and extreme poverty lines. Each table shows the change between the estimated measures for each year, the standard error of the change, and the corresponding confidence interval at a 95 percent level of significance. The change is statistically significant if it is possible to reject the null hypothesis of equality between the measure in 1997 and 2003. Each row indicate with an asterisk (*) whether the change is statistically significant.

For the Gini, the changes are statistically significant nationally, for rural areas, and for indigenous areas, but not for urban areas. Among FGT measures, only the moderate poverty headcount at a national level and the extreme poverty gap and severity of poverty in the indigenous area show no statistically significant change between 1997 and 2003.

Table A.1.3.1: Tests of Statistical Significance for Changes in the Gini Coefficient

Source: Own estimation based in 1997 and 2003 ENV data. Note: (P) denote the percentile method and (N) denote the normal-approximation method.

Table A.1.3.2: Tests of Statistical Significance for Changes in the FGT Poverty Measures Calculated with the Moderate Poverty Line

Source: Own estimation based in 1997 and 2003 ENV data. Note: (P) denote the percentile method and (N) denote the normal-approximation method.

Table A.1.3.3 - Tests of Statistical Significance for Changes in the FGT Poverty Measures Calculated with the Extreme Poverty Line

Source: Own estimation based in 1997 and 2003 ENV data. Note: (P) denote the percentile method and (N) denote the normal-approximation method.

Annex 2.1: Rates of Chronic Malnutrition in Same Age Cohort (between 1997 and 2003)

One hypothesis offered to explain this discrepancy is that the 1997 indicator might have been badly constructed due to measurement errors in the field. To examine this, we look at the malnutrition rates among children who were aged six to eleven at the time of the ENV-2003, i.e. children who are in the cohort that was in the 0 to 5 years of age range at the time of the ENV-1997. As can be seen in Table A.2.1.1, at the national level the differences in chronic malnutrition in the age cohort are very small between the two points in time. However, when we look at the differences within specific subgroups (by geographic area) the differences are striking⁴⁵.

Table A.2.1.1: Rates of Chronic Malnutrition in Same Age Cohort between 1997 and 2003

1997: Children ages 0 to 5 2003: Children ages 6 to 11 Differences 1997 to 2003 Chronic (height for age) National 14.3 15.4 -1.1 Urban 5.7 6.2 -0.5 Rural 14.5 15.8 -1.3 Comarca 48.5 58.7 -10.2 Underweight (Weight for Age) National 6.7 4.2 2.5 Urban 2.8 2.4 0.4 Rural 7.1 4.1 3 Comarca 21 12.9 8.1 Acute (weight for height) National 1.1 0.8 0.3 Urban 0.9 1.1 -0.2 Rural 1.1 0.4 0.7 Comarca 1.8 0.5 1.3

Source: Censo de Talla, MINSA/ MEDUC, 2001.