13.07.2015 Views

A Simple Poverty Scorecard for the Philippines

A Simple Poverty Scorecard for the Philippines

A Simple Poverty Scorecard for the Philippines

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

etween <strong>the</strong> estimate and <strong>the</strong> true value is within 0.5 percentage points of <strong>the</strong> averagedifference. In <strong>the</strong> specific case of <strong>the</strong> national line and <strong>the</strong> validation sample, 90 percentof all samples of n = 16,384 produce estimates that differ from <strong>the</strong> true value in <strong>the</strong>range of 0.6 – 0.4 = 0.2 to 0.6 + 0.4 = 1.0 percentage points. This follows because 0.6 is<strong>the</strong> average difference, and +/–0.4 is its 90-percent confidence interval. The averagedifference is 0.6 because <strong>the</strong> average scorecard estimate is too high by 0.6 percentagepoints; it estimates a poverty rate of 32.3 percent <strong>for</strong> <strong>the</strong> validation sample, but <strong>the</strong>true value is 31.7 percent (Figure 2).7.2 Sample-size <strong>for</strong>mula <strong>for</strong> estimates of poverty rates at a pointin timeHow precise are <strong>the</strong>se point-in-time estimates? For a range of sample sizes,Figure 9 reports average differences between estimated and true poverty rates at apoint in time as well as precision (confidence intervals <strong>for</strong> <strong>the</strong> differences) <strong>for</strong> <strong>the</strong>scorecard applied to 1,000 bootstrap samples from <strong>the</strong> validation sample.A related question is, How many households should an organization sample if itwants to estimate <strong>the</strong>ir poverty rate at a point in time <strong>for</strong> a desired confidence intervaland confidence level? This practical question was first addressed in Schreiner (2008). 21As in <strong>the</strong> previous paragraph, <strong>the</strong> answer lies in Figure 9.21IRIS Center (2007a and 2007b) says that n = 300 is sufficient <strong>for</strong> USAID reporting. Ifa scorecard is as precise as direct measurement, if <strong>the</strong> expected (be<strong>for</strong>e measurement)poverty rate is 50 percent, and if <strong>the</strong> confidence level is 90 percent, <strong>the</strong>n n = 30029

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!