Public Policy 253 Sept. 22, 2003Notes on Poverty Reduction Strategy Paper for GuatemalaThese are some notes that came to my mind while I was reading your reports. First I wanted to share with you thecriteria that I used: I gave one point for each question:2.1 Review GUAPA and summarize/ critique2.2 Develop a strategy to use the LSMS data to investigate your critique3.1 Describe the poor and their poverty. Use a t-test or other test for significant difference. (The addition of a t-testrequirement was done by email, and most of you did it, so there was a ½ point penalty if you didn’t.)3.2 Draw the poverty profile, and calculate P0, P1, and P2 for the sample population.3.3 Draw the poverty profile, and calculate P0, P1, and P2 for a subgroup.3.4 Look at the subgroup’s 3 P-statistics and discuss.3.5 Look into the possibility of gender and age biases in poverty.4.1 Create an index and estimate the probability of being poor given some indicators.4.2 Compare your predicted and the actual values in the sample.I added one general point for making it through the first assignment and made the score out of ten. The mean on theassignment was 8.8 out of 10, which was pretty good. I’m not sure about how the numbers translate to letters- that’sup to the professors.*** I was very disturbed to find very frank evidence that groups are copying each other’s work. As I said before, Iencourage you to consult each other frequently, and some results are going to be the same, but each group shouldhand in their own work. Not all groups followed this. I hope that this does not recur on the next assignment.***Some specific notes:1. PresentationPresentation was by and large pretty good. Your reports were well-labeled and very readable- thank you! The onedifficulty many people had was in presenting numbers. Although it’s easier to read text than numbers, sometimes asyou know it’s easier to flip through a report and scan for the numbers, so your reports should be amenable to suchscanning as well. Please do NOT tell me to go digging through the several hundred pages of Stata output youattached- going through that is YOUR job. Please present the relevant information and interpret it for me, rather thangiving me a log file. (I’m happy to have the log file separately, attached to the rest, but I don’t want to have to digthrough that unless I’m curious about how you got some result.) Also, please follow the assignment’s structure,preferably labeling each section clearly.One important help in being clear is to present your poverty results in a table. It makes contrasting the groupseasier. In addition, do not report too many decimals, as they are not meaningful. They give the false impression thatthose numbers are precise. In fact if you computed the standard error on these numbers, you would be surprise tosee how imprecise they are. For example for the number of people in poverty you might get something like.4604503% with a standard deviation of 1.10%. Hence it is not meaningful to report a more precise number than46.0%If you use weights (which is the correct way to compute aggregates, but this was OK not to use them for thisexercise), then the population that is reported is the number of observations inflated to the real population size by theweights. Always report the population size, so that one sees how important those groups are relative to each other.Example (from another dataset):Table 1. Poverty in Nicaragua, 1998Number Headcount Depth Severityof persons P0 P1 P2Overall 4,794,131 39.2 13.9 6.6By locationUrban 2,606,139 22.9 7.0 3.0Rural 2,190,539 58.5 22.2 10.9By genderMen 2,351,186 40.8 14.6 6.9Women 2,442,945 37.7 13.3 6.3By education of head of householdNo education….Source: Nicaragua, LSMS 1998 PO, P1, and P2 are all in percentPoverty indicators2. StrategyIn the first part, you were asked to summarize the report and look for weaknesses, such as the fact that the reportdidn’t detail the breakdown of poverty among different groups. Next you were asked to say how you’d remedy thisusing the LSMS, i.e. “With the LSMS data, we could check to see if there are gender associations with poverty.”3. Poverty profileMost of you did a pretty good job drawing the required graph, but because of outliers it doesn’t look like much. (Abackwards L doesn’t tell us much about the shape of the graph for the majority.) Be sure to lop off the extreme folksso that we can get a better picture.4. Significant differencesI sent out an email a few days before the assignment was due asking you to use a t-test to check whether thedifferences you talked about in parts 3.3 - 3.5 were statistically significant. Most of you did this, and for those whodidn’t, it’s quite easy to do: in Stata, the command is ttest variable, by(group). For example, ttest belowpov,by(urban) tests whether the variable belowpov takes different values, on average, for urban and rural households.Also some of you weren’t sure about what the results mean: all you need to note is whether the difference issignificant, and if so, which one is significantly larger.5. Gender biasMost of you interpreted this to mean checking on the significance of male vs. female-headed households. That’s notwrong, and that interpretation received full credit, but think about what else you could do- you have the number ofmales & females in each household, so you could see if more men than women overall are in poverty. Here’s anexample of some relevant output:. ******* question 3.5. *******;. ****** poverty rate by male/female *******;. g nummale=hhsize*por_male/100;. g numfemale=hhsize-nummale;. sum p0 [w=nummale];(analytic weights assumed) Variable | Obs Weight Mean Std. Dev. Min Max-------------+--------------------------------------------------------------- p0 | 6914 18497.9998 57.01697 49.50875 0 100. sum p0 [w=numfemale];(analytic weights assumed)Variable | Obs Weight Mean Std. Dev. Min Max-------------+----------------------------------------------------------------- p0 | 7065 19273.0001 55.90723 49.65333 0 100. sum p1 [w=nummale];(analytic weights assumed) Variable | Obs Weight Mean Std. Dev. Min Max-------------+----------------------------------------------------------------- p1 | 6914 18497.9998 22.91091 25.36843 0 92.25044.
View Full Document
Unlocking...