Josiah Johnston and Suzie ShinDecember 5, 2008ARE253Homework Assignment 3Impact Evaluation of the PROGRESA Program in MexicoI. The schooling situation in control villagesThis section illustrates and discusses the problems of low enrollment in school and high failure rates in rural Mexico. Table I.a below shows the proportion of children in control villages that were not enrolled in school in 1998.Table I.a. Proportion of Children Not Enrolled in 1998Age in 1998 % Not Enrolled11 3.2%12 12.5%13 22.2%14 35.5%15 50.9%16 67.8%The proportion of children that were not enrolled increases with age—the older the child, the less likely he or she is to be enrolled in school. Table I.b below shows the dropout rate in control villages by grade level in 1998.Table I.b Dropout Rate by Grade Level in 1998SchoolGrade Level in 1998Dropout Rate1 26.2%2 13.4%3 9.0%4 7.5%5 4.0%6 33.6%1 7.4%2 2.8%3 50.7%1 18.6%2 0.0%PrimaryJunior highSenior highThe dropout rate is extremely high between primary school and junior high (33.6%) and between junior and senior high (50.7%). It is interesting that dropout rates are lowest in the second to last year of primary school and junior high, immediately preceding theyears with the highest dropout rates. This suggests that students are motivated to complete primary school or junior high, but not to enter the next phase of schooling. The dropout rate in the second grade of senior high is 0%, but this is only out of 20 students, as opposed to over 1,300 students in the sixth grade of primary school. Table I.c shows the failure rate in 1997 by grade level in control villages.Table I.c Failure Rate in 1997 by Grade LevelSchoolGrade Level in 1997Failure Rate1 0.0%2 18.2%3 17.0%4 17.0%5 11.8%6 13.0%1 24.4%2 15.1%3 16.6%1 40.2%2 28.0%3 100.0%PrimaryJunior highSenior highEither the first grade of primary school is so easy that all children pass, or teachers have apolicy of not failing their students in the first grade. The failure rates throughout primary school are fairly consistent, ranging from 11.8% to 18.2%. The relatively high failure rate in second grade may compensate for the fact that all children passed first grade, including some that should have stayed back. The failure rate jumps to 24.4% in the first grade of junior high, perhaps due to a more difficult curriculum or the shock of entering anew school, before settling between 15 and 17%. The failure rate again jumps in the first year of senior high to 40.2%, which is extremely high compared to any of the previous years. Indeed, the failure rate in any grade of senior high is greater than the failure rate inany grade of primary school or junior high. The failure rate for the third grade of senior high was 100%, meaning that none of the XX children enrolled in the third grade managed to graduate. The high failure rates in high school suggest that greater attention should be given to enabling children to advance through high school in a timely fashion. Table I.d shows the dropout rate in 1998 for children that completed the sixth grade of primary school by gender, poverty status, and distance to secondary school.Table I.d Dropout Rate in 1998 for Children Having Completed 6th Year of Primary SchoolCategoryDropout RateSimple Differencep-value from prtestGirl 36.8%Boy 30.4%Poor 36.9%Non-poor 27.8%> 1km to secondary school 15.3%<= 1km to secondary school 40.5%0.00079.2%25.2%06.4%0.0145The p-values, obtained by running a two-group test of proportion in Stata, are less than 2% for all three categories, which means that we can reject the null hypotheses that there is no difference in dropout rate according to gender, poverty status, and distance to secondary school. Taking those results together, describe the schooling problem in the poor communities of rural Mexico. Evaluate the coverage and targeting schemes of PROGRESA (i.e., to whom should transfer go in priority and to whom do they go).II. Impact analysis on continuation rate and performance based on random assignment of the programThis section compares the continuation rate in 1998 for eligible children (i.e. from poor families) in the treatment and control communities. The simple difference between the treatment and control populations is an indicator of the average impact of the program. Table II.a shows the III. Assessing the quality of randomizationVariable Treatment Control p-valueSecondary school in village 23.2% 23.1% 0.8454Doesn't speak Spanish 2.6% 4.2% 0Worked in 1997 14.2% 11.7% 0.0001Indigenous 39.3% 40.3% 0.2886Education of household head 2.6 2.5 0.5684Household size 7.6 7.5 0.8124Dependency ratio: # of dependables to workers 4.254488 4.309737 0.2499Monthly per capita consumption 112.1 103.2 0Total land owned 1.5 1.5 0.6017Running water in house 5.7% 4.6% 0.0099Bathroom 57.9% 60.1% 0.0219Electrified house 66.9% 70.3% 0.0001IV. Robustness check for the continuation rate in secondary schoolTable 4.a. Program impact when controlling for individual covariatesVariableImpact on continuation likelihood Std Err p-valueProgram * 0.138 0.022 < 0.0005Secondary school in village * 0.158 0.020 < 0.0005Doesn't speak Spanish * -0.200 0.091 0.014Worked in 1997 * -0.055 0.043 0.175Indigenous * 0.079 0.021 < 0.0005Education of household head (years) 0.027 0.005 < 0.0005Household size (people) -0.009 0.005 0.063Dependency ratio: # of dependants to workers 0.016 0.004 < 0.0005Monthly per capita consumption in 1998 (pesos) 0.000 0.000 0.289Total land owned by family -0.002 0.004 0.712Running water * 0.063 0.047 0.225Bathroom * 0.060 0.021 0.005Electrified house * 0.105 0.023 < 0.0005The impact of the program when controlling for the presence of a secondary school in thevillage and various individual covariates is shown in Table 4.a. Impact was measured as the continuation likelihood for children who graduated from 6th grade in 1997 and are continuing into secondary school in 1998. Variables that have a statistically significant impact are shown in bold. The impact column for binary variables (denoted with a *), describes the change in likelihood caused by changing the binary value from 0 to 1. The impact column for non-binary variables describes the change in likelihood caused by increasing the given variable by one unit. Results were calculated in stata with the dprobitcommand, which builds a probit regression model then assesses the impact of each variable for an individual with average value of each characteristic.The impact of the program in a
View Full Document
Unlocking...