Education Program Meets
November 10-11, 2011
Karthik Muralidharan, University of California at San Diego and NBER
Muralidharan presents results from a five-year randomized evaluation of group and individual teacher performance-pay programs implemented across a large representative sample of government-run rural primary schools in the Indian state of Andhra Pradesh. He finds consistently positive and significant impacts of the individual teacher incentive program on student learning outcomes across all durations of program exposure. Students who completed their full five years of primary school under the program performed significantly better than those in control schools, by 0.54 and 0.35 standard deviations in math and language tests respectively. These students also scored 0.52 and 0.3 standard deviations higher in science and social studies tests, even though there were no incentives in these subjects. The group teacher incentive program also had positive (and mostly significant) effects on student test scores, but the effects were always smaller than those of the individual incentive program, and were not significant at the end of primary school for the cohort exposed to the program for five years.
Debopam Bhattacharya, Shin Kanaya, and Margaret Stevens, University of Oxford
Selective colleges are often accused of unfair admission practices which favor specific socioeconomic groups. Bhattacharya, Kanaya, and Stevens develop an empirical framework for testing if current admissions are academically fair in that they maximize expected college performance of the entrants. Such maximization would imply that the expected performance of the marginal admitted candidate - the admission threshold- is equalized across socioeconomic groups. The authors show that the inter-group threshold differences are identified if 1) the unobserved heterogeneity affecting admission decisions is median-independent, but not necessarily scale-independent, of applicant covariates; and 2) the density of the conditional expected performance is positive around the admission threshold for each demographic group. They develop the theory of large-sample inference for the identified thresholds. Lastly, they consider identification of the short-run academic productivity loss from the use of different thresholds, as would happen under affirmative action policies. They apply these methods to three recent cohorts of UK-based applications to Oxford's flagship PPE degree focusing on first-year results as the outcome of interest. They find that applicants who are male or from private schools face expected performance thresholds that are about 2 percentage points higher; the implied productivity loss is about 1.5 percentage points, per applicant.
Robert Fairlie, University of California at Santa Cruz; Florian Hoffmann, University of British Columbia; and Philip Oreopoulos, University of Toronto and NBER
Fairlie, Hoffmann, and Oreopoulos use detailed administrative data from one of the largest community colleges in the United States to quantify the extent to which academic performance depends on students being of similar race or ethnicity to their instructors. To address the concern of endogenous sorting, they use both student and classroom fixed effects and focus on those with limited course enrollment options. They also compare sensitivity in the results from using within versus across section instructor type variation. Given the computational complexity of the 2-way fixed effects model with a large set of fixed effects, they rely on numerical algorithms that exploit the particular structure of the model's normal equations. They find that the performance gap in terms of class dropout and pass rates between white and minority students falls by roughly half when taught by a minority instructor. In models that allow for a full set of ethnic and racial interactions between students and instructors, they find African-American students perform particularly better when taught by African-American instructors.
Sa Bui and Steven G. Craig, University of Houston, and Scott A. Imberman, University of Houston and NBER
Bui, Craig, and Imberman identify the impact of gifted and talented services on student outcomes by exploiting a discontinuity in eligibility requirements. They find no impact on standardized test scores of marginal students, even though peers and classes improve substantially. They then use randomized lotteries to examine the impact of attending a GT magnet program relative to programs in other schools. They find that, despite exposure to higher quality teachers and peers, only science achievement improves. They further find that the relative ranking of students change, as do their grades, indicating that either invidious comparison peer effects or teaching targeting may be important.
Rajashri Chakrabarti, Federal Reserve Bank of New York
No Child Left Behind (NCLB) mandated the institution of Adequate Yearly Progress (AYP) objectives, and schools are assigned an AYP pass/fail based on performance in these objectives. AYP fail status is associated with negative publicity and often sanctions. Chakrabarti studies the incentives and responses of schools that failed AYP once. Using data from Wisconsin and regression discontinuity designs, she finds improvements in high stakes reading and spillover effects to low stakes language arts in these schools. The patterns are consistent with a focus on marginal students around the high stakes cutoff, but this did not come at the expense of the ends of the distribution. There is not much evidence in favor of improvement in high stakes math, or in low stakes science and social studies. Performance in low stakes grades suffered, as did performance in weaker subgroups, in spite of their inclusion in AYP computations. While there is no robust evidence in favor of effects in test participation and graduation, attendance improved in threatened schools where they mattered for AYP. There is strong evidence in favor of response to incentives-schools that failed AYP by failing only in reading and/or math did substantially better in these subject areas. The credibility of the threat mattered: AYP-failed schools that faced more competition responded both more strongly and more broadly, with robust evidence in favor of improvements in all AYP objectives.
Fernanda Brollo, Eliana La Ferrara, and Katja Kaufmann, Bocconi University
The effectiveness of conditional welfare programs crucially depends on their design, such as the exact rules, benefit amounts, and structure. Another important factor is the quality of enforcement of those rules. While the former relationship has been studied, the importance of the quality of enforcement for program effectiveness has rarely been addressed. Brollo, La Ferrara, and Kaufmann study the implementation of a large-scale conditional cash transfer program, "Bolsa Familia" in Brazil, which conditions transfers to poor families on children's school attendance. They analyze how people learn about the quality of enforcement and how this affects their behavior. They find that individuals respond to incentives and fine tune their behavior in response to signals about the quality of enforcement of program conditions. They learn both from private signals, observing the consequences of own noncompliance, and from public signals, that is observing the consequences from peers' noncompliance. Thus enforcement has not only a direct effect on the family affected but also an important multiplier effect on other families, who learn from the experiences of their children's peers.
Joshua Goodman, Harvard University
Prior biological research has shown that left- and right-handed individuals have differing brain structures, particularly in relation to language processing. Goodman argues that left-handedness can be considered a proxy for differential neural wiring generated in part by poor infant health. Using five data sets from the United States and the United Kingdom, he shows that low birth weight and complicated labors increase the likelihood of a child being left-handed. Even conditional on observable measures of infant health and family background, lefties exhibit economically and statistically significant human capital deficits relative to righties. Compared to righties, lefties score a tenth of a standard deviation lower on measures of cognitive skill, are more likely to have emotional or behavioral problems, and are more likely to have learning disabilities such as dyslexia. Differences between left- and right-handed siblings are similar or larger in magnitude. Lefties have lower educational attainment and have hourly wages that are 6 percent lower than righties. Parents and schools could use this easily observable characteristic to identify a subgroup of children whose cognitive and behavioral development may warrant additional attention.
Todd R. Stinebrickner, University of Western Ontario and NBER, and Ralph Stinebrickner, Berea College
Stinebrickner and Stinebrickner estimate a simple structural learning model of the college drop-out decision which takes advantage of unique data to reduce the assumptions that are traditionally necessary for identification. They find that approximately 45 percent of the drop-out that occurs before the start of the third year can be attributed to student learning about academic performance. They also find that students who perform poorly tend to learn that staying in school is not beneficial; they do not leave simply because they have lost the option to stay, or believe that they are more likely to lose the option in the future. As to why students find that staying in school is no longer beneficial, the most important avenue is that performing poorly reduces how enjoyable it is to be in school. However, the reduction in the financial returns to graduating that accompanies poor performance would also be sufficient to create substantial drop-out.
Stephanie Riegg Cellini, George Washington University, and Claudia Goldin, Harvard University and NBER
For-profit higher education is among the most rapidly growing sectors in the economy. There is a strong presumption that the size and growth of its institutions are largely due to the implicit subsidy received from federal student aid programs under Title IV. But not all for-profit institutions are eligible for Title IV aid. Because only Title IV eligible institutions are captured in the official data, those that are not eligible are not counted. Cellini and Goldin use administrative data from five states to provide the first estimates of the total for-profit higher education sector. They find that the actual number of for-profit institutions is double the official count, and that the actual number of students is between one-quarter and one-third greater. Many for profit institutions that are not Title IV eligible offer programs and certificates that are similar, if not identical, to those given by institutions that are part of Title IV. The researchers find that Title IV institutions charge tuition that is about 55 log points higher than that charged by comparable institutions whose students cannot apply for federal financial aid. The dollar value of the premium is about equal to the amount of financial aid received by students in eligible institutions, lending credence to the Bennett hypothesis that aid-eligible institutions raise tuition to maximize aid.
C. Kirabo Jackson, Northwestern University and NBER
Existing studies on single-sex schooling suffer from biases because students who attend single-sex schools differ in unmeasured ways from those who do not. In Trinidad and Tobago, students are assigned to secondary schools based on an algorithm allowing one to address self-selection bias and estimate the causal effect of attending a single-sex school versus a similar coeducational school. Jackson finds that while students (particularly females) with strong expressed preferences for single-sex schools benefit, most students perform no better at single-sex schools. Girls at single-sex schools take fewer sciences courses and more traditionally female subjects.