Members of the NBER's Education Program met April 11-12 in Stanford. Program Director Caroline M. Hoxby of Stanford University organized the meeting . These researchers' papers were presented and discussed:
Andrew Morgan, Minh Nguyen, and Ben Ost, University of Illinois at Chicago; Eric A. Hanushek, Stanford University and NBER; and Steven G. Rivkin, University of Illinois at Chicago and NBER
Getting Effective Educators in Hard-to-Staff Schools
C. Kirabo Jackson and Laia Navarro-Sola, Northwestern University, and Diether Beuermann and Francisco Pardo, Inter-American Development Bank
What is a Good School, and Can Parents Tell? Evidence on the Multidimensionality of School Output (NBER Working Paper No. 25432)
Is a school's impact on high-stakes test scores a good measure of its overall impact on students? Do parents value school impacts on high-stakes tests, longer-run outcomes, or both? To answer the first question, Jackson, Beuermann, Navarro-Sola, and Pardo apply quasi-experimental methods to data from Trinidad and Tobago and estimate the causal impacts of individual schools on several outcomes. Schools' impacts on high-stakes tests are weakly related to impacts on low-stakes tests, dropout, crime, teen motherhood, and formal labor market participation. To answer the second question, the researchers link estimated school impacts to parents' ranked lists of schools and employ discrete choice models to estimate parental preferences. Parents value schools that causally improve high-stakes test scores conditional on average outcomes, proximity, and peer quality. Consistent with parents valuing the multidimensional output of schools, parents of high-achieving girls prefer schools that increase formal labor market participation, and parents of high-achieving boys prefer schools that reduce arrests.
Sarah Cohodes, Columbia University and NBER; Elizabeth Setren, Tufts University; and Christopher R. Walters, University of California, Berkeley and NBER
Can Successful Schools Replicate? Scaling Up Boston's Charter School Sector
Can schools that boost student outcomes reproduce their success at new campuses? Cohodes, Setren, and Walters study a policy reform that allowed effective charter schools in Boston, Massachusetts to replicate their school models at new locations. Estimates based on randomized admission lotteries show that replication charter schools generate large achievement gains on par with those produced by their parent campuses. The average effectiveness of Boston's charter middle school sector increased after the reform despite a doubling of charter market share. An exploration of mechanisms shows that Boston charter schools reduce the returns to teacher experience and compress the distribution of teacher effectiveness, suggesting the highly standardized practices in place at charter schools may facilitate replicability.
Hessel Oosterbeek, Sandor Sovago, and Bas van der Klaauw, VU Amsterdam
Why are Schools Segregated? Evidence from the Secondary-School Match in Amsterdam
Oosterbeek, Sovago, and van der Klaauw use rich data from the secondary-school match in Amsterdam to nonparametrically decompose school segregation by ethnicity and by household income into five additive sources: i) ability tracking, ii) noise, iii) residential segregation, iv) preference heterogeneity, and v) capacity constraints. Important features of the Amsterdam school district are its diverse population, that students can freely choose any school at their ability level, that school density is high and that private schools are absent. The researchers find that school segregation is mainly driven by ability tracking and students from different groups having different preferences. Residential segregation, capacity constraints and noise play only a minor role. Policy simulations indicate that even hard quota reduce segregation by only a modest amount, while it is costly in terms of student welfare.
Jeffrey T. Denning, Eric R. Eide, and Merrill Warnick, Brigham Young University
Why Have College Completion Rates Increased?
In influential work, Bound, Lovenheim, and Turner (2010) showed that college completion rates declined from the 1970s to the 1990s and the majority of the decline was due to "supply side" reasons. Denning, Eide, and Warnick document that this trend has reversed -- since the 1990s, college completion rates have increased. Next, they focus on the reasons for the increase in college graduation rates. Collectively, student characteristics, institutional resources, and institution attended do not explain much of the change. However, the researchers find support that standards for degree receipt may explain the change in graduation rates.
Ulf Zoelitz, University of Zurich, and Ingo E. Isphording, IZA - Bonn
The Value of a Peer - A New Way to Quantify Individual Spillovers
Zoelitz and Isphording introduce a new approach to peer effects: Peer value-added isolates the total contribution of an individual to the performance of others without relying on observable peer characteristics as measures of peer quality. Using data from a setting with repeated random assignment of students to peer groups the researchers show that there is significant variation in peers' value-added. Peer observable characteristics, most notably previous performance, are poor predictors of individual spillovers. The researchers validate their peer value-added measures in out-of-sample social interactions and show that peer value-added captures performance and earnings spillovers among randomly re-assigned peers. They establish that the ability to raise others' performance is a malleable trait. Students interacting with peers generating positive spillovers become more valuable peers themselves.
Mauricio Romero, ITAM; Justin Sandefur, Center for Global Development; and Wayne A. Sandholtz, University of California, San Diego
Outsourcing Service Delivery in a Fragile State: Experimental Evidence from Liberia
In 2016, the Liberian government delegated management of 93 randomly-selected public schools to private providers. Providers received USD 50 per pupil, on top of USD 50 per pupil annual expenditure in control schools. After one academic year, students in outsourced schools scored 0.18σ higher in English and mathematics. Teachers in outsourced schools were 50% more likely to be in school and 43% more likely to be engaged in instruction during class time. Romero, Sandefur, and Sandholtz do not find evidence of heterogeneity in learning gains or enrollment across students. They do find significant heterogeneity across providers, both in learning gains and other unintended consequences of outsourcing.
Ying Shi, Stanford University, and John D. Singleton, University of Rochester
Expertise and Independence on Governing Boards: Evidence from School Districts
Shi and Singleton study the roles of expertise and independence on governing boards in the context of education. In particular, they examine the causal influence of professional educators elected to local school boards on education production. Educators may bring valuable human capital to school district leadership, thereby improving student learning. Alternatively, the independence of educators may be distorted by interest groups. The key empirical challenge is that school board composition is endogenously determined through the electoral process. To overcome this, the researchers develop and implement a novel research design that exploits California's randomized assignment of the order that candidates appear on election ballots. The insight of the empirical strategy is that ballot order effects generate quasi-random variation in the elected school board's composition. The approach is made possible by a unique dataset that combines election information about California school board candidates with district-level data on education inputs and outcomes. The results reveal that educators on the school board causally increase teacher salaries and reduce district enrollment in charter schools relative to other board members. The researchers do not find accompanying effects on student test scores. They interpret these findings as consistent with educators on school boards shifting bargaining in favor of teachers' unions.
Tarek Azzam, Claremont Graduate university, and Michael D. Bates and David Fairris, University of California, Riverside
Do Learning Communities Increase First Year College Retention?
Testing the External Validity of Randomized Control Trials
Azzam, Bates, and Fairris (1) utilize a randomized control trial (RCT) to estimate the impact of a learning community on first year college retention, (2) introduce simple tests for the external validity of RCT results, and apply these tests to the data, and (3) compare observational estimates to those from the RCT, considering the internal and external validity of both approaches. Intent-to-treat and local average treatment estimates reveal no discernable programmatic effects, whereas observational estimates are positive and statistically significant. The experimental sample, though negatively selected on observed characteristics, is positively selected on unobserved characteristics implying limited external validity of the RCT.
James Berry, University of Delaware; Rebecca Dizon-Ross, University of Chicago and NBER; and Maulik Jagnani, Cornell University
(Not) Playing Favorites: An Experiment on Parental Preferences for Educational Investment
How do parents choose to allocate investments across children? Do they maximize the returns to their investments (total household earnings), or equalize across their children because of an aversion to cross-sibling inequality? Berry, Dizon-Ross, and Jagnani conduct the first experiment that identifies parents' preferences for investing in their children's education. The experiment exogenously varies the short-run returns to educational investments to identify the degree to which parents care about (a) maximizing total household earnings, (b) minimizing cross-sibling inequality in "outcomes" (i.e., child-level earnings), and (c) minimizing cross-sibling inequality in "inputs" (i.e., the investments each child receives). The researchers find that parents care about both maximizing total household earnings and minimizing inequality in inputs. Parents place a high value on equality of inputs, choosing exactly equal inputs 35% of the time and foregoing roughly 40-50% of their potential experimental earnings in order to equalize inputs. By contrast, the researchers find no evidence that parents are averse to inequality in outcomes.
Krzysztof Karbownik, Northwestern University, and Umut Özek, American Institutes for Research
Setting a Good Example? Examining Sibling Spillovers in Educational Achievement Using Regression Discontinuity Design.
Karbownik and Özek identify externalities in human capital production function arising from sibling spillovers. Using regression discontinuity design generated by school-entry cutoffs and school records from one district in Florida, they find positive spillover effects from an older to a younger child in less affluent families and negative spillover effects from a younger to an older child in more affluent families. The results are consistent with direct spillovers dominating in economically disadvantaged families and with parental reinforcement in more affluent families.
Kendall J. Kennedy, Mississippi State University
Hidden Schooling: Repeated Grades and the Returns to Education and Experience
Over the past four decades, nearly 25% of all American public school students repeated at least one grade in primary or secondary school, and ninth grade repeating increased four-fold. Despite its prevalence, few economists have attempted to account for grade repeating when estimating the returns to education and experience. Kennedy shows that 10% of the increase in ninth grade repeating was caused by changes in compulsory schooling laws (CSLs). Because CSLs increase both grade repeating and educational attainment, compulsory education-based IV estimates of the returns to education are positively biased by up to 38%. Additionally, grade repeating causes endogenous proxy error in labor market experience. Through this error, Kennedy shows that the residual black-white wage gap is overstated by 10%, the wage return to a high school diploma is overstated by 11% relative to dropouts, and the labor supply gap between dropouts and high school graduates is overstated by 23%. Controlling for age instead of experience reduces this bias, suggesting age should be a standard control variable for reduced-form analysis, not experience.