Science, Technology, Engineering, and Math (STEM) jobs are a key contributor to economic growth and national competitiveness. Yet STEM workers are perceived to be in short supply. Deming and Noray show that the "STEM shortage" phenomenon is explained by technological change, which introduces new job tasks and makes old ones obsolete. They find that the initially high economic return to applied STEM degrees declines by more than 50 percent in the first decade of working life. This coincides with a rapid exit of college graduates from STEM occupations. Using detailed job vacancy data, the researchers show that STEM jobs changed especially quickly over the last decade, leading to flatter age-earnings profiles as the skills of older cohorts became obsolete. The findings highlight the importance of technology-specific skills in explaining life-cycle returns to education, and show that STEM jobs are the leading edge of technology diffusion in the labor market.
This paper was distributed as Working Paper 25065, where an updated version may be available.
Improving schools depends largely on attracting and retaining high-quality teachers. Succeeding in this regard, with limited resources, is made straightforward by understanding teacher preferences for compensation and working conditions. Unfortunately, for a variety of reasons, naturally occurring choice data are neither attainable nor informative. Johnston deploys a discrete-choice experiment in a setting where teachers have reason to reveal their preferences. Estimated utility functions suggest that districts overpay in retirement and class size while underpaying in salary and merit rewards under various objectives. If the best teachers have distinct preferences, compensation can be structured to differentially attract and retain them, inducing a separating equilibrium. High value-add teachers have preferences similar to other teachers, but they have stronger preferences for schools offering merit rewards.
Despite a dramatic increase in the prevalence of private education services in the developing world, there is a limited understanding of pricing policies and the behavior of agents in this market. Little is known about the factors influencing demand, utilization conditional on take-up (attendance), and the effectiveness of these services. Berry and Mukherjee explore these issues with the help of a field experiment in Delhi's slums. Using a two-part pricing design, they identify whether willingness-to-pay is associated with higher utilization of these classes, and whether prices have causal impacts on attendance and dropout. Using assigned prices as instruments for take-up of the classes, the researchers estimate the impacts of enrollment in the classes on student test scores. They find strong evidence that higher willingness to pay is associated with higher utilization, but that lower prices reduce dropout. While tuition classes do not impact test scores for the average student, test scores for children belonging to poorer households increase significantly at endline.
Asher, Novosad, and Rafkin propose a method for the partial identification of conditional expectation functions with interval censored conditioning variables. They focus on two out of many potential applications. First, they resolve a known problem in the estimation of mortality as a function of education: because individuals with high school or less are a smaller and thus more negatively selected group over time, estimates of their mortality change are likely to be biased. The researchers method makes it possible to estimate mortality at constant education percentiles, revealing that current estimates of rising mortality for less educated women are biased upward in some cases by a factor of three. The method is also applicable to the estimation of education gradients in patterns of fertility, marriage and disability, among others, where similar compositional problems arise. Second, the researchers apply the method to the estimation of intergenerational mobility, where researchers frequently use coarsely measured education data in the many contexts where matched parent-child income data are unavailable. They show that conventional measures like the rank-rank correlation may be uninformative once interval censoring is taken into account. The researchers propose a new CEF-based measure of mobility that can be bounded tightly. Earlier methods in both mortality and mobility produced results that were either biased or had bounds too wide to be meaningful. The researchers' bounds are applicable in a wide range of contexts, rely on minimal structural assumptions, and generate parameters that are useful for policy.
College Access Field Experiment
Despite well-documented large and growing returns to attending college, youth from socioeconomically disadvantaged backgrounds continue to attend college at lower rates than their more-advantaged peers. This gap is not fully explained by differences in academic preparation, and college enrollment and completion appear to have become more dependent on family income in recent decades. Phillips and Reber reports the results of a randomized field trial of the effects of two variants of V-SOURCE, a college-counseling intervention designed to address informational and social support barriers to college enrollment, as well as students' tendency to forget or procrastinate about deadlines. V-SOURCE served students from the spring of the junior year through the summer after high school graduation. The program was delivered "virtually" via the internet, phone, email, text message, and social networking platforms, making it relatively low-cost and scalable. The researchers find that students took advantage of V-SOURCE services and found them helpful, and that the program had modest effects on the college application process, including the application portfolio. The program did not improve college enrollment outcomes on average, though it may have improved outcomes for Hispanic students who spoke Spanish at home. While some low-touch interventions have improved college outcomes in other studies, these effects may be dependent on the population served and context of the study.
Lavecchia, Oreopoulos, and Brown estimate long-run impacts to the Pathways to Education program, an intensive set
of coaching, tutoring, and group activities, offered to disadvantaged high school students
beginning in Grade 9, along with short- and long-term financial incentives for participation. Using
a gradual roll-out process and eligibility requirements based on residing in public housing, the researchers
compare similar eligible and ineligible students before and after the program's introduction. High
school administrative records are matched to income tax records to follow individuals up to the
age of 26, even when they leave the household or province. Lavecchia, Oreopoulos, and Brown find significant positive effects
on persistence in postsecondary education institutions, earnings and employment. Eligibility for
Pathways increased annual earnings by 9 percent and employment by 6 percentage points (10
percent) and reduced the likelihood of ever receiving social assistance (welfare) by more than a
third. These gains suggest long-term benefits - even from government net revenue changes alone,
can justify the high program's high cost.
One hypothesis regarding financial aid is that increased generosity in aid programs gives incentives for universities, particularly for-profit colleges, to increase tuition rates. However, this causal effect is difficult to estimate because identification requires exogenous shocks to a program's generosity. The Post 9/11 GI Bill represented one of the largest expansions of college benefits for veterans and their dependents. In the first version of the bill, the Department of Veterans Affairs offered full funding for public colleges and set the maximum tuition reimbursement for private institutions on a state level. However, in 2010, the VA set the maximum tuition benefit for private schools to one nationwide amount, while continuing to fully fund attendance at public colleges. Baird, Kofoed, Miller, and Wenger use a difference in differences estimator and find that in states where the benefit increased, for-profit universities increased their sticker price tuition by $461. The researchers also find that for-profit institutions in states where the VA cut benefits lowered tuition significantly, exhibiting symmetric behavior with pass through rates of around 5.6 percent. However, for-profit institutions did not increase admissions or enrollments because of the policy change.
and Educational Segregation
Standard psychometric methods aggregate test items into achievement scores without considering how items relate to economic outcomes. Nielsen constructs alternative achievement measures using the estimated relationship between individual test items and economic outcomes such as school completion and labor market earnings. Item-anchored scales rank students differently than standard psychometric scales, and these ranking differences have important implications for estimated achievement gaps by race, gender, and parental income. Typically, though not always, item-anchored achievement gaps are substantially larger than gaps calculated using standard psychometric scores. Black/white and high-/low-income achievement gaps are generally about 0.2-0.6 standard deviations larger (about 20-60%) when test items are related to labor market outcomes and 0.06-0.2 standard deviations larger (6-20%) when high school and college completion are used. Test items can fully explain blackwhite earnings differences, but can explain only half of the earnings difference between youth from high- versus low-income households. Finally, conditional on item-anchored math scores, item-anchored reading scores have significantly positive relationships with various economic outcomes in contrast to psychometric scales where the reading relationships are zero or even negative conditional on math.
Low-income students are unlikely to attend a highly selective college even when they have strong academic credentials. Changing this has the potential to increase college graduation and reduce income inequality. With a field experiment , we test an intervention intended to increase the enrollment of low-income students at the highly selective University of Michigan. Susan Dynarski, Carmello Libassi, Katherine Michelmore, and Stephanie Owen contact students (as well as their parents and principals) with an encouragement to apply along with a promise of four years of free tuition and fees upon admission. Materials emphasize that this offer is not contingent on completing aid applications (e.g., the FAFSA or PROFILE). The intervention produced large effects on application, enrollment, and persistence. Treated students were more than twice as likely to apply to (67 percent vs. 26 percent) and enroll at (27 percent vs. 12 percent) University of Michigan. There was no diversion from schools as (or more) selective as UM; the enrollment effect of 15 percentage points (pp) is comprised of students who would otherwise attend a less selective four-year college (7 pp), a community college (4 pp), or no college at all (4 pp). Nearly all admitted students eventually completed aid applications and qualified for more than the promised tuition waiver. Enrollment effects persist through two years of follow-up. The intervention closed by half income gaps in college attendance and selectivity among Michigan's high-achieving students. The researchers conclude that an encouragement to apply, paired with an upfront promise of aid, when communicated to both students and influential adults, can substantially reduce socioeconomic differences in college choices.
Barrera-Osorio, de Barros, and Filmer report on a randomized experiment to investigate the longterm effects of a primary school scholarship program in rural Cambodia. In 2008, fourth-grade students in 207 randomly assigned schools (103 treatment, 104 control) received scholarships based on the student's academic performance in math and language or on their level of poverty. Three years after the program's inception, an evaluation showed that both types of scholarship recipients had more schooling than non-recipients, however, only merit-based scholarships led to improvements in cognitive skills. This new study reports impacts, nine years after program inception, on the educational attainment, cognitive skills, socioemotional outcomes, socio-economic status and well-being, and labor market outcomes, of individuals who are, on average, 21 years old. The results show that both types of scholarships led to higher long-term educational attainment (about 0.21-0.29 grade levels), but only merit-based scholarships led to improvements in cognitive skills (0.11 standard deviations), greater self-reported well-being (0.18 standard deviations), and employment probability (3.4 percentage points). Neither type of scholarship increased socioemotional skills. The results also suggest that there are labeling effects: the impacts of the scholarship types differ even for individuals with similar characteristics.