Abstract: I study whether an AI-Tutor embedded in the Macmillan Achieve homework platform improves learning outcomes for honors macroeconomic principles students. The setting provides assignment-level telemetry on 32 students across 11 homework assignments, linked to exams, course grades, and surveys. AI use was not associated with better grades. In the baseline fixed-effects model, each additional AI hint per question is associated with 0.683 more attempts per question, but not with higher homework scores, exam scores, or final course grades. Adoption was swift but short-lived: 87.1% of students requested at least one hint on the first homework, compared with 37.5% on the eleventh. Students using more hints tended to start homework earlier, but the extra activity did not translate into measured learning gains. Survey responses fit the same pattern. Nearly all surveyed students tried the tool (96.0%), but only 34.8% reported being satisfied with its performance. In this implementation, the AI-Tutor engaged students without improving learning outcomes, though the sample’s high baseline achievement may limit detection of small performance gains.
Abstract: In a recession, increased competition forces inexperienced job market entrants to accept lower wages than those who start their careers during an economic boom. Despite years of improvement in labor market conditions following a recession, a wage disparity, known as scarring, persists between these cohorts. Recently implemented Salary History Ban laws (SHBs) are intended to reduce wage disparities between advantaged and disadvantaged groups. In this study, I test how these laws affect a unique and often less salient disadvantaged group – scarred workers. For scarred workers who began their careers during a moderate-to-severe recession, or a five percentage point higher state unemployment rate, I find SHBs increase job mobility by 0.6%, hourly wages by 2.65%, and weekly earnings by 5% relative to cohorts who graduated in baseline labor market conditions. These estimates represent a substantial reduction in the original scarring effect and provide a broader understanding of the mechanisms behind both scarring and SHB laws.
Abstract: Are there long-term labor consequences in migrating to the US during a recession? For most immigrants, credibly estimating this effect is difficult because of selective migration. Some immigrants may not move if economic conditions are not favorable. However, identification is possible for refugees as their arrival dates are exogenously determined through the US Refugee Resettlement program. A one percentage point increase in the arrival national unemployment rate reduces refugee wages by 1.98% and employment probability by 1.57 percentage points after 5 years.
Abstract: The literature has traditionally focused on the local unemployment rate to estimate how initial economic conditions affect long-run outcomes. Using Job Openings and Labor Turnover Survey, or JOLTS, State Estimates for job openings, hires, and separations along with Local Area Unemployment Statistics, I test how changes in other aggregate measures of labor market activity affect long run outcomes. I find that for every one point increase in the local unemployed-to-job-opening ratio, annual earnings are reduced by 4.45% and remain depressed for over 13 years. Conversely, I find that a one percentage point increase in the local job openings rate or a one point increase in the local vacancy/unemployment ratio, increases initial annual earnings by 8.18% and 17.16%, respectively, which persists for nearly 11 years. I similarly find a positive and persistent effect on annual earnings from other measures of labor market tightness like the job-to-job transition rate, quits rate, job finding rate, and the labor-leverage ratio.