Lu & Song (2020) showed that a multi-component information-and-communications-technology (ICT) program adopted at elite Chinese universities from the end of 2012 — bundling e-learning platforms and digital courseware, investment in computer labs and campus internet infrastructure, and online instruction — raised final-year students' probability of receiving a job offer and the wage offered, with the gains operating through students' increased use of computers and the internet for job search. Because the program arrived as a bundle, the data contain no variation exposing the components separately; because outcomes are observed only at graduation, nothing is known beyond the first job; and the job-search channel is consistent with several distinct micro-mechanisms that the survey cannot separate. The open problem is to supply causal evidence that decomposes this ICT bundle into its component-specific labor-market returns, identifies the transmitting micro-mechanisms, and establishes whether the first-offer and starting-wage gains persist over graduates' subsequent careers.

Background

Whether computers and the internet improve education has been studied mostly through academic achievement, and the early evidence was discouraging: classroom computers in Israeli schools did not raise, and may have lowered, pupil test scores (Angrist & Lavy 2002), and substituting online delivery for in-person instruction in US colleges reduced student performance and persistence (Bettinger et al. 2017). Whether educational technology pays off in the labor market — the margin that ultimately matters to students — had essentially no causal evidence at the college level.

Lu & Song (2020) supplied the first such evidence. Starting at the end of 2012, China's Ministry of Education promoted a program at a subset of elite universities that bundled three components: e-learning platforms and digital courseware, investment in computer labs and campus internet infrastructure, and technology-aided (online) instruction. The authors use the Chinese College Students Survey (CCSS), a large repeated cross-sectional survey of students at roughly three dozen elite colleges (fifteen of which implemented the program), interviewed in their final year, when CCSS records whether the student has received a job offer and the wage attached to the best offer. Identification is by difference-in-differences — the before-after change in outcomes at implementing colleges relative to the same change at non-implementing colleges — supported by event-study checks for pre-existing trend differences (plotting cohort-by-cohort treatment-control gaps around the rollout) and by Lee bounds, a nonparametric technique that brackets the average treatment effect on wage when wages are observed only for the selected subsample that receives an offer. The program significantly raised both the offer probability and the offered wage; the effect operated through students' increased use of computers and the internet for job search; and effects differed descriptively by gender, paternal education, and academic major.

Three features of the design leave the economics of the question open. First, the program was a bundle: all components arrived together, so the estimates measure the effect of the package, not of infrastructure access, computer-assisted learning, or online instruction separately, and the available data contain no component-specific exposure variation. Second, the documented channel — more job search via computers and the internet — is consistent with several distinct micro-mechanisms (such as more effective search, digital or task-specific skill accumulation valued by employers, signaling of technological competence, expanded networks, or better job-match quality) that first-job survey outcomes cannot separate. Third, CCSS observes students only in the final year, so nothing is known about labor-market outcomes beyond the best first job offer and its starting wage. In addition, the evidence covers elite Chinese colleges only — the estimation sample spans 36 elite colleges across 20 provinces — and whether the effects extend to other institution tiers or settings remains untested.

Problem Statement

The problem is to causally account for the labor-market returns of the multi-component college ICT program studied by Lu & Song (2020), through three linked core questions. (1) Component decomposition — which components of the bundle (e-learning platforms and digital courseware; computer-lab and internet-infrastructure investment; online instruction) causally generate the increases in job-offer probability and offered wage? (2) Mechanism identification — through which micro-mechanisms does the induced increase in computer and internet use raise labor-market outcomes? The source documents the job-search channel using self-reported search and time-use measures that cannot identify the underlying micro-mechanisms; candidate channels — such as more-effective job search, accumulation of digital or task-specific skills valued by employers, employer signaling of technological competence, network expansion, or improved job-match quality — are illustrative research hypotheses, not a closed or exhaustive answer space. (3) Persistence — do the first-offer and starting-wage gains persist, attenuate, or grow over graduates' subsequent employment histories? The question therefore asks for more than a re-estimate of the bundle-level effect already reported by Lu & Song, or a relabeling of correlational mediation as an identified mechanism.

The problem requires joint treatment of (1), (2) and (3): decomposing the bundle into component-specific causal returns, credibly identifying one or more transmitting micro-mechanisms, and establishing whether the gains persist, attenuate, or grow beyond the first job. Because the source data contain neither component-specific exposure variation nor outcomes beyond the first job, and cannot separate the candidate micro-mechanisms, none of these quantities can be recovered from the available data; for each of (1), (2) and (3) the problem therefore requires a design and identifying assumptions under which the component-specific returns, the transmitting micro-mechanisms, and the post-graduation outcomes would be recovered.

References

  1. Yi Lu, Hong Song (2020), "The effect of educational technology on college students' labor market performance", Journal of Population Economics 33(3):1101-1126, DOI 10.1007/s00148-019-00756-3, https://doi.org/10.1007/s00148-019-00756-3
  2. Nicola Bianchi, Yi Lu, Hong Song (2022), "The effect of computer-assisted learning on students' long-term development", Journal of Development Economics 158:102919, DOI 10.1016/j.jdeveco.2022.102919, https://doi.org/10.1016/j.jdeveco.2022.102919
  3. Lihan Jiang, Xueke Xia, Ruonan Chi (2026), "The Effects of ICT-Aided Learning on Non-Farm Employment: Evidence From China", Review of Development Economics (early view), DOI 10.1111/rode.70173, https://doi.org/10.1111/rode.70173
  4. Muthoni Nganga, Aimable Nsabimana, Christine Niyizamwiyitira (2026), "Smart classrooms and education outcomes: Evidence from Rwanda", Economics of Education Review 110:102736, DOI 10.1016/j.econedurev.2025.102736, https://doi.org/10.1016/j.econedurev.2025.102736
  5. M. Niaz Asadullah, Anindita Bhattacharjee (2022), "Digital Divide or Digital Provide? Technology, Time Use, and Learning Loss during COVID-19", Journal of Development Studies 58(10):1934-1957, DOI 10.1080/00220388.2022.2094253, https://doi.org/10.1080/00220388.2022.2094253
  6. Joshua Angrist, Victor Lavy (2002), "New Evidence on Classroom Computers and Pupil Learning", Economic Journal 112(482):735-765, DOI 10.1111/1468-0297.00068, https://doi.org/10.1111/1468-0297.00068
  7. Eric P. Bettinger, Lindsay Fox, Susanna Loeb, Eric S. Taylor (2017), "Virtual Classrooms: How Online College Courses Affect Student Success", American Economic Review 107(9):2855-2875, DOI 10.1257/aer.20151193, https://doi.org/10.1257/aer.20151193