Moving Attention but Not Enrollment: Nudging College Information Search at National Scale in Colombia

Students often underuse available information when making high-stakes higher education choices, and a planner rarely knows which option any student should choose—which makes nudging search the natural instrument. We randomized informational pop-up messages across 519,000 Colombian students—the full fall national cohort (502,494) and the smaller spring cohort (16,861)—as they logged in to view mandatory exit-exam results, linking each message to a purpose-built portal centralizing earnings, costs, and financial aid. The pop-ups moved attention at scale: 15–25% of treated students clicked through. But message design defied prediction: every belief-targeting variant—earnings levels, earnings ranges, personalized figures, gain or loss frames—underperformed the plain invitation to search by 4–8 percent-age points, and loss framing, predicted by prospect theory to dominate, performed worst (13.1% versus 16.7% for gain framing). Enrollment did not move: we rule out average effects above 2.6 percentage points under the conservative inference we report, and above roughly 0.6 points under design-based inference. Students who clicked genuinely searched—nearly half opened the portal’s earnings tables or aid pages—yet generic information, even when consulted, did not change what they did. Digital platforms make experimentation cheap; our results show why it remains necessary.