![[ Visit Client Website ]](images/banner.gif)
CollgeBound Boost is a three-arm randomized controlled trial (RCT) embedded in CollegeBound Saint Paul, a citywide initiative that automatically opens CDA accounts with an initial seed deposit for newborns and young children. Because CollegeBound Boost’s eligibility requirements target families who live at or below 300 percent of the federal poverty level (FPL) guideline, the resulting sample included families with higher wealth than most Guaranteed Income programs, which typically set eligibility requirements at 100 to 200 percent of the FPL. This experimental study examines the treatment effects of quarterly supplemental deposits and guaranteed income on college savings behavior, with particular attention to wealth threshold effects.
Methods
In CollegeBound Boost, all participating families receive a seeded CDA and are randomly assigned to one of three arms: (1) CDA only (control group, n = 274); (2) CDA plus quarterly Boost deposits that mimic a small baby bond contribution over 12 months (QD group, n = 290); or (3) CDA plus quarterly Boost deposit for 12 months and a monthly guaranteed income payment over 24 months (GI group, n = 324). Baseline surveys and administrative data were collected before random assignment, with two follow up surveys at 12 and 24 months. This design allows us to make causal inferences regarding intervention impacts on college savings behavior. Blocked fixed-effects models with clustered standard error models are employed to estimate the heterogeneous effects of treatment conditions.
Results
The treatment effect varied markedly by household wealth. Among families with assets above $35,000, 67.1% of GI group participants reported saving for college, compared to 55.4% in the control group (weighted analysis; p = .0135). By contrast, there was no statistically significant difference among lower‑wealth families (assets below $35,000), where college saving rates were nearly identical across the GI group (62.8%), QD group (60.4%), and control group (58.2%; p=.6314). Heterogeneous treatment‑effect models confirm that this is not merely a descriptive comparison. The interaction between GI group assignment and higher-wealth status was statistically significant (p = .009) in the weighted model, confirming that wealthier families responded to the intervention at a substantially greater rate.
Implications
These findings have implications for the design of guaranteed income programs serving low-income families with children enrolled in CDAs. Specifically, a $500 monthly GI payment combined with a $1,000 asset deposit appears most effective at promoting college savings among families at the upper end of the low-income bracket (up to 300% FPL) who hold approximately $35,000 in assets, while producing no statistically significant effect among lower-wealth families within the same income range. This suggests that relatively modest GI transfer amounts may be insufficient to shift savings behavior among the lower-wealth families, and that program design should consider wealth-sensitive parameters – such as calibrating payment size to baseline family wealth or setting an asset requirement – to maximize impact across wealth distribution.