By Tamra Thetford, VP of Impact Evaluation, CNote
For most of its history, community finance operated without much academic research behind it. That is beginning to change. In 2024, Julie Birkenmaier and Jin Huang of Saint Louis University published a systematic review of 171 studies on financial access, arguing that holding an account and having meaningful access are not the same thing. This piece is about what happened when we tried to translate that research into an actionable approach for measuring financial access, using the data we can actually collect.
Academic Research Can Test and Strengthen Impact Practice
Most impact metrics in this space, including CNote’s until recently, were built from available data and reasonable judgment, not from academic research findings. That is not a criticism. It describes a field that grew out of practice before the research caught up to it. The two are now close enough to talk to each other. Citing the paper in a footnote is not the same as putting it into practice which requires taking a research construct, figuring out what it would look like in the data you can actually collect, applying it, and being honest about what breaks in translation. This year, we set out to do exactly that with our own Impact Cash data.
Financial Access Is More Than Having an Account
The central argument in Birkenmaier and Huang’s research is one that community finance practitioners will recognize immediately: financial access is not the same thing as account ownership. Someone can hold a checking account and still lack meaningful access if they are treated unfairly, if the branch environment is unwelcoming, or if the products available to them carry barriers they cannot clear.
Their work organized these dimensions into what they call the Financial Access Scale, which includes a measure of institutional practices, how financial service providers actually treat the people they serve. Subsequent papers from the same team validated that structure empirically and linked financial services mistreatment to measurable effects on financial well-being.
CNote Turned Research Into Measurable Indicators
The Financial Access Scale measures individual perception. It asks people how they were treated and what they experienced. CNote does not have that relationship: our Impact Cash® program places deposits with mission-driven banks and credit unions, and we collect data from those institutions, not from their account holders.
So we had a choice: cite the research and keep measuring what we had always measured, or attempt a translation and accept its imperfections. We took the second path, mapping the research constructs onto things we can actually observe. The result is three categories, each scored relative to the portfolio:
1. Wealth building opportunity access. Lending to underserved communities, supported by CDFI, MDI, and LID designations.
2. Inclusive and responsive service environment. Branch presence in LMI and majority-minority census tracts, plus whether the institution has structured ways of hearing from its community.
3. Reduction of structural and operational barriers. ChexSystems flexibility, underwriting that does not require a credit score, and multiple acceptable types of documentation.
We applied this to the mission-driven banks and credit unions in the Impact Cash® portfolio.
Lending Alone Does Not Show the Full Picture of Financial Access
What follows is a first application of this approach. We’re reporting these results in aggregate, not by individual institution or exact count, to avoid disclosing the performance of specific partner institutions.
- The service environment scored highest across the portfolio. Roughly two-thirds of institutions reporting results in 2025 rated High, driven largely by branches physically located in the communities they serve and by structured input channels such as advisory boards, member surveys, and frontline staff feedback loops.
- Barrier reduction scored lowest and split into extremes. Scores clustered at the ends rather than the middle, with the High and Low groups each holding close to 40 percent of the portfolio. That pattern suggests institutions adopt barrier-reducing practices as a package, rather than adjusting them incrementally, which points to a cultural and strategic decision more than a product one.
- Lending performance and barrier reduction did not consistently move together. A meaningful subset of institutions scored High on lending and Low on barrier reduction, meaning strong capital deployment did not guarantee that account-opening practices had kept pace with it. We think this is the pattern most worth watching, though we are not ready to generalize it beyond this portfolio.
- CDFI and MDI designation correlated with lending outcomes, but some of the highest scores in the portfolio came from institutions with no formal designation at all.
Institutional Data Cannot Replace Customer Experience
The underlying research measures how people experience their financial institutions, including whether they felt fairly treated and whether a branch felt welcoming to them. We are measuring branch location and whether an institution has a structured way of gathering member input. Those seem like reasonable stand-ins, but we don’t yet know how closely they track what account holders actually experience, and that remains an open question rather than one we’ve answered. Birkenmaier and Huang were generous enough to review our application of their research and share their perspective on it, which shaped where we landed.
Better Data and Benchmarks Will Strengthen the Approach
On this measurement approach itself, we are adding indicators for 2026, including low-fee account availability and language access, and we are evaluating whether portfolio-relative scoring should give way to external benchmarks, so an institution’s score means something outside our own portfolio, not only relative to its peers here. Turning academic research into usable metrics is not a one-time exercise, we expect to keep revising this approach as we learn more, both from our own data and from anyone else working through the same translation problem. If you’re doing something similar, we’d like to compare notes.
