8  Interpretation

The aggregate rural leverage gap is almost entirely a between-CDE selection phenomenon. This chapter spells out what that means substantively, what it does and does not say about NMTC and rural markets, and what the policy implications are.

8.1 What the within-CDE collapse means

The 80% reduction in the rural coefficient from M0 to M4 — from −0.262 to a statistically insignificant −0.047 — is the empirical content of the paper. To unpack it:

  • The between-CDE component (about 80% of the raw gap) reflects the fact that rural-specialist CDEs (Rural Development Partners LLC at 80% non-metro share, Montana Community Development Corporation at 70%) deploy capital that ends up at lower leverage than urban- specialist CDEs (ESIC New Markets Partners, Consortium America LLC, and Capital Impact Partners, all at 0% non-metro). This difference is not about where they deploy — it is about who they are as organizations.
  • The within-CDE component (the remaining ~20%, statistically indistinguishable from zero) reflects what happens when a single CDE takes its existing operational capacity to a rural deal versus an urban one. The same CDE achieves essentially the same leverage in rural and urban tracts.

This is a much sharper finding than “rural deals leverage less.” It relocates the empirical action from the rural market itself to the intermediary structure that selects into rural markets.

8.2 What this does NOT say

Three things the paper is careful not to claim:

  1. It does not say rural markets are not different. Rural credit markets are well-documented to differ from urban ones in capital availability, lender networks, deal-flow density, and information structure (Backman et al. 2022). What the paper says is that those differences do not translate into a within-CDE leverage penalty when intermediary identity is held constant. The composition of CDEs operating rurally is what produces the aggregate gap, not the rural markets themselves at the within-intermediary margin.

  2. It does not provide a causal estimate of the effect of NMTC on leverage. The fixed-effects decomposition is selection-on- observables identification within CDE × year × type cells. It tells us about the deals CDEs actually do, not about what would happen if a CDE were forced to do a randomly-selected rural deal. The fully causal piece — the LIC-eligibility regression discontinuity — is the planned extension.

  3. It does not say any individual CDE could not deploy more leverage in rural tracts. The result is an average across the CDE fixed- effects distribution. Particular CDEs (typically rural specialists) may have systematically lower mobilization capacity for reasons that would respond to capacity-building, not to changes in program design.

8.3 Policy implications, in order of strength of evidence

(i) The aggregate rural mobilization gap is amenable to intermediary- allocation policy levers. The marginal rural deal is not constrained by market structure under the existing program design. Reallocating allocation awards toward CDEs with demonstrated mobilization capacity — regardless of those CDEs’ current rural orientation — could close the aggregate gap without redesigning the credit itself.

(ii) Allocation reform plausibly dominates program redesign. If the binding constraint were rural-market structure, the policy answer would be to reshape the credit (richer subsidy in rural, longer compliance period, layered grants). The empirical evidence here suggests that’s the wrong knob. The right knob is which CDEs are chosen to deploy rurally — and the standards by which they are selected.

(iii) The 20% statutory mandate may be binding at the wrong margin. The mandate, as currently enforced, requires CDEs to commit to ≥20% non-metro deployment in their competitive applications. But the realized-deployment distribution shows no bunching at 20% — suggesting either that the mandate binds at the allocation-award stage but is violated downstream (worth investigating), or that the realized-deployment margin is bimodal and the 20% line is in the trough between modes. Either way, the policy goal (more rural mobilization) and the mandate’s actual margin of operation appear to be misaligned.

These are suggestive implications given the descriptive nature of the decomposition. The causal RDD piece, when added, will sharpen the mechanism story but is unlikely to change the headline.

8.4 Connection to the broader blended-finance literature

The blended-finance literature has long debated the right mobilization ratio benchmark for public-credit-subsidized capital — and has done so without a setting in which the ratio is directly observable at the project level. NMTC is unusual precisely because the CDFI Fund makes both numerator (QLICI) and denominator (project cost) public.

Bringing the blended-finance vantage to a U.S. place-based subsidy is, to our knowledge, not previously done at the econometric level. Every existing causal paper on NMTC examines downstream neighborhood effects (Freedman 2012; Harger and Ross 2016; Theodos et al. 2022; Freedman and Kuhns 2018); this paper is the first to take the upstream capital-stack — the leverage and the mobilization ratio — as the central outcome variable.

The descriptive precedent is Theodos et al. (2021), who classify CDEs into five mutually exclusive types and tabulate funding patterns descriptively. They do not run any regression, observing only that “CDE type also associates with project types, with important correlations among CDE types and certain types of projects” (p. 7). This paper’s contribution is the regression extension of that observation: we use CDE fixed effects to formally absorb the heterogeneity Theodos et al. document, and decompose the rural-leverage gap into between-CDE and within-CDE components.

8.5 The methodological template

Three transferable elements of the strategy that travel across institutional settings:

  1. Leverage as outcome. Any blended-finance program with project- level disclosure of public + total financing supports the same measurement framework. International examples include EIB project finance and EU Cohesion Fund deployments (European Investment Bank 2024).
  2. Intermediary FE for decomposition. Wherever capital flows through a regulated intermediary structure, the within-intermediary versus between-intermediary decomposition asks the same question.
  3. Rural-vs-urban heterogeneity framing. The U.S. metro / non-metro binary maps onto Eurostat’s NUTS3 urban-rural typology, USDA RUCA codes, OECD predominantly-rural classifications, and analogous country-specific schemes globally.

The within-program decomposition is one paper; the comparative work that asks whether the same finding holds in different institutional settings is what gives the methodology external validity.