4  Descriptive findings

The empirical fact this paper exists to explain is visible before any regression. This chapter walks through six figures and one summary table that establish the rural mobilization gap as real, persistent across project types, and tied to enormous heterogeneity in CDE-level deployment behavior.

4.1 Headline numbers

metric value
Total QLICI deployed, FY2001–FY2022 $66.6 billion
Total project cost (public + private combined) $120.9 billion
Implied program-wide mobilization ratio 0.82×
Number of QLICI transactions 19,907
Number of unique projects 8,024
Number of unique CDEs ~600
Number of unique census tracts ~6,500
Non-metro share of dollars 19.6%
Non-metro share of transactions 19.3%
Statutory non-metro target 20.0%

For every $1 of federal credit deployed, the program pulled in approximately $0.82 of additional non-federal capital. That’s the program-wide blended-finance number. It implies the federal credit is roughly 55% of total project cost on average — a program designed to catalyze private investment is being only barely matched dollar-for- dollar by private capital. NMTC is closer to a direct grant than a high-leverage instrument.

4.2 The metro-vs-non-metro split

metro non-metro
Number of projects 6,463 1,561
Total QLICI deployed $53.6 B $13.0 B
Total project cost $97.2 B $23.8 B
Mean leverage 1.99× 1.73×
Median leverage 1.19× 1.07×
Mobilization ratio (aggregate) 0.81× 0.82×

Two things to highlight:

  1. The aggregate mobilization ratios are nearly identical (0.81× metro vs 0.82× non-metro) — what CDFI tells the world. The aggregate hides the median story.
  2. The median leverage gap is real: 1.19× metro versus 1.07× non- metro. The right tail of leverage is long; the median is more informative than the mean. The typical rural deal mobilizes essentially zero additional private capital (1.07× is just barely above the 1× floor).

4.3 Annual deployment

Figure 4.1: Annual NMTC dollars deployed, FY2001–FY2022, stacked metro vs non-metro.

The program ramped from almost nothing in 2001 to roughly $5 billion per year through the 2010s, then settled into a $3–4 billion annual steady state. The non-metro share (red base of each bar) is visibly present every year.

4.4 Non-metro share over time

Figure 4.2: Non-metro share of dollars each year, with the 20% statutory mandate overlaid.

The non-metro share hovers above the 20% target in most years — suggesting CDEs in aggregate over-perform the floor — and dips slightly below it in only a few. The line and the floor are tightly co-moving, which is the descriptive signature of a binding constraint.

4.5 The leverage distribution

Figure 4.3: Density of project-level leverage (winsorized 1×–20×), metro vs non-metro. Both distributions pile against the 1× floor; the non-metro mass at 1× is substantially higher.

This is the empirical fact the rest of the paper is built around:

  • Both distributions have a heavy mass at leverage = 1× — the floor meaning 100% NMTC-financed, zero private capital mobilized.
  • The non-metro mass at the floor is substantially higher than metro. Rural deals very disproportionately do not mobilize any additional private capital.
  • Both have long right tails, but metro has a noticeably fatter right shoulder (more deals at 1.5×, 2×, 3×). Metro deals more often stack additional private capital on top.

Median: 1.19× metro vs 1.07× non-metro, gap = 0.12. Mean: 1.99× vs 1.73×, gap = 0.26.

4.6 QALICB-type composition differs by metro status

Figure 4.4: QALICB-type composition, metro vs non-metro projects.

A natural objection: maybe the rural leverage gap is just a project- type composition story. Real-estate deals leverage more than operating- business deals (you can stack mezzanine debt, second mortgages, etc.), and rural skews toward operating businesses. Maybe the metro-vs-non- metro gap is really an RE-vs-NRE gap.

This is the correct objection. We address it directly:

4.7 The gap holds within every QALICB type

Figure 4.5: Median leverage by QALICB type, metro vs non-metro. The leverage gap holds within every project type.
within QALICB type metro median non-metro median
Real estate (RE) 1.32× 1.11×
Non-real-estate (NRE) 1.08× 1.04×
Special-purpose (SPE) 1.18× 1.14×
Loan-to-CDE 1.00× 1.00×

The gap is real and it is not composition.

4.8 CDE-level heterogeneity is enormous

The top-20 CDEs together do roughly 50% of all NMTC dollars. Their non-metro shares span from approximately 0% to 80%:

CDE total $M non-metro share
Rural Development Partners LLC 636 80%
Montana Community Development Corporation 660 70%
Midwest Minnesota Community Development Corporation 662 60%
Coastal Enterprises, Inc. 679 40%
Advantage Capital Community Development Fund 1,439 30%
Truist Community Development Enterprises 654 20%
Stonehenge Community Development 797 20%
Chase New Markets Corporation 717 20%
USBCDE LLC 935 10%
Local Initiatives Support Corporation 1,123 10%
ESIC New Markets Partners 1,057 0%
Consortium America LLC 759 0%
Capital Impact Partners 660 0%
National Trust Community Investment Corporation 657 0%

Same federal credit. Same statute. Wildly different organizations, wildly different deployment patterns. This heterogeneity is the variation the empirical strategy of Chapter 5 exploits.