We asked a simple question that no one had formally answered: in the aggregate, how much of an FHA denial is actually about the borrower?
Using eight years of federal HMDA data — 4,124 lender-state-year observations across eleven major FHA lenders and all fifty-one U.S. jurisdictions — we decomposed denial rate variance into four components: lender identity, market timing, state geography, and the residual (which includes borrower characteristics like FICO score, DTI, documentation quality).
The math: 63.4% (lender) + 5.3% (timing) + 2.4% (geography) = 71.1% outside the borrower's control. The remaining 28.9% is the residual — FICO, DTI, income type, documentation quality — the factors a borrower can actually influence.
One objection to this finding: maybe the lender effect is concentrated in a single unusual year. It isn't. The lender identity component was the dominant factor in every year from 2018 to 2025 — through COVID, through the fastest rate-hike cycle in four decades, through the 2021 boom and the 2023 freeze.
Each number represents the share of denial variance explained by lender identity alone in that calendar year, across all 51 jurisdictions.
This finding applies to aggregate variance, not individual outcomes. An individual borrower's denial may be entirely about their credit profile. The 71% figure describes the system as a whole: if you could randomly reassign all FHA applicants to different lenders while keeping everything else constant, the lender reassignment alone would explain the majority of changes in who gets approved.
This is not a finding about discrimination, fraud, or regulatory failure. It is a finding about institutional heterogeneity — lenders applying the same federal guidelines with radically different interpretations, overlays, and risk appetites.
If the majority of denial variance is outside your control, the most productive response to a denial is not necessarily to repair your credit score. It may be to route your application differently. The same file that fails at one lender may succeed at another — not because one lender is better or worse, but because they have different institutional behaviors.
This is why routing exists. Borrower optimization (improve FICO, reduce DTI) addresses the 29%. Routing optimization addresses the 71%.
This analysis uses annual lender-state aggregates from HMDA public data, not individual application records. The residual term captures everything not explained by lender, year, and state fixed effects — including borrower characteristics, but also unmeasured lender policies, application mix selection, and data noise. The true "borrower-controlled" share may be higher or lower than 28.9%. The 71% figure is a lower bound on systemic influence, not a precise estimate of individual blame attribution.
This is an FHA-specific finding. Conventional lending may show different patterns.