## 1.    Parametric approach

## Source details

**Canonical URL:** [1.    Parametric approach](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11215.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11215.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11215.pdf.json)

---

### Overview
- Section summarizes the paper’s examination of the role of less-regulated non-bank mortgage originators (independents) in the U.S. mortgage boom and subsequent rise in foreclosures, using county-level variation and multiple empirical strategies (parametric regressions, instrumental variables, matching, and robustness checks).
- Adjacent methodological sections referenced: Matching methods; Parametric Results (credit, house prices, and unemployment); The Abadie-Imbens Estimator; OLS on the matched subsample; Discussion; Exploring the Role of Regulation and Alternative Hypotheses; Conclusion; Data Appendix.

### Main empirical findings
- Independents accounted for around one-third of mortgage lending in 2003, and they contributed to more than 60% of the increase in mortgage lending between 2003 and 2005.
- Market share of independents as of 2005 is a strong predictor of:
  - the rise in foreclosures (county-level variation),
  - the contraction in mortgage credit during the downturn,
  - the contraction in house prices during the downturn,
  - the rise in unemployment starting in 2008.
- The independents–foreclosure relation remains significant after controlling for:
  - county economic and demographic differences,
  - measures of credit and house price growth during the boom,
  - measures of securitization,
  - measures of geographical diversification and lender competition,
  - a weighted measure of core deposit ratio of lenders in a county.
- Instrumental variable approach: house price growth is instrumented using housing supply elasticity; evidence indicates the early rise in foreclosures preceded the fall in house prices.
- Matching and semi-nonparametric robustness: Abadie-Imbens bias-adjusted matching estimator used; results repeated on matched subsamples (including intrastate matching) to reduce confounding.

### Data and scope
- Primary mortgage data source: Home Mortgage Disclosure Act (HMDA) comprehensive sample, 2003–2008.
- Sample restrictions and coverage:
  - Focus on home purchase, conventional, one-to-four-family, owner-occupied applications.
  - Counties restricted to those in a Metropolitan Statistical Area (MSA) with available house price growth and housing supply elasticity.
  - Final sample: 773 counties, accounting for around 80% of total mortgage originations in 2005 reported in HMDA.
  - 2003–2008 sample period consists of around 28 million applications aggregated to the county level.
- Foreclosure data: Realty Trac Foreclosure Market Trend Reports, using 2005Q2, 2006Q2, 2007Q2 to measure early increase in quarterly foreclosure filings.
- Supplementary data: ICPSE county characteristics; FHFA house prices at MSA level; TransUnion Trend Data for credit scores; housing supply elasticity from Saiz (2010); state regulation data from Pahl (2007) and Bostic et al. (2008); core deposit ratios from Call Report and SDI (FDIC).

### Summary statistics and stylized facts
- Loan-level patterns:
  - Originated loans increased between 2003 and 2005, then decreased between 2005 and 2007; 2005 was the peak year.
  - Independents made around 31% of loans in 2003 and accounted for more than 60% of the increase in originations between 2003 and 2005 and the decrease between 2005 and 2007.
  - 2005 jumbo cutoff: $359,650 (banks more active in jumbo market).
  - Loan-to-income (LTI): independents gave higher LTI loans in 2003 and 2007 but lower LTI loans in 2005.
- County-level patterns:
  - In the average county, mortgage credit grew by around 30% between 2003 and 2005.
  - Mortgage credit contracted by more than 80% between 2005 and 2007.
  - Distribution of independents’ market share: mean and median around 23%.
- Mortgage credit expansion (2003–2005) — summary figures:
  - Change in total mortgage volume (banks + independents) between 2003 and 2005: around 33%.
  - On average, credit growth by independents was around 23% higher than that of banks.
  - Market share of independents grew by around 4% on average between 2003 and 2005.
  - New foreclosures filed for 0.1% of properties during 2005Q2; standard deviation around 0.11.
  - New foreclosure filings doubled between 2005Q2 and 2006Q2 and nearly tripled between 2005Q2 and 2007Q2.
  - House prices: average growth rate 27% and median 19% between 2003 and 2005; growth rate substantially declined between 2005 and 2007.

### Geographic patterns and endogeneity handling
- Independents expanded disproportionately into areas with higher house price growth.
- House price growth (2003–2005) instrumented with housing supply elasticity (Saiz (2010)):
  - Instrument expected to be negatively correlated with house price growth (lower elasticity → larger price response).
  - Regression of change in independents’ market share on housing supply elasticity (controlling for 2003 market share) yields a negative and significant coefficient: independents expanded more in areas with lower housing supply elasticity.
  - IV second-stage: positive relation between instrumented house price growth and independents’ expansion; coefficient larger and significant at the 10% level for independents (banks’ coefficient small and far from significant).

### Parametric benchmark specification
- Outcome variable: ∆05Q2−07Q2 Forci (change in new foreclosure filing rates between 2005Q2 and 2007Q2 in county i).
- Benchmark OLS:
  - ∆05Q2−07Q2 Forci = β0 + β1 Independenti,05 + β1 Xi,05 + β2 ∆03−05 Zi + β3 Securitizationi,05 + εi
  - Independenti,05: market share of independent lenders in 2005.
  - Xi,05 controls include per-capita income and unemployment in 2005; per-capita income growth 2003–2005; categorical credit score measures; six census-tract income decile shares; share Black, share Hispanic, average immigration rate 2000–2005; housing supply elasticity.
  - ∆03−05 Zi captures 2003–2005 mortgage boom (house price and mortgage credit growth).
  - Securitizationi,05: share of originated loans sold for private securitization.
- Rationale: control for county confounders, local housing boom extent, and securitization because independents securitized a higher share of loans.

### Parametric results (highlights and exact numerical implications)
- Securitization:
  - A one standard deviation increase in the securitization rate → increase of 0.04 in the foreclosure filing rate.
  - Interpreted as: 4 properties in every 10000 properties per quarter or 1/5th of the increase in average filings between 2005Q2 and 2007Q2.
- Market share of independents:
  - A one standard deviation increase in the market share of independents → increase of 0.08 in the foreclosure filing rate.
  - Interpreted as: about 40% of the increase in average filings between 2005Q2 and 2007Q2.
  - Including market share reduces the magnitude and significance of the securitization coefficient.
- Instrumental variables diagnostics:
  - First IV (instrumenting house price growth by housing supply elasticity and lagged house price growth): First stage F-statistic=14.1; partialR2=0.05. Sargan and Bassmann test p-values: 0.96 and 0.97.
  - Second IV (instrumenting 2007 house price growth): First stage F-statistic=31.2; partialR2=0.11. Sargan and Bassmann p-values: 0.52 and 0.55.
- Timing and alternative windows:
  - Dependent variable using 2006–2007 foreclosures: coefficient on independents decreases in magnitude but remains significant. Controlling for 2003–2005 house price and credit growth, mortgage credit growth is positive and significant and captures some independents’ effect. Instrumented 2007 house price growth yields a negative and significant coefficient.

### Credit, house prices, and unemployment (parametric outcomes)
- Mortgage credit (2005–2007):
  - A one standard deviation increase in independents’ market share → contraction of around 5% in mortgage credit between 2005 and 2007 (noted as -0.498*0.1 in source).
  - Higher market share of securitization associated with sharper contraction in credit, but loses significance when controlling for boom expansion.
- House prices (2005–2007):
  - Negative relation between independents’ market share and house price growth between 2005 and 2007; significant when controlling for boom-period credit and price growth.
  - Stronger prior housing boom predicts higher subsequent house price persistence.
- Unemployment:
  - Change in unemployment 2005–2007: coefficient on independents positive but not significant.
  - Change in unemployment 2005–2008: a one standard deviation increase in independents’ market share → increase of 0.16 in the unemployment rate (0.16 percentage points).
  - Context: U.S. unemployment rate in 2007 was 5% versus 4.9% in 2005.

### Matching methods — motivation and implementation
- Motivation: address poor distributional overlap, undue linear-model weight, potential non-linearity, and imbalance in county characteristics correlated with independents’ market share.
- Approach:
  - Use Abadie-Imbens bias-corrected matching estimator to estimate treatment effects and re-run linear regressions on matched subsamples.
  - Treatment: counties in upper 15% of independents’ market share in 2005 (cutoff = 0.3854). Treated counties: 107.
  - Matching covariates selected where causality from treatment to covariate is unlikely: per capita income, average credit score, housing supply elasticity, unemployment rate.
  - Benchmark interstate matching (no geographic restriction) and intrastate matching (within-state) to address policy and legal confounders.

### Matching diagnostics and matched-sample findings
- Balancing tests used: Kolmogorov-Smirnov (KS) distributional tests and t-tests for means.
- Interstate matching:
  - Treated vs control: treated group experienced significantly worse change in foreclosures; covariates show strong similarity (KS cannot reject same distribution; t-tests cannot reject equality of means).
- Intrastate matching:
  - Imposing within-state restriction makes matching harder but KS and t-tests indicate covariates’ distributions and means are not significantly different; KS p-value on income = 0.12 (treated slightly higher income).
  - Foreclosure outcomes remain significantly worse in treated sample.
- Abadie-Imbens estimator results (benchmark matching, Table10):
  - Treatment associated with increase in foreclosure filings rate by around 0.26 (higher than average increase over period).
  - SATE, SATT, and PATT estimators produce similar magnitudes; impacts on mortgage credit growth and unemployment confirm earlier findings though unemployment impact varies by estimator.
- Intrastate matching (Table11):
  - SATE yields substantially lower differences than benchmark but SATE, SATT, PATT remain significant for foreclosures, credit growth, and unemployment.
  - SATT and PATT yield very similar results on foreclosures in both benchmark and intrastate matches.

### OLS on matched subsamples (exact coefficients preserved)
- Regressions run with state dummies and state-clustered errors; controls included as in benchmark.
- Market share of independents, 2005 — coefficients on change in foreclosures (t-statistics in parentheses):
  - Full sample (foreclosure): 0.840***  (3.37)
  - Interstate matched (foreclosure): 1.685***  (3.97)
  - Intrastate matched (foreclosure): 1.982***  (5.54)
- Market share of independents, 2005 — credit growth (05–07):
  - Full sample: -0.499***  (-3.83)
  - Interstate matched: -0.423*  (-1.76)
  - Intrastate matched: -0.536**  (-2.14)
- Market share of independents, 2005 — unemployment change (05–08):
  - Full sample: 1.900***  (2.90)
  - Interstate matched: 2.078**  (2.45)
  - Intrastate matched: 1.177*  (1.75)
- Sample sizes and adj. R2 (examples from Table12):
  - Full-sample N: 583 (foreclosure), 599 (credit), 599 (unemployment). adj. R2 foreclosure full-sample: 0.504.
  - Interstate matched N: 278 (foreclosure), 280 (credit), 280 (unemployment). adj. R2 interstate matched foreclosure: 0.572.
  - Intrastate matched N: 162 (foreclosure), 162 (credit), 162 (unemployment). adj. R2 intrastate matched foreclosure: 0.414.
- Interpretation: coefficients on matched subsamples are larger in magnitude; intrastate matched results yield the largest estimated effect on foreclosures.

### Robustness to alternative hypotheses and controls
- Competition and diversification:
  - Herfindahl indexes (top 15, 30, 50) are far from significant; coefficient on independents remains positive and significant and often larger in smaller samples.
  - Geographic diversification index small and not significant.
- Deposit-taking heterogeneity:
  - Median core deposits-to-assets ratio in sample banks (2005) = 0.51; many banks have ratios <0.2; independents assigned ratio = 0.
  - Controlling for county shares of loans from banks with CD>0.51 and CD>0.61 suggests deposit-taking differences unlikely to drive results.
- Securitization controls:
  - Controlling for private securitization explains at best a small fraction of independents’ effect.
  - Percent sold to GSEs: negative (not significant) coefficient; slightly reduces independents coefficient (GSEs require minimum standards).
  - Percent not securitized: positive and not significant.

### Role of state regulation (exact interactions and findings)
- State regulation datasets:
  - Bostic and others (2008) — state anti-predatory lending laws.
  - Pahl (2007) — state regulation of mortgage brokers.
- Constructed dummies for upper quartile (most regulated states) for broker regulation and anti-predatory laws; interacted these with Independents,2005.
- Selected results (Table14):
  - Broker regulation dummy coefficient: negative and significant (example: -0.0596*** in column (1)).
  - Interaction Independents×broker: -0.956*** (column (2)), implying weaker independents effect in highly broker-regulated states. Interaction remains significant at 5% when clustering at state level.
  - Interaction Independents×anti-predatory: -0.660*** (column (5)) and -0.660* when clustered at the state level (column (6)), implying a weaker independents effect in states with stronger anti-predatory laws (smaller significance than broker interaction).
- Controls included state per capita GDP (2002) and foreclosure-law dummies (judicial foreclosure, statutory right of redemption, deficiency judgment rules).
- Interpretation: the independents–foreclosure association is weaker in states with tighter mortgage-related regulation, consistent with the regulatory-explanation hypothesis (while acknowledging regulation proxies are imperfect).

### Conclusions and policy implications
- Summary conclusions:
  - Lightly regulated independent lenders contributed disproportionately to the mortgage boom and to worse outcomes in the subsequent downturn.
  - Market share of independents as of 2005 strongly predicts the increase in foreclosure rates between 2005 and 2007 and is associated with larger contractions in credit and house prices and higher unemployment between 2007 and 2008.
  - Results robust to parametric controls, IVs, matching methods (interstate and intrastate), and multiple specifications.
  - Robustness tests point to weak regulatory structure governing independents as a plausible explanation.
- Policy implication:
  - Findings lend support to the view that more stringent regulation of independent mortgage originators might have reduced some of the volatility during the boom-bust episode.
- Suggested further research:
  - Use disaggregated data to identify specific characteristics that made independent lending riskier than bank lending.

*Source: _wp11215 - 1.    Parametric approach (table of contents and section listing) and chapter excerpts from the specified IMF PDF.*

### 1.    Parametric approach  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .16

### 1.    Parametric approach

### Overview
- Section headings and related subsections as listed in the source PDF.

### Adjacent methodological sections (from table of contents)
- 2.    Matching methods
- C. Parametric Results
  - (a)  Credit, house prices, and unemployment
- D. Matching results
  - 1.  Balancing tests
  - 2.  The Abadie-Imbens Estimator
  - 3.  OLS on the matched subsample
- E. Discussion

### Further exploration and appendices
- V. Exploring the Role of Regulation and Alternative Hypotheses
  - A.  Alternative Hypotheses
  - B. State regulation
- VI.  Conclusion
- VII. Data Appendix

### Tables referenced (table numbers and titles preserved)
- 1. Summary statistics: loan originations.
- 2. Statistical summary of the county level variables.
- 3. Expansion of the Independents
- 4. Expansion and house prices.
- 5. The relation between the rise in foreclosure rates and the market share of independents
- 6. Early rise in foreclosure rates in counties with high and low mortgage credit growth
- 7. The relation between market share of independents and the change in credit, house prices, and unemployment
- 8. Comparison between treated and control samples obtained from the interstate matching procedure
- 9. Comparison between treated and control samples obtained from the intrastate matching procedure
- 10. Matching estimators, interstate.
- 11. Matching estimators, intrastate.
- 12. OLS regressions on the full, interstate matched, and intrastate matched samples
- 13. The relation between the rise in foreclosure rates and market share of independents: Robustness analysis

*Source: _wp11215 - 1.    Parametric approach (table of contents and section listing) from the specified IMF PDF.*

### 14.    The relation between the rise in foreclosure rates and market share of indepen-

### 14.    The relation between the rise in foreclosure rates and market share of independents: the role of regulation

### Introduction and research question
- The global financial crisis had its roots in the U.S. housing market following an unprecedented boom in mortgage lending and a sharp increase in mortgage delinquency in 2006.
- The paper examines whether less regulated non-bank mortgage originators (independents) contributed disproportionately to the mortgage boom and whether their lending was associated with a sharper increase in foreclosures.
- Key institutional contrast: banks were more regulated under federal banking laws and subject to federal examinations (CRA, TILA, fair lending, risk-based capital, FDIC supervision), while independents were lightly regulated and supervised at the state level.

### Main empirical findings
- Independents accounted for around one-third of mortgage lending in 2003, and they contributed to more than 60% of the increase in mortgage lending between 2003 and 2005.
- The market share of independents as of 2005 is a strong predictor of:
  - the rise in foreclosures (using county-level variation),
  - the contraction in mortgage credit during the downturn,
  - the contraction in house prices during the downturn, and
  - the rise in unemployment starting in 2008.
- The relation between the market share of independents and the rise in foreclosures remains significant after controlling for:
  - economic and demographic county differences,
  - measures of credit and house price growth during the boom,
  - measures of securitization (independents securitized a significantly larger share of originations, but market share of independents largely explains the securitization–foreclosure relation),
  - measures of geographical diversification and lender competition,
  - a weighted measure of core deposit ratio of lenders in a county (derived from bank and thrift balance sheet data).
- Instrumental variable approach: house price growth is instrumented using housing supply elasticity; the early rise in foreclosures preceded the fall in house prices.
- Matching and semi-nonparametric robustness: Abadie-Imbens bias-adjusted matching estimator is used, and results are repeated on matched subsamples (including intra-state matching) to minimize confounding and specification dependence.

### Data and scope
- Primary mortgage data source: Home Mortgage Disclosure Act (HMDA) comprehensive sample of mortgage applications and originations between 2003 and 2008.
- Sample restrictions:
  - Focus on home purchase, conventional, one-to-four-family, owner-occupied applications.
  - Limit to counties situated in a Metropolitan Statistical Area (MSA) for which house price growth and housing supply elasticity are available.
  - Final sample: 773 counties, which account for around 80% of total mortgage originations in 2005 reported in HMDA.
  - 2003–2008 sample period consists of around 28 million applications aggregated to the county level.
- Foreclosure data: Realty Trac Foreclosure Market Trend Reports, using the second quarter of 2005, 2006, and 2007 to measure the increase in quarterly foreclosure filings prior to the liquidity crisis and official start of the recession.
- Supplementary data:
  - County characteristics from ICPSE.
  - FHFA house prices at the MSA level.
  - TransUnion Trend Data for average consumer credit score and percentage of low consumer credit score in a county.
  - Housing supply elasticity measure constructed by Saiz (2010).
  - State regulation data on mortgage brokers from Pahl (2007) and state-level anti-predatory lending laws from Bostic et al. (2008).
  - Core deposit to total assets ratio for depository institutions from Call Report and SDI (FDIC).

### Summary statistics and stylized facts
- Loan-level patterns:
  - Number of originated loans increased between 2003 and 2005, then decreased between 2005 and 2007 for both banks and independents.
  - 2005 was the peak year in loan originations.
  - Independents made around 31% of loans in 2003 and accounted for more than 60% of the increase in mortgage originations between 2003 and 2005 and the decrease between 2005 and 2007.
  - Banks on average made loans to higher income applicants; part of this difference reflects bank activity in the jumbo loan market (2005 jumbo cutoff: $359,650).
  - Excluding jumbo loans narrows differences in applicant income and loan amount between banks and independents.
  - Loan-to-income (LTI) findings: independents gave higher LTI loans in 2003 and 2007 but lower LTI loans in 2005.
- County-level patterns (Table 2 highlights):
  - In the average county, mortgage credit grew by around 30% between 2003 and 2005.
  - Mortgage credit contracted by more than 80% between 2005 and 2007.
  - The share of loans originated by independents varies substantially across counties; the distribution is relatively symmetric with mean and median market share around 23%.
- Coverage and measurement notes:
  - HMDA requires income reporting when it was relied upon in the credit decision; around 4% of loans lack applicant income.
  - The sample focuses on larger counties (MSA counties) to ensure availability of house price growth and housing supply elasticity and to reduce noise from small-population areas.

### Role of regulation and heterogeneity across states
- Hypothesis: state-level regulation that constrains risky lending should have a larger impact on independents (who were less constrained federally) than on banks.
- Empirical evidence:
  - The impact of independents on foreclosures was smaller in states that tightened regulation prior to and during the boom, using datasets on anti-predatory laws and broker laws.
- Policy-relevant interpretation: the heterogeneous regulatory and supervisory environment across lender types plausibly contributed to differences in lending standards and subsequent default outcomes.

### Methodological approach and robustness
- Main identification strategy: exploit cross-county heterogeneity in market share of independents and in house price growth (instrumented by housing supply elasticity) to estimate effects on foreclosures and downturn severity.
- Controls and robustness checks include:
  - county economic and demographic controls,
  - measures of securitization (private securitization defined as in Mian and Sufi (2009a); additional securitization measures tested),
  - measures of lender competition and geographical diversification,
  - bank core deposit ratios,
  - semi-nonparametric matching (Abadie-Imbens bias-adjusted estimator) and regressions on matched subsamples,
  - intra-state matching.
- Time framing: focus on early rise in foreclosures prior to the liquidity crunch to limit contamination from subsequent financial-sector disruptions.

### Conclusions and implications
- The expansion of independents during the mortgage boom was substantial and concentrated in higher house-price-growth areas.
- Market share of independents as of 2005 is strongly associated with worse county outcomes in the subsequent downturn (higher foreclosures, larger credit contraction, larger house-price declines, higher unemployment).
- Results are robust to a wide array of controls, alternative specifications, matching procedures, and instruments for house price growth.
- Findings are consistent with the interpretation that less-regulated independents contributed to a deterioration in lending standards, with regulatory differences across lender types playing a meaningful role in the housing boom–bust dynamics.

*Source: IMF working paper chapter 14, "The relation between the rise in foreclosure rates and market share of independents: the role of regulation."*

### 2003. This market share has increased by 4% in 2005, due to the faster expansion of inde-

### _wp11215 - 2003. This market share has increased by 4% in 2005, due to the faster expansion of inde-

### Mortgage credit expansion: 2003-2005 — summary findings
- The year 2005 constituted the peak of a mortgage boom that started in early 2000s and accelerated to register unprecedented levels of mortgage growth between 2003 and 2005.
- The change in total mortgage volume (banks + independents) between 2003 and 2005 is estimated at around 33%.
- On average, credit growth by independents was around 23% higher than that of banks.
- The market share of independents grew by around 4% on average between 2003 and 2005.
- Independents expanded disproportionately into areas where they had lower presence in the past (inward vs. outward expansion).
- New foreclosures were filed for 0.1% of properties during 2005Q2; the measure shows significant variation with a standard deviation around 0.11.
- New foreclosure filings doubled between 2005Q2 and 2006Q2 and nearly tripled between 2005Q2 and 2007Q2.
- House prices were increasing rapidly between 2003 and 2005 with an average growth rate of 27% and a median of 19%. The growth rate substantially declined between 2005 and 2007; house prices entered their downturn trend only later in 2007 and early 2008.

### Geographic patterns and association with house price booms
- A scatter of county-level market shares shows the expansion in the market share of independents took place in the vast majority of U.S. counties.
- Independents tended to expand more into areas that experienced higher house price growth.
- To address endogeneity (independents potentially affecting house prices), house price growth between 2003 and 2005 is instrumented with the region’s housing supply elasticity (from Saiz (2010)), which is based on geographical characteristics of an MSA and is exogenous to changes in mortgage credit.
- The housing supply elasticity is expected to be negatively correlated with house price growth between 2003 and 2005 (lower elasticity → more responsive prices to demand shocks).
- A regression of the change in the market share of independents on housing supply elasticity, controlling for 2003 market share, yields a negative and significant coefficient: independents expanded more in areas with lower housing supply elasticity.
- IV results: the instrument is strongly correlated with house price growth (first stage). Second stage regressions show:
  - A positive relation between house price growth and bank lending growth, but the coefficient is small and far from significant.
  - For independents, the coefficient on instrumented house price growth is larger in magnitude and significant at the 10% level, suggesting independents expanded relatively faster in areas experiencing a house price boom.

### The rise in foreclosures and the role of independents — empirical associations
- Controlling for county characteristics, the market share of independents is a strong predictor of the early rise in foreclosures.
- Counties with higher market shares of independents tended to have worse outcomes during the crisis, including larger increases in foreclosure filings, larger contractions in credit, larger declines in house prices, and larger rises in unemployment (figures referenced for 2005–2008 comparisons).
- The rise in foreclosures began as early as 2006 and was pronounced in areas that historically had low foreclosure rates.
- The pre-crisis market share of independents is not fully explained by county economic and demographic characteristics or by factors directly related to the housing boom; historical presence and proximity to strong-presence areas explain a large share of 2005 market share, and the 2003–2005 increase was around 4%.

### Empirical methodology — parametric benchmark
- Outcome of interest: ∆05Q2−07Q2 Forci — change in new foreclosure filing rates between 2005Q2 and 2007Q2 in county i.
- Benchmark OLS specification:
  - ∆05Q2−07Q2 Forci = β0 + β1 Independenti,05 + β1 Xi,05 + β2 ∆03−05 Zi + β3 Securitizationi,05 + εi
  - Independenti,05: market share of independent lenders in 2005 (base/peak year).
  - Xi,05: county-specific controls from or prior to 2005, including:
    - per-capita income and unemployment in 2005;
    - per-capita income growth during 2003–2005;
    - categorical variables capturing average consumer credit score and percentage of low credit score consumers;
    - six variables capturing the share of census tracts in a county with median income in one of six deciles below 60K;
    - share of Black population, share of Hispanic population, average immigration rate between 2000 and 2005;
    - housing supply elasticity.
  - ∆03−05 Zi: extent of the mortgage boom between 2003 and 2005, captured by growth in house prices and mortgage credit during that period.
  - Securitizationi,05: share of originated loans in a county that were sold for private securitization.
- Rationale for controls:
  - Control for county factors that could confound the relation between independents’ market share and foreclosure outcomes.
  - Control for extent of local housing boom (∆03−05 Zi) because higher Independenti,05 might be associated with faster expansion in credit and house prices.
  - Control for securitization because independents securitized a higher share of loans and securitization has been linked to worse lending standards.

### Interpretation and implications noted in the text
- Patterns are suggestive of faster deterioration in lending standards among independents, potentially due to lighter regulation and reliance on an originate-to-distribute (OTD) model.
- The interpretation that independents’ lending was associated with worse outcomes aligns with calls during the crisis for tighter regulation on the “shadow banking” sector, including independent mortgage lenders, though the text notes this remains an interpretation without direct causal proof beyond the presented empirical associations.

*Source: _wp11215 (excerpt).*

### 2.  Matching methods

### 2.  Matching methods

### Parametric results
- Dependent variable: change in new foreclosure filing rates between the second quarter of 2005 and the second quarter of 2007.
- Regressions include state dummies and cluster errors at the state level.
- Securitization:
  - Associated with an increase in foreclosure filings.
  - Coefficient implication: a one standard deviation increase in the securitization rate leads to an increase of 0.04 in the foreclosure filing rate.
  - Interpreted as: 4 properties in every 10000 properties per quarter or 1/5th in the increase in average filings between 2005Q2 and 2007Q2.
- Market share of independents:
  - When included, the coefficient is positive and significant at the 1% level.
  - Coefficient implication: a one standard deviation increase in the market share of independents is associated with an increase of 0.08 in the rate of foreclosure filings.
  - Interpreted as: this is about 40% of the increase in average filings between 2005Q2 and 2007Q2.
  - Including market share of independents reduces the significance and magnitude of the securitization coefficient, suggesting earlier estimates were capturing the effect of independents via their higher securitization rate.
- Controls and robustness:
  - Adding house price growth between 2003 and 2005 and mortgage credit growth over that period does not significantly affect the coefficient on independents; their coefficients are not significant in this early-rise period.
  - Second-stage regression instrumenting house price growth (instrumented by housing supply elasticity and lagged house price growth):
    - Estimate of instrumented house price growth is negative but not significant.
    - First stage F-statistic=14.1 and gives a partialR2=0.05.
    - The Sargan and Bassmann overidentification tests yield p-values of 0.96 and 0.97, respectively.
  - When the dependent variable is the change in foreclosures between 2006 and 2007:
    - Coefficient on independents decreases in magnitude but remains significant.
    - Controlling for house price growth and mortgage credit growth, mortgage credit growth is positive and significant and captures some of the effect of independents.
    - Instrumenting for house price growth in 2007 yields a negative and significant coefficient.
      - First stage F-statistic=31.2 and the partialR2=0.11.
      - The Sargan and Bassmann overidentification tests yield p-values of 0.52 and 0.55, respectively.
- Credit, house prices, and unemployment (early downturn period):
  - Market share of independents as of 2005 has a strong and significant negative impact on mortgage credit growth between 2005 and 2007.
    - A one standard deviation increase is associated with a contraction of around 5% in mortgage credit between 2005 and 2007 (noted as -0.498*0.1 in the text).
  - Higher market share of securitization is associated with a sharper contraction in credit, but this loses significance when controlling for expansion in credit and house prices during the boom.
  - House prices:
    - Negative relation between market share of independents and house price growth between 2005 and 2007, significant when controlling for credit and house price growth during the boom.
    - A more substantial housing boom predicts an increase in house prices between 2005 and 2007 (persistence in house prices).
  - Unemployment:
    - Change in unemployment between 2005 and 2007: coefficient on independents is positive but not significant.
    - Change in unemployment between 2005 and 2008: market share of independents is a significant predictor.
      - A one standard deviation increase in the market share is associated with an increase of 0.16 in the unemployment rate by 0.16 percentage points.
  - Context note: The U.S. unemployment rate in 2007 was 5% in comparison to 4.9% in 2005.

### Matching methods and implementation
- Motivation:
  - Supplement parametric regressions to address poor distributional overlap of control variables, undue weight on a linear model, and possible non-linearity or imbalance of county characteristics correlated with market share of independents.
  - Matching used as nonparametric pre-processing to reduce correlation between key variable and controls and to produce a subsample of counties similar on chosen covariates.
- Estimation approach:
  - Use Abadie-Imbens bias corrected matching estimator to compare treated and control samples and to compute differences in means; re-run linear regressions on matched samples as parametric follow-up.
  - Matching is used to create a pre-processed sample that makes subsequent parametric estimates less dependent on modeling choices.
- Treatment definition and sample:
  - Explanatory variable: percentage of loans originated by independents in 2005 (continuous).
  - Ad-hoc cutoff procedure used for continuous variable; benchmark cutoff is the upper 15% of counties in terms of market share of independents as of 2005.
    - This cutoff corresponds to a market share of independents of 0.3854.
    - Choosing a cutoff corresponding to the higher 10% or 20% of counties gives similar results.
  - Treated subsample size: 107 treated counties.
  - Matching covariates chosen so that causality from treatment to covariate is highly unlikely; covariates are likely correlated with both market share of independents and rise in foreclosures:
    - per capita income
    - average credit score
    - housing supply elasticity
    - unemployment rate
  - Matching performed benchmark without geographical restrictions; an intrastate (within-state) matching exercise is also performed to address concerns about state foreclosure laws.

### Balancing tests and matched-sample findings
- Balancing tests used:
  - Kolmogorov-Smirnov (KS) test of distributional differences.
  - t-test to compare means.
- Benchmark (interstate) matching results:
  - First row of Table 8: change in foreclosures between treated and control counties shows treated group experienced distinctly worse outcomes; KS and t-tests indicate these differences are significant.
  - Next four rows: distribution and means of covariates (per capita income, average credit score, housing supply elasticity, unemployment rate) show strong similarity between treated and control subsamples:
    - KS test cannot reject that covariates are generated by same distribution.
    - p-values from t-tests show inability to reject equality of means.
- Intrastate matching results:
  - Results in Table 9 show similar balancing: imposing same-state restriction makes matching harder but KS and t-tests suggest covariate distributions and means are not significantly different between treated and control subsamples.
  - KS test p-value on income is relatively small (0.12); treated counties on average have slightly higher per-capita income.
  - First row in Table 9: foreclosure outcomes are significantly worse in the treated sample, consistent with interstate matching.

### Key statistics and robustness notes
- Treated sample definition and size:
  - Benchmark cutoff: upper 15% of counties (market share cutoff = 0.3854).
  - Treated counties: 107.
- Instrumental-variable diagnostics:
  - First IV (house price growth instrumented by housing supply elasticity and lagged house price growth):
    - First stage F-statistic=14.1; partialR2=0.05.
    - Sargan and Bassmann test p-values: 0.96 and 0.97.
  - Second IV (house price growth in 2007 instrumented):
    - First stage F-statistic=31.2; partialR2=0.11.
    - Sargan and Bassmann test p-values: 0.52 and 0.55.
- Exact numerical effect statements preserved from source:
  - Securitization: one standard deviation increase → increase of 0.04 in foreclosure filing rate.
  - Independents: one standard deviation increase → increase of 0.08 in foreclosure filing rate.
  - Independents and mortgage credit: one standard deviation increase → contraction of around 5% in mortgage credit between 2005 and 2007 (noted as -0.498*0.1).
  - Independents and unemployment: one standard deviation increase → increase of 0.16 in unemployment rate by 0.16 percentage points.
- Robustness checks:
  - Regression specifications varying time windows (2005–2007, 2006–2007, 2005–2006) yield consistent positive relation between market share of independents and increases in foreclosure rates.
  - Subsample analysis by counties’ mortgage growth in 2005 and 2006 shows relation holds for counties still experiencing mortgage boom and for other subsamples; aggregate patterns and IV regressions minimize concern that results are driven by house price and credit movements alone.

*Source: _wp11215 - 2.  Matching methods*

### 2.  The Abadie-Imbens Estimator

### 2.  The Abadie-Imbens Estimator

### Overview of estimators
- The Abadie-Imbens matching estimator is used to estimate the effect of the “treatment” defined as having a high share of loans originated by independents.
- Three estimators are reported: the sample average treatment effect (SATE), the sample average treatment effect on the treated (SATT), and the population average treatment effect on the treated (PATT). 19

### Benchmark matching exercise (Table10)
- Outcome variables analyzed: differential change in foreclosure filings rate, mortgage credit growth, and change in unemployment rate between treated and control samples.
- Main findings:
  - Treated counties experienced a significantly sharper increase in foreclosures across all estimators, with results of a similar magnitude.
  - The treatment effect (having a high market share of independents) is estimated to be associated with an increase in foreclosure filings rate by around 0.26, which is higher than the average increase in foreclosure filings rate over that period.
  - Results on mortgage credit growth and unemployment confirm earlier findings, though the impact on unemployment varies substantially depending on the estimator used.
- Tables referenced: Table10 reports the benchmark matching results.

### Intrastate matching exercise (Table11)
- Results from the intrastate match are reported in Table11.
- Main findings:
  - The SATE estimator yields substantially lower differences compared with the benchmark.
  - Results from all estimators (SATE, SATT, PATT) are significant for the three variables (foreclosure filings rate, mortgage credit growth, unemployment).
  - The SATT and PATT yield very similar results on the main outcome variable, foreclosures, in both the benchmark and the intrastate matching exercises.

*Source: _wp11215 - 2.  The Abadie-Imbens Estimator*

### 3.  OLS on the matched subsample

### _wp11215 - 3.  OLS on the matched subsample

### OLS results on matched subsamples
- Method:
  - Ran the benchmark linear regression on the subsample of matched counties (see Table12).
  - Controlled (but did not show) for all previously used economic and demographic controls and for state dummies (see Table5).
  - Clustered errors at the state level.
  - First three columns: regressions on the full sample for three endogenous variables: change in foreclosures, credit growth and unemployment.
  - Next three columns: benchmark matched subsample (interstate matching).
  - Last three columns: intrastate matched subsample.
- Key regression coefficient patterns (Table12):
  - Market share of independents, 2005:
    - Full sample (foreclosure): 0.840***  (t-statistic reported as (3.37))
    - Interstate matched (foreclosure): 1.685***  (t-statistic (3.97))
    - Intrastate matched (foreclosure): 1.982***  (t-statistic (5.54))
    - Interpretation: estimated coefficients on the matched subsample are significantly larger in magnitude; column (4) coefficient is twice the size of column (1); intrastate match (column (7)) is even higher.
  - Market share of independents, 2005 on credit growth (05-07):
    - Full sample: -0.499***  (t-statistic (-3.83))
    - Interstate matched: -0.423*  (t-statistic (-1.76))
    - Intrastate matched: -0.536**  (t-statistic (-2.14))
    - Note: coefficient on mortgage credit growth becomes only significant at the 10% level in some matched specifications.
  - Market share of independents, 2005 on unemployment change (05-08):
    - Full sample: 1.900***  (t-statistic (2.90))
    - Interstate matched: 2.078**  (t-statistic (2.45))
    - Intrastate matched: 1.177*  (t-statistic (1.75))
- Goodness of fit and sample sizes (Table12):
  - Full-sample regressions N: 583 (foreclosure), 599 (credit), 599 (unemployment).
  - Interstate matching regressions N: 278 (foreclosure), 280 (credit), 280 (unemployment).
  - Intrastate matching regressions N: 162 (foreclosure), 162 (credit), 162 (unemployment).
  - adj. R2 examples: full-sample foreclosure 0.504, interstate matched foreclosure 0.572, intrastate matched foreclosure 0.414.

### Findings from matched-subsample exercise (implications)
- When restricting analysis to characteristically similar counties (matched samples), the key positive association between market share of independents and the rise in foreclosures becomes stronger.
- The intrastate matched results yield the largest estimated effect of independent market share on foreclosures.
- Coefficients on mortgage credit and unemployment are broadly similar in magnitude across matches, with some differences in statistical significance (e.g., mortgage credit growth coefficient sometimes significant only at the 10% level).

### Discussion: lending standards, independents vs. banks (Section E)
- Empirical patterns:
  - Independents contributed disproportionately to the lending boom and expanded market share more in areas with higher percentage of low credit score consumers and higher house price growth.
  - Counties where a higher share of mortgage lending was channeled by independents experienced a sharper rise in foreclosures, even after controlling for county characteristics.
- Interpretation:
  - Findings strongly indicate that the expansion of independents came at the expense of a significant deterioration in lending standards.
  - Possible mechanisms (not empirically disentangled in this paper):
    - Independents lent to a riskier category of borrowers.
    - Independents expended less effort in collecting soft information.
    - Independents designed riskier contracts (possibly more attractive to less risk-averse borrowers).
  - The correlation between presence of independents and county characteristics is imperfect; many high-price-growth, lower-income, lower-credit-score counties were still bank-dominated as of 2005.
- Limitations and avenues for future research:
  - Exploring risks associated with independents’ lending more directly is beyond the scope of this paper and would benefit from disaggregated data.
  - County-level findings establish quantitatively important correlations at the aggregate level.

### Exploring the role of regulation and alternative hypotheses (Section V)
- Hypothesis: weak regulatory oversight of independents explains their contribution to the boom-bust cycle.
- Approach:
  - Tested alternative hypotheses and exploited variation in mortgage-related regulations across states.
  - Controlled for local market competition using Herfindahl indexes (top 15, 30, 50 lenders) and for lender geographical diversification (per Loutskina and Strahan (2011)).
  - Examined deposit-taking heterogeneity by computing county shares of loans originated by banks with core deposits-to-assets ratios above median (0.51) and upper quartile (0.61), assigning independents a core deposits ratio equal to zero for comparison.
  - Controlled for securitization measures: private securitization share, share sold to GSEs, and share kept on originator balance sheet.
- Alternative hypotheses results (Table13 and related text):
  - Herfindahl indexes (top 15, 30, 50) are far from significant; coefficient on independents remains positive and significant and often larger as sample size shrinks.
  - Geographic diversification index has virtually no impact on the coefficient of independents; estimated coefficient on diversification is small and not significant.
  - Deposit-taking heterogeneity:
    - Median core deposits-to-assets ratio in sample banks (as of 2005) is 0.51.
    - Many banks have ratios lower than 0.2; independents have ratio equal to zero by definition.
    - Regressions controlling for shares of loans from banks with CD>0.51 and CD>0.61 suggest the relation is unlikely driven by deposit-taking differences.
  - Securitization:
    - Controlling for private securitization explains at best a small fraction of the effect of independents.
    - Controlling for Percent sold to GSE yields a negative (not significant) coefficient on Percent sold to GSE and slightly reduces coefficient on independents; implication: GSE securitization is negatively correlated with rise in foreclosures (GSEs require minimum standards).
    - Controlling for Percent not securitized yields a positive and not significant coefficient.
- State regulation analysis (Section V.B and Table14):
  - Two datasets on state regulation used:
    - Bostic and others (2008) — state restrictions on predatory lending (anti-predatory lending laws).
    - Pahl (2007) — state regulation of mortgage brokers.
  - Constructed dummies for upper quartile (most regulated states) for both broker regulation and anti-predatory laws.
  - Interacted market share of independents (2005) with these regulation dummies.
  - Results (Table14):
    - States with high broker regulation dummy coefficient: negative (e.g., -0.0596*** in column (1)) indicating smaller average increase in foreclosures in those states.
    - Interaction Independents×broker: -0.956*** (column (2)), implying the effect of independents on foreclosures is weaker in states with high broker regulation.
    - When clustering errors at the state level the interaction remains significant at the 5% level.
    - States with high anti-predatory lending laws: coefficient alone negative but not significant; interaction Independents×anti-predatory: -0.660*** (column (5)) and -0.660* when clustered at state level (column (6)), implying a weaker independents effect in states with stronger anti-predatory laws (significance smaller than broker interaction).
  - Controls included state per capita GDP (2002) and foreclosure-related law dummies (judicial foreclosure, statutory right of redemption, deficiency judgment rules).
  - Interpretation:
    - Results suggest the association between independent lending share and rise in foreclosures is less pronounced in states that implemented stricter mortgage-related regulations during the boom.
    - Regulation variables are imperfect proxies for effective regulatory enforcement, but findings are consistent with the regulatory explanation.

### Conclusion (Section VI)
- Main conclusions:
  - Lightly regulated independent lenders contributed disproportionately to the recent boom-bust housing cycle.
  - The mortgage boom was fueled to a large extent by a fast expansion of credit from independent lenders.
  - Market share of independents as of 2005 is a strong predictor of the increase in foreclosure rates between 2005 and 2007.
  - Robust to parametric and semi-nonparametric controls and matching methods.
  - Robustness tests suggest the strong association is most likely due to weak regulatory structure governing independents.
  - Presence of independents also predicts contraction in credit and house prices and the subsequent rise in unemployment between 2007 and 2008.
- Policy implication:
  - Findings lend support to the view that more stringent regulation could have averted some of the volatility in the housing market during the recent boom-bust episode.
- Suggested future work:
  - Use disaggregated data to identify characteristics that made independent lending riskier than bank lending.

*Source: _wp11215 - 3.  OLS on the matched subsample (PDF chapter/section).*

### 2005. For comparison we plot the 45 degree line to underline the upward shift.

### _wp11215 - 2005. For comparison we plot the 45 degree line to underline the upward shift.

### Foreclosure trends (Figure 4)
- Foreclosures in percentage of originated mortgage credit plotted over time with axis values shown as: .2, .4, .6, .8, 1.
- Years plotted: 1985, 1990, 1995, 2000, 2005, 2010.
- Notes: "This figure shows the evolution over time of foreclosure filings in percentage of originated mortgages. Source: U.S. Department of Housing and Urban Development (HUD)."

### Relationship between independents’ market share and foreclosures (Figure 5)
- X-axis label shown as: 0.2.4.6 (Independents’ share, 2005).
- Y-axis markers shown as: −.5, 0, .5, 1, 1.5, 2.
- Dependent variable: "Change in foreclosures: 05Q2−07Q2".
- Fitted values plotted against independents’ share.
- Key finding stated in the figure note: "We find a positive and significant coefficient."

### Mortgage credit boom and bust (Figure 6)
- Bars: "Log Total Loan Origination" plotted for years 2003, 2004, 2005, 2006, 2007, 2008.
- Line: "FHFA House Price Index" with numeric markers shown as: 170, 180, 190, 200, 210, 220.
- Secondary axis labels for house-price-related scale shown as: 14, 14.5, 15, 15.5.
- Notes: "This figure plots the logarithm of total mortgage credit in our sample (bars) and an index of house prices in the U.S. (line, yearly average of quarterly data). Source: Home Mortgage Disclosure Act (HMDA) data (our sample, see Data Appendix) and U.S. Federal Housing Finance Agency (FHFA)."

### Credit, house prices, and unemployment versus independents’ share (Figure 7)
- Three scatter panels against "Independents’ share, 2005" with X-axis label: 0.2.4.6 in each panel.
- Left panel (Credit growth: 2005 − 2008) Y-axis markers: −2.5, −2, −1.5, −1, −.5, 0.
- Middle panel (House price growth: 2005 − 2008) Y-axis markers: −.4, −.2, 0, .2, .4.
- Right panel (Change in unemployment: 2005 − 2008) Y-axis markers: −5, 0, 5.
- Notes: "This figure shows a scatter of mortgage credit growth between 2005 and 2008 (left) house price growth between 2005 and 2008 (middle) and change in unemployment rate between 2005 and 2008 (right), against the market share of independents as of 2005."

### Substantive findings and interpretation (from figure notes and labels)
- Foreclosure filings, measured as a percentage of originated mortgages, exhibit an upward shift over the sample period shown.
- A positive and significant relationship is observed between the market share of independents in 2005 and the change in foreclosure filing rate from 05Q2 to 07Q2.
- The period 2003–2008 shows concurrent dynamics in total mortgage credit (log total loan origination) and the FHFA House Price Index, indicating a mortgage credit boom and subsequent bust in that window.
- Cross-sectional scatter evidence links higher independents’ share (2005) with patterns in:
  - Credit growth (2005–2008)
  - House price growth (2005–2008)
  - Change in unemployment (2005–2008)

### Data sources cited in the figures
- U.S. Department of Housing and Urban Development (HUD).
- Home Mortgage Disclosure Act (HMDA) data (sample, see Data Appendix).
- U.S. Federal Housing Finance Agency (FHFA).

*Source: _wp11215 - 2005. For comparison we plot the 45 degree line to underline the upward shift.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11215.pdf_