The Federal Reserve Bank of New York recently released a paper that looked at the impact of HARP revisions on loan defaults and pricing.  The paper, Payment Changes and Default Risk: the Impact of Refinancing on Expected Credit Losses was written by Joseph Tracy and Joshua Wright. 

When the Home Affordable Refinance Program (HARP) was initiated, its goal to stimulate the economy and reducing defaults by lowering mortgage payments in households with high loan-to-value mortgages. These were borrowers who were otherwise unable to refinance.

HARP was implemented in 2009 but refinancing activity was much lower than expected.  Just over one million refinances have been done under HARP rather than the 3 to 4 million expected.  This confirms, the authors say, the original rationale for HARP, that in the wake of the housing bust borrowers need help refinancing.

HARPS lackluster results have provoked discussion about the impediments to refinancing including credit risk fees, limited lender capacity, a costly and time consuming appraisal process, limitations on marketing, and legal risks for lenders.  HARP was recently revised to better address these impediments. .

Concerns about revising HARP included doubts about its fairness and about macroeconomic efficiency.  The Federal Housing Finance Agency (FHFA) has a responsibility to weigh the value of any proposed changes in terms of a possible impact on the capital of the government sponsored enterprises (GSEs).  These could include a reduction in the income generated through interest on the GSE's Holdings of MBS, on the expected revenues from the put-backs of guaranteed mortgages that default, and finally on the impact of refinancing on expected credit losses to the GSE fees. 

One outcome of an improved program would be more borrowers in a position to refinance.  Estimates can be made of the average reduction in monthly mortgage payments that would result from a refinance; the question is how this payment reduction would affect future defaults.  Ideally a study could determine the difference in expected credit losses from two identical borrowers with identical mortgages where one borrower refinances and the other does not.  However, once the existing mortgage is refinanced it disappears so both mortgage/borrower sets cannot be similarly tracked and the impact of the payment change on a borrower's performance must be inferred.

A recent congressional budget office working paper estimates that reduced credit losses would produce an incremental 2.9 million refinances of agency and FHA mortgages and that such a program would reduce expected foreclosures by 111,000 or 38 per 1000 refinances, reducing credit losses by $3.9 billion.

Using data from Lender Processing Services the authors selected eligible borrowers from among borrowers who had been current on mortgage payments for at least 12 months and had estimated loan-to-value ratios (LTV) over 80 percent. To measure motivation the authors selected loans where the borrower could recover refinancing costs in two years.  Using these parameters it was determined that refinancing would reduce the required monthly payment by 26 percent on average.

The authors found impacts on results from various combinations of local factors such as house prices, employment rates, the local legal methods of handling delinquencies, and contagion risk, i.e. the exposure of the borrower to others who had defaulted.  There were also effects from FICO scores and debt to income ratios and loan specific factors such as the purpose of the loan, full documentation of the loan, and length of loan term.  Various methods were used to control for these variables including excluding loans from the sample.

It is acknowledged that LTV ratios have a significant correlation with default and the authors did test and confirm this relationship.   The next step was to estimate the impact of a 26 percent payment reduction on the average default rate.  The authors used estimated ARM default and prepayment hazards to do a five-year cumulative default forecast holding the local employment rate and home prices stable.  At five years the models imply that the expected cumulative default rate would be 17.3 percent. When the payments are reduced by 26 percent the expected default rate is reduced to 13 percent a 24.8 percent reduction.   The same analysis was run for borrowers with prime conforming fixed-rate mortgages obtaining a cumulative default rate of 15.2 percent which refinancing reduced to 11.4 percent, a decline of 3.8 percentage points. 

These figures were used to conduct a simple pricing exercise to measure the difference between a refinancing fee that maximizes fee income for a certain category of borrowers and a fee that maximizes the combination of the fee income and the reduction in the expected future credit losses using the GSE pricing categories for their loan level adjustments.  It was found that FICO score strongly impacted both payment reductions and default rates ultimately resulting in reductions in the default rate of 1.9 percentage points for a high FICO borrower and 9.1 points for a low one.  This implies that incorporating the impact of expected credit losses into the pricing decision should generate higher price discounts for weaker credit borrowers as measured by FICO score and LTV.

The authors found that incorporating the impact of expected credit losses after refinancing on average lowed the desired pricing by 17 basis points.  Looking at the averages by LTV intervals shows an impact of 15 basis points for mortgages with a current LTV of 80 to 85 and increases to 14 basis points for mortgages with a current LTV of 105 or higher.  Basing fees on FICO scores involves a much more complicated set of factors. 

The authors conclude that the average HARP refinance would result in an estimated 3.8 percent lower default rate.  Assuming a conservative average loss-given default of 35.2 this indicates an expected reduction in future credit losses of 134 basis points of a refinanced loan's balance.

The paper concludes that the impact of refinancing on future default risk is important to the current debate of the GSE fee structure for HARP loans.  "These results suggest that refinancing can be fruitfully employed as a tool for loss mitigation by investors and lenders.  The optimal refinance fee will be lower if this reduction in credit losses is recognized."  Reducing fees, the authors say, will increase incentives to refinance but at the cost of fee income to the GSEs.  "Our analysis shows, however, that there is an offset to this lower fee income today which is lower credit losses in the future."