The non-partisan Congressional Budget Office (CBO) just released a working paper entitled An Evaluation of Large-Scale Mortgage Refinancing Programs in which it analyzed a large-scale program that would relax current income and loan-to-value restrictions for borrowers who wish to refinance mortgages currently held by Fannie Mae, Freddie Mac, or the Federal Housing Administration. The analysis uses an estimate of the increased refinancing that would occur and how future default and prepayment behavior would be affected by such refinancing.   

The paper was written by Mitchell Remy and Damien Moore, both of the Financial Analysis Division of the CBO and Deborah Lucas, Sloan School of Management at the Massachusetts Institute of Technology.

Despite the gradually recovering economy, the paper says, the housing market remains weak and, in addition to the large number of mortgage delinquencies and defaults many homeowners are unable to refinance and take advantage of record low interest rates.  This has prompted a number of proposals for federal programs to provide opportunities for refinancing which would free up household income for non-housing expenditures as well as help some homeowners avoid future default.

The lack of refinancing is driven by weakened household balance sheets that make it difficult for some borrowers to meet debt-to-income (DTI) ratios; other homeowners find their equity eroded by falling house prices and cannot qualify for refinancing based on the loan-to-value (LTV) ratios of their homes.  The increased stringency of underwriting requirements on the part of the government sponsored enterprises (GSEs) and FHA have also contributed to the inability of others to qualify for refinancing. 

The government has launched several efforts to assist such homeowners but program features and eligibility criteria have excluded a significant number of homeowners who might have benefited.  These efforts have also been restricted to homeowners with mortgages insured by the GSEs or other Federal programs representing 56 percent of all outstanding loans.

But, while mass refinancing of these loans might save the government some eventual losses because of future defaults, mortgage investors would experience losses through loans that are prepaid more quickly than without such programs.  Many of these losses would hit taxpayers through mortgage-backed securities held by the Federal Reserve, the GSEs and the Treasury.  The stylized refinancing program studied by the authors would increase the availability of refinancing and slightly lower its cost. 

Homeowners are generally free to refinance at will, weighing positive factors such as a lower rate or better terms against the financial costs of refinancing such as appraisal costs, settlement fees, and the time and effort involved in the process.  That freedom, however, comes at a cost as investors evaluate prepayment risk and factor that into the price of their loans.

Over the years refinancing has periodically affected large percentages of the outstanding mortgage debt, particularly in periods of declining rates and several changes in the market such as automated underwriting and property-valuation models, the widespread use of credit scores, auto have increased the likelihood homeowners will refinance. 

However, unlike similar periods of declining rates, the fact that rates have dropped more than 1.5 percentage points since 2007 has not sparked a refinancing wave.  This is largely due to the financial crisis.  First, a significant number of mortgages were actually originated during this period of extraordinarily low rates and earlier transactions have more or less drained the refinancing pool.  Then there are the widely acknowledged impediments to refinancing the remaining high interest loans - home price declines, reduced income or unemployment, continued high levels of consumer debt, and significantly higher underwriting standards.  Many loan products such as interest only, no documentation, and negatively amortizing loans have been curtailed or have disappeared.

These impediments to refinancing are reflected in the unusually high premium over face value that investors are willing to pay for MBS with high coupon rates.  The possibility of prepayment usually prevents MBS from rising in price much above par value but the current pricing indicates that investors expect these impediments to last and thus their investment to continue considerably longer than usual.  In January of 2011, weekly pricing for premium ($100 par) Fannie Mae 30-year FRM ranged from just below $105 to considerably over $107.  The authors say that, while many factors influence MBS pricing "a strong case can be made for lower prepayment expectations as an important component in the price premium investors are currently willing to pay for those securities."

The authors examined the parameters and criteria of existing refinancing programs such as the GSE's Home Affordable Refinance Program (HARP) and some programs proposed under pending legislation before laying out their design considerations for a stylized large-scale program.  These include whether to include only GSE guaranteed mortgages or those guaranteed by FHA and private-label mortgages as well.  Should only borrowers who are current on their existing loans be allowed to participate, only those who are delinquent, or both?  Should guarantee fees be increased to reduce the federal cost of the program even if it reduced the potential pool of borrowers or should the fees be reduced to generate the opposite effect?  The degree to which underwriting procedures are waived or fees are waived or subsidized could significantly affect participating rates but could result in loans that are less appealing to investors and thus increase the interest rate on the loans.   

A final consideration, they said, is the degree to which the LTV is utilized.  Limiting the level of negative equity may mitigate the severity of the loss in the case of default, a risk that the guarantor already carries and hopes to reduce through granting a lower monthly cost.  Another question is whether a current LTV is required.  Waiving an appraisal would save the borrower a fee but would deny investors an up-to-date assessment of the potential risk of default inherent in the loan.

The stylized program that emerged from these considerations had the following characteristics:

• It would start in the first quarter of 2012 and be available for one year.
• Eligible loans would be existing mortgages guaranteed by the GSEs or FHA.
• The borrow must be current and never more than 30 days late on the existing mortgage during the prior year but there are no limits on the borrowers current income or on the LTV of the new loan.
• The GSEs and FHA will assess a guarantee fee equal to that charged initially on the existing loan. This will be incorporated into the interest rate by the GSEs and charged as an annual premium by FHA.
• The loan will have a fixed rate of interest at the prevailing rate and a 30 year term.
• Lenders and third party fees will be the lesser of 1 percent or $1000 to process the loan.

The authors estimate that the potential target audience for the program is borrowers with mortgages in existing MBS with outstanding balances of $4.3 trillion as of June 1.  Of that total, the GSEs guarantee $3.5 trillion.  These are all fixed rate mortgages.  Although borrowers with adjustable rate mortgages (ARM) would also benefit from the program they, as well as non-traditional and delinquent borrowers were excluded from the analysis to eliminate certain complexities in estimating incentives.

Borrowers were classified by the interest rate on their current mortgage, the term, and risk characteristics that affect their propensity to default on their mortgages.  The groups were further characterized by the borrower FICO score, the mark-to-market LTV on their existing mortgage and its origination year and the imputed interest rate on a new mortgage.  This resulted in classification into 108 distinct groups representing the combination of two maturities, nine coupon rate pools, and three risk categories.  Total program costs are the sum of the cost per dollar of outstanding loans multiplied by the estimated total principal balances across each of the 108 groups.

Each group was run through a model to simulate a scenario and determine the incremental refinancing attributable to the program and a base case scenario to estimate expected repayments, defaults, and loss severities over each group without the program.  The analysis then presumed that all borrowers had an LTV of 50 percent, FICO scores of 780, and assumed the new mortgage had a rate reflecting market conditions and discounted transaction fees.

The difference in prepayment volumes between the base case and the program model represent the incremental refinancing attributed to the program over the course of calendar year 2012 and those rates, applied to estimated outstanding principle balances, determine the expect dollar volume of the incremental financing.

The relaxed underwriting constraints result in identical refinance rates between high, medium and low risk categories with the same MBS coupon in the program scenario.  Borrowers in the high and middle risk categories (higher LTV, lower FICO) benefit the most from the program based on the increase in refinance rates between the base case and program scenarios.  Because these groups are a small portion of the borrower population, however, the greatest number of incremental loans is expected in the low risk group where borrowers are likely to be attracted by the lower costs of the program.

A program such as that analyzed offers benefits to homeowners including lower monthly payments resulting from lower interest payments and lower principal payments because of the re-amortization of the loan.  This results in an increase in disposable income which could help the economy, assist the borrower in paying down other debt, and in some cases help prevent eventual default. 

The authors estimate that the program would result in incremental refinancing of 2.9 million mortgages with first year gross cash savings from reduced mortgage payments of roughly $2,600 per borrower or a total of $7.4 billion.

The model predicts that approximately 111,000 fewer loans will default as a result of this program, in comparison to the approximate 4 million borrowers currently past due on their mortgages.

The results have cost implications for the GSEs, FHA, the Federal Reserve, and the Treasury Department.  Because of their mix of credit guarantees and portfolio investments the GSEs will experience both gains and losses; FHA will see gains from credit guarantees and the Fed and Treasury will face losses through their investments in MBS.  Total federal losses from investment will be 4.5 billion resulting in a net federal loss from the program of $0.6 billion.

Non-government investors hold 65 percent of the outstanding MBS included in the analysis and also held a greater share of older, higher coupon securities than the Federal share.  Non-federal investors were expected to suffer a fair-value loss of $13 to $15 billion, triple the federal loss.

The authors see several impediments to program implementation.  One is the possible inability or unwillingness of lenders to scale up quickly to meet program demand, especially in light of the low origination volumes in recent years.  A second obstacle is the willingness of subordinate lien holders to make the financial or operational considerations necessary to facilitate refinancing the first mortgage and private mortgage insurers must agree to the terms of the refinancing if insurance will be needed.  The authors assume that the potential of reducing default may be attractive enough to convince these third parties to participate.  If this is not correct the number of participants in the program may be reduced significantly.  Finally, the program could be imperiled by rising interest rates during the program and could be affected either positively or negatively by the direction of housing prices which, if rising, may allow more homeowners to qualify for this or other programs and if falling may lessen the numbers who are eligible or drive more borrowers into default.