Working to find actionable solutions to the nation's key challenges.

Two Sides of the Same Coin: Measuring Credit Availability and Credit Risk in the Housing Finance System

By Michael A. Stegman

Friday, July 22, 2016

Share on FacebookTweet about this on TwitterPin on PinterestShare on TumblrEmail this to someonePrint this page

Michael Stegman, who served as the Obama administration’s point person on housing finance and affordable housing policy, has joined the BPC as a fellow focused on housing policy. This post is the first in our Housing America Series, aimed at elevating the need for housing policy reforms—from housing finance to affordability—as we move through a pivotal federal election and transition. The series will also provide an update of the policy landscape since the release of the BPC Housing Commission’s well-received 2013 report, Housing America’s Future: New Directions for National Policy.

Let us start with a simple proposition: credit availability and default risk are opposite sides of the same coin. The amount of credit risk that the mortgage finance system takes on rises as credit becomes more widely available, while the reverse is true as credit tightens; both measures always move in the same direction. In anticipation of the next round of congressional debate on housing finance reform, including the contours of the applicable credit box, stakeholders of varying ideological persuasions must agree on ways to measure how well the existing system is making credit available and how well it should in a reformed system.

Two indexes—the American Enterprise Institute’s (AEI) National Mortgage Risk Index (NMRI) and the Urban Institute’s Housing Credit Availability Index (HCAI)—were developed to measure credit risk and credit availability, respectively. Both measure changes in the amount of credit risk a particular channel or the entire housing finance system holds as new loans are originated. Because they measure mortgage activities over time, each allows a comparison of current credit conditions with previous periods. Importantly, AEI refers to its measure as a “risk” index, while Urban refers to its metric as a “credit availability” index.

Despite underlying similarities, the connotation or subtle signals to stakeholders that each measure conveys are not trivial. As lenders begin to open the credit box relative to a benchmark period before the housing crisis, AEI’s metric signals that systemic risk is on the rise and, according to some observers, putting the taxpayers in greater jeopardy. Conversely, under the same circumstances, Urban’s metric would indicate that credit is becoming more widely available, which to many stakeholders would be viewed as a good thing. These metrics and their interpretation have important implications for the debate over a reformed housing finance system.

While there are several measures out there, AEI’s International Center on Housing Risk and Urban Institute’s Housing Finance Center have developed two of the more interesting measures and merit our attention.

Salience to the Housing Finance Reform Debate

A pillar of any serious proposal for housing finance reform is that affordable mortgage credit be widely available to all qualified borrowers. If the federal government is to guarantee mortgage-backed securities in a reformed system, the reasoning goes, the lower borrowing costs that the guaranty makes possible should not just accrue to those who could get a mortgage without a government backstop. The raison d’etre for continuation of a government backstop, then, is to ensure that the reformed system provides affordable mortgage credit across the eligible credit quality continuum. This reality requires two metrics: a measure of credit availability in as close to real time as possible, and an agreement on the four corners of a credit box within which the federal government guarantee would apply.

At some point in the continuing journey to a successful bipartisan legislative outcome, Democrats and Republicans will have to agree on many attributes of a reformed housing finance system. Yet, in my view, few are as important as those that would ensure that the reformed system will provide at least as much access and affordability as does the GSE system it replaces.

A pillar of any serious proposal for housing finance reform is that affordable mortgage credit be widely available to all qualified borrowers.

While the housing finance reform bill crafted by then-Senate Banking Committee Chairman Tim Johnson and Ranking Member Mike Crapo (R-ID) garnered bipartisan support in a 2014 committee vote, some senators felt the bill did not do enough to promote broad access, affordability, and a duty to serve. However, more recent legislative proposals are at least conceptually committed to remedying this issue. Beyond broad principles, serious proposals will eventually have to define the risk characteristics of the credit box that lawmakers determine will characterize the new system. This requires defining acceptable expected and stress default risks for individual mortgages, mortgage pools, vintage books, or entire portfolios that receive a government backstop on the mortgage securities the loans collateralize.

While philosophically committed to ensuring broad access and affordability, the Johnson-Crapo bill fell short of convincing some senators that it offered a workable path forward on ensuring broad access and affordability. In establishing its credit box, the legislation defined an “eligible” mortgage as one that meets all of the requirements of a Qualified Mortgage (QM) as promulgated by the Consumer Financial Protection Bureau. Yet the QM definition does not directly address credit risk. In addition, the definition’s “GSE patch” (allowing borrowers’ debt-to-income (DTI) levels to exceed 43 percent in the presence of compensating factors) is temporary in nature—the patch disappears in 2021 or when the GSEs end conservatorship, whichever comes first. At that time, the applicable DTI limit returns to a hard 43 percent without regard to compensating factors. There is no record on whether this is more restrictive than the bill’s authors might have had in mind when they defined an eligible mortgage in the reformed system as a QM loan.

Digging into the Metrics

The published views on housing finance reform expressed by experts at AEI and those releasing their work under the Urban Institute’s banner have naturally built themselves constituencies on the center-right and center-left of the political spectrum. AEI’s experts largely espouse free-market views about limiting government’s role in backstopping private sector losses in a reformed housing finance system.1 Comparatively, Urban’s housing experts have written extensively on why continuing the government guarantee of qualified mortgage-backed securities is an essential element of GSE reform, and that pre-crisis levels are a reasonable standard of credit availability the mortgage system should provide. If center-right stakeholders in the housing finance reform debate are more likely to cite AEI’s approach regarding the measurement of mortgage risk when evaluating alternative proposals, while progressive stakeholders are more likely to rely upon Urban’s metrics, the more we understand their similarities and differences, the better.

Let us dig a little deeper into the inner workings of these measures. To begin with, both measures assume that the default risk of loans originated at any given time will mirror the historical performance of loans with the same risk factors originated during specified economic conditions. For AEI’s NMRI, the expected default risk of newly originated loans is benchmarked to the performance of loans with the same risk factors as those originated in 2007, the start of the Great Recession. Thus, for example, an NMRI index value of say, 7.0 for newly originated government-guaranteed mortgages means that 7 percent of those loans would default in the event of another housing market collapse equal in severity to that which characterized the recent financial crisis. While not reflected in the index value, the default risk associated with that same bundle of loans in benign economic times would be significantly lower than the NMRI value for those loans.

The architects of the HCAI concluded that there is about a one-in-ten chance that the economy would experience a bout of severe housing market distress.

While Urban also benchmarks loan performance against historically-defined economic conditions, the HCAI index value is a weighted average of the expected default risk of a loan or cohort of loans during a period of normal economic conditions and during severe economic decline; that is, over a full business cycle. Here, expected loan performance in good times is benchmarked against loans with identical risk factors that were originated in 2001-2002, while the stress scenario benchmarks performance with those same risk factors for loans that were originated during 2005-2006.

Citing research that there have been 19 full economic cycles for the 100-year period from 1913-2013, with just two of those cycles causing severe housing market collapses (the Great Depression and the Great Recession), the architects of the HCAI concluded that there is about a one-in-ten chance that the economy would experience a bout of severe housing market distress. Thus, the HCAI assigns a weight of 90 percent to the default risk associated with a bundle of loans originated during good times, and a 10-percent weight to the higher default profile of loans originated during a stress scenario, with the index value being a blended value of the two factors.2

The HCAI measures default risk (or credit availability) as a function of expected loan performance during both good times and bad, while the NMRI benchmarks performance only during a stress scenario. Therefore, for the same cohort of mortgage loans, HCAI will always produce a significantly lower index value than that generated by the NMRI. To illustrate how these differences in methodology play out, see these recent quotes from each think tank:

From AEI: “GSE Purchase 30 year fixed from November 2012 through May 2016, the NMRI rose 1.6 percentage points from 4.9 percent to 6.5 percent.”

From Urban: “From Q2 2011 to Q4 2015, the total risk taken by the GSE [purchase] channel increased from 1.4 percent to 2.1 percent.”

While the times of the two releases are not perfectly aligned, the difference in the absolute values of the HCAI and NMRI is not just a function of timing differences, but is largely due to the low probability that Urban assigns to the likelihood of another financial crisis.


It is important for stakeholders to appreciate that these seemingly different indexes are essentially measuring the same phenomena. Since both indexes are being cited as a basis for opining on the current state of the mortgage market and to influence policy, it is also critical to understand the causes for the differences in their values. It is not a stretch to expect that various stakeholders will use both in future debates to buttress recommendations as to whom a reformed housing system should serve and the level of risk that taxpayers should assume.

In fact, AEI researchers contend that had their index been available before the crisis, the NMRI could have alerted stakeholders of the looming dangers of rising credit risk building up in the system long before 2007 when the dam broke. In addition, in the thick of the Johnson-Crapo debate, citing NMRI data, they argued that the QM standard was “too broad and deep to support a stable credit market.” For the good of borrowers and for financial stability, they consistently argue that government guarantees should be limited to low-risk loans, which they define as having an average portfolio stressed default rate of no more than 3 percent, with a 6 percent maximum stressed default rate for the last loan-in.

While to our knowledge, Urban Institute researchers have not used the HCAI to opine on the contours of a future system, they regularly use the index to monitor monthly changes in credit conditions in the current system. It is not a far reach to use the index more prescriptively to define the credit box in a successor system to the GSEs. So, if mortgage credit should be as widely available as it was during Urban’s pre-crisis baseline period, then the average HCAI value during 2001-2002 can be used to set the desired risk tolerance for the reformed system. If policymakers want to benchmark credit availability in a reformed system to the profile of the GSEs’ more recent, post-crisis credit books in conservatorship, they would define the eligible credit box more tightly. This is not a trivial matter. According to Urban, in 2001-2002, GSE credit was about three times more widely available then it was in 2014-2015.3

The dramatic differences in who the GSEs have served over time is also a reminder to those who advocate for releasing the GSEs from conservatorship because of their broad credit box, be careful what you wish for because the credit box can quickly narrow.

Finally, in practical terms, credit availability is also dependent upon its cost to the end user. A mortgage term sheet that purports to show that credit is available to borrowers at the outer edges of the eligible credit box, but assesses a substantial surcharge on those borrowers that essentially prices them out of the market, is not really demonstrating access to credit in any real sense. This is where levels of acceptable credit risk, capital requirements, and required rates of return interact to define the effective credit box, but that is for a future discussion.

1 The Urban Institute publishes essays from unaffiliated guest authors. This post is referring to researchers employed by Urban.
2 In measuring historical loan performance, both AEI and Urban developed lookup tables or matrices of loans divided into hundreds of different “risk buckets,” defined in AEI’s case as credit score (FICO), combined loan-to-value ratio and debt-to-income ratio (DTI), and in Urban’s case as FICO, LTV, DTI, and product type, for both government-guaranteed loans and non-agency loans, including loans held on bank balance sheets. While the HCAI is only for purchase first mortgage loans, the NMRI database is limited to government-guaranteed loans, but includes home purchase and refinance, although the NMRI and HCAI data cited are limited to the GSE purchase channel.
3 In the eight quarters spanning 2001-2002, the HCAI for GSE home purchase loans ranged between 5.1 and 5.8. In the eight quarters spanning 2014-2015, the HCAI ranged from 1.9-2.1 (data provided to author by Urban Institute).