The roots of the mortgage crisis have been microscopically examined by everyone from economists to consumer groups and volumes of conten has been written on underwriting guidelines, irrational exuberance, greed, and hubris. 

Now comes a study which offers a stunningly simple explanation, not for the whole crisis, but for a contributing factor: Americans can't add and subtract

Kristopher Gerardi, Lorenz Goette, and Stephan Meier recently produced a working paper for the Federal Reserve Bank of Atlanta which examined the relationship between financial literacy and subprime mortgage delinquencies.  They found that the ability of individuals to make simple financial calculations is strongly related to the amount of time they spend being delinquent on their mortgages.

The study used loan level data collected by FirstAmerican Loan Performance on a group of non-agency mortgages written between the late 1990s and 2007 in three New England States, Massachusetts, Rhode Island, and Connecticut, and used as collateral for mortgage-backed securities purchased by investors. The data set was cross referenced with data collected by the Warren Group which has been the reporter of record for real estate data in New England for over a century. This resulted in a pool of over 70,000 mortgages.  A random sample of borrowers was pulled from this universe and divided into two groups, one of which was cold called and the other contacted by mail. Unfortunately less than 400 respondents ultimately participated in the study.

Respondents were questioned to determine their financial literacy as measured by their numerical ability and basic economic literacy as well as their general cognitive ability.  They were also questioned to derive a measure of time and risk preference and to gather an extensive list of socio-demographic characteristics and were asked for details about the terms of their mortgages (the study already had extensive details on terms) and their experience shopping for that loan.

To measure numerical ability, respondents were asked five questions.  One example:  "A secondhand car dealer is selling a car for $6,000.  This is two-thirds of what it cost new.  How much did the car cost new?"  Cognitive ability was measured by asking respondents to name as many animals as they could in 90 seconds.  Two questions were asked to measure basic economic literacy, for example, "Imagine that the interest rate on your savings account was 1% a year and inflation was 2% a year.  After one year, how much would you be able to buy with the money in this account?  More than today?  Less than today?  Exactly the same as today?"

While cognitive ability was found to have a bearing in one area of the study as explained below, numerical ability was the factor that had the highest correlation to mortgage delinquency which was taken from loan data in three categories; the fraction of time the borrower was behind on any mortgage payment; the ratio of missed mortgage payments to total payments billed, and whether the borrower had ever been in the state of foreclosure.  

Respondents were placed in one of four groups corresponding to their perceived numerical ability.  Borrowers in the group with the lowest numerical ability spent, on average, about 25 percent of the time in delinquency while those with the highest ability spent 12 percent.  The lowest group had missed 15 percent of mortgage payments on average compared to the highest group at 6 percent.  There was an almost perfect waterfall from high literacy to low in the category of time periods of delinquency; the middle two quartiles of literacy were nearly equal and fell right in the middle of the 9 percentage point of the percent of payments missed.

In the foreclosure category, the differences across numerical ability groups was also very large; approximately 18 percentage points between the group with the highest foreclosure rate and the lowest.  However, there was a deviation from the waterfall with the group with the second lowest numerical ability scoring highest on foreclosures.  The lowest distress was registered again by the highest numerical ability group.  Foreclosure was the only delinquency category in which cognitive ability also was a factor with the higher cognition correlating with lower foreclosures.  

The study's authors make the point that much of the control in foreclosures is in the hands of the lender and speculate that perhaps they are more reluctant to foreclose on more intelligent borrowers.

The authors state that the results of their study suggest that the correlation between financial literacy and mortgage delinquency is not due to financially illiterate borrowers taking on too much debt, or choosing excessively risky mortgages but that the limited numerical ability might lead to other mistakes over the course of time, like too much spending, too little savings, or inappropriate reaction to income and/or consumption shocks. 

The results also suggest that subprime borrowers with limited numerical ability were no more likely than others to enter into unfavorable contracts.  However, the authors say that the timing of the originations of respondents' mortgages relative to the emergence of the subprime crisis prohibit any solid conclusions in that area.

The study, according to its authors, has several implications for future research and applications.  "First, a normally unobservable characteristic or ability can explain part of the heterogeneity in default behavior."  This could provide insights to lenders on designing contract terms and default reduction strategies; i.e. financial institutions may have an interest in applying tests of numerical ability to loan applicants.

Second, there is a question as to whether literacy could be especially important in the environment of rapidly falling home prices.  Because the experts differed on the characteristics of the housing boom, many individuals may have relied even more heavily on the ability of rising equity to keep them in their homes.  As the market reversed, "individuals with high financial literacy may have found it easier to adjust their consumption and savings decisions in order to continue making their mortgage payments."  

The authors conclude that their results suggest that more intensive financial education could substantially improve financial decisions later in life.  However, the next logical step is to randomize financial education and then track the financial decisions of these individuals over time.