In previous blogs we have discussed appropriate debt levels for nonfinancial companies and for large banks.  What about for the individual household?  Sometimes, households are ‘credit-constrained’ in that they cannot find borrowers to lend them their desired amounts.  In economic booms, like the housing boom that preceded the 2007-2008 crisis, lending standards tend to soften and credit is more widely available.  In that situation, where there are little or no credit constraints, how much should the individual household borrow?

The Federal Reserve’s Flow of Funds database provides detailed information on the aggregate Household Balance Sheet.  As of the latest reading (March 2016), total assets held by households were $101 trillion and total debt was $14 trillion, leaving a cool $87 trillion in net worth.  Thus, household leverage is about 16% ($14 trillion/$87 trillion), quite a bit lower than leverage for non-financial companies and massively lower than typical leverage for financial institutions.  So, in contrast to comments often heard about the “over-leveraged” household sector, maybe households should be taking on more debt.

One problem with the Flow of Funds database is that it is purely an aggregate; from it you can get total assets and debt, and average assets and debt, but you cannot learn anything about the distribution of assets and debt across households.  Fortunately, for this type of information you can go to another Federal Reserve database – the triennial Survey of Consumer Finances (SCF).  The SCF is based on a detailed survey of more than six thousand households and provides data on income, assets, debts, wealth, savings behavior and a host of other financial variables along with information on age, income, gender, education of the head of household.  Thus, with this data you can get an idea of the entire distribution of income or wealth or debt, broken down if desired by age, education, etc.  This detailed data set shows that the typical or median household is a lot more leveraged than the average household.  The median family has a net worth of about $80,000 and combined mortgage and consumer debt of about $150,000.

Is this level of debt too much?  Well, it depends on what the debt is for.  Debt can be highly valuable in terms of financing the acquisition of an asset like a home or a college degree, or it can be useful in allowing “smoothing” of consumption (generally, younger people can look forward to rising income over time, so it may make sense to augment income when young to finance a desired consumption pattern).

Still, even if taking on debt can be a smart strategy, there can always be too much of a good thing.  You have too much debt if you do not have a viable strategy for paying it off, or if it leaves you highly susceptible to an economic downturn.

Weakened underwriting standards and optimal borrowing

Mortgage standards weakened dramatically in the years preceding the financial crisis.  This enabled increased demand for homes and thereby accelerated the housing price boom.  Eventually, the boom ended and housing prices fell dramatically.  Some of the housing price decline was surely due to a decline in housing demand as underwriting standards tightened significantly.  Another factor was generalized economic weakness in the aftermath of the financial crisis of 2007-2009.

There were many reasons for the weaker underwriting standards preceding the crisis.  One was the direction of public policy which strongly supported expansion of the rate of homeownership, particularly among minorities.  Another was strong credit performance – in part due to rising housing prices – that led to the assumption that underwriting standards were too strict.  A third reason, perhaps, is the ascension of the “mortgage banking” or “originate to sell” model of mortgage origination, under which the lender is not the final investor in the loan.  Instead, the lender underwrites and funds a large number of loans that are subsequently combined into a mortgage backed security (MBS) that is sold to investors.  In this model, the underwriting standards are set by the ultimate mortgage investor or guarantor, not the lender.  The lender’s job is to originate loans in accordance with these standards.

In retrospect, many people agree that mortgage underwriting standards were excessively loose during the mid-2000s, and financial reform legislation includes efforts to toughen underwriting standards and make sure lenders are on the hook for bad loans.  One means of doing this is to force lenders to retain credit on loans that are sold to investors.  Another is to impose a standard of “suitability” on lenders, analogous to that applied to securities brokers.  In the securities world, brokers have a legal obligation to put their clients into investments that are “suitable” for them (investment advisors, as opposed to brokers, have a higher legal standard).

Jack Guttentag is a long-time observer of the mortgage markets and is the author of the popular “Mortgage Professor” website.  In fact, he is the Mortgage Professor (MP).  The MP argues against imposing a standard of suitability on mortgage lenders because it would make lenders responsible for things that they cannot control.  There are many potentially legitimate motives for taking out a loan, including such wide-ranging goals as reducing cash outflows so as to pay down other debts, or reducing cash outflows so as to enable alternative investments, or reducing the minimum payment so as to be consistent with volatile income, and so on.  There is no way the lender will have sufficient information on the borrower to determine if these motives are valid or not.  At the end of the day, the borrower has to make this judgment.

In addition to credit retention and imposing a suitability standard, a third way to reign in risky lending is by requiring more and better disclosure.  Better disclosure surely is desirable, but there already is massive disclosure, much of which is ignored.  Additional disclosure that the MP favors is a clear statement of the maximum payment that is possible under the terms of the loan and the nearest period that this maximum payment could occur.  This would be helpful in addressing cases in which the initial payment is very low and the borrower is led to believe it would always remain low.

Weaker underwriting standards increase the probability of failure, forced sale or foreclosure of the property.  Some borrowers were no doubt aware of this during the run up to the crisis but proceeded anyway.  Think of the guy buying multiple “second homes” in Las Vegas or Arizona with the objective of a quick and profitable flip.  The downside risk of taking on low or zero down payment loans to buy these properties is pretty small relative to the potential upside.  In effect, the mortgage industry offered speculators a nearly free call option on housing prices.

The major loser was the property owner who had gradually built his equity over years of timely payments only to see it disappear almost overnight.

It seems to me that a useful addition to the debate would be an information service for borrowers that estimated the safety of the loan, where non-safety means a high probability of severe financial stress due to taking out the loan.

How can this calculation be made?  One answer would be the probability of default at the time of origination.  However, even with highly risky loans, this probability is generally pretty small.  After all, even if the borrower has difficulty making payments, so long as housing prices drift higher over time, the borrower can refinance or sell.  For example, during periods of rising housing prices, sub-prime loans have default rates in the mid-single digits.  However, during periods of falling housing prices, sub-prime loans may experience failure rates of 50% or more.

I think a better answer would be the probability of default in the event of a moderately severe economic downturn.  Good proxies for economic conditions would be local unemployment rates and local housing prices.  All we need is for some econometrician (i.e., an economist with data analysis skills) to build a statistical model relating probability of default to unemployment rates and housing prices, given a set of underwriting criteria.  Any takers?