Tuesday, March 5, 2013

Deleveraging by Default

The new Household Credit and Debt Report came out last week from the New York Fed. Fun!

The stuff about student debt got the scary headlines, and with reason -- especially once you notice that the 17 percent delinquency rate on student debt, bad enough, understates the problem, since that figure includes debt on which no payment is due:
when we remove the estimated 44 percent of all borrowers for whom no payment is due or the payment is too small to offset the accrued interest, the delinquency rate rises to over 30 percent.
Student debt is not really my beat, though, so I want to call attention to something which has gotten less attention: how much household "deleveraging" is really about defaults, rather than reduced borrowing.

The Credit and Debt Report is based on the New York Fed's Consumer Credit Panel, with the underlying data from the credit bureau Equifax. It's unique, as far as I know, in its comprehensive coverage of the various flows that make up changes in household liabilities. The Flow of Funds, by contrast, sees household debt only from the creditors' side, and doesn't directly observe flows, only changes in stocks of debt. So with this data we can see much more clearly what's actually driving the fall in household debt-to-income ratios. [1] And here's what we find:

The heavy lines are the year-end ratios of mortgage and total household debt, respectively, to disposable personal income. The dotted lines are the path these ratios would have followed if defaults were held fixed at their pre 2007 levels. So we see that total household debt peaked at 119 percent of income at the end of 2007, and has since fallen to 100 percent, a substantial decline. But when we break out the various factors accounting for changes in debt -- new borrowing, repayment, and default -- we find that the fall is entirely the result of higher defaults. If households had continued defaulting on debt at the same rate after 2007 as before, household debt would not have fallen at all. (It is true that since 2009, there has been some deleveraging even net of defaults, but even over those two years two-thirds of the fall in debt-income ratios is due to elevated default rates.) Mortgage debt follows the same pattern: If default rates had continued at their 2003-2006 level, mortgage debt would have been greater, relative to income, at the end of 2011 than at the end of 2007.

It's interesting to compare the debt writeoffs reported by households with the writeoffs reported by commercial banks. The biggest difference between the two series is that banks report their net losses, i.e. after recoveries. But both show the same dramatic rise in the Great Recession.

As we can see, the two series move more or less together. It's noteworthy, though, that before the crisis the amount of debt discharged by default was consistently about five times greater than banks' default losses; after 2007, this ratio dropped to more like like two to one. This represents some mix of lower recovery rates -- underwater homes are worth less than their mortgages -- and a worse default performance among mortgages owned by entities other than commercial banks.

So why does all this matter? Well, the obvious reason is that we want to get the story of the recent past right. The usual debate about falling debt is how much it's due to banks' unwillingness to lend, and how much to households' unwillingness to borrow. If it's really due largely to higher default rates, our stories of the financial crisis and its aftermath should reflect that. But they seldom do. Richard Koo, just to pick one example at random, treats changes in household liabilities as simply a measure of household borrowing.

A couple other reasons to care. For one, the role of defaults is further evidence against the idea that demand is being constrained by a lack of access to credit. 

More broadly, it's evident that the relationship between defaults and changes in income is nonlinear. Over a normal business cycle, household defaults are stable and fairly low. (This is not true of business and especially commercial real-estate defaults.) It takes an exceptionally deep fall in income to produce a noticeable rise in household defaults. The macroeconomic significance of this is that defaults, like Koo-style deleveraging, weaken the link between current income and current expenditure; in both cases, a higher share of changes in income show up as changes in the flow of payments to creditors, rather than changes in spending on currently produced goods and services. This dampening of the income-expenditure link helps put a floor under demand fluctuations, as discussed in the previous post (provided that defaults don't limit other units' access to credit -- this is an important difference  between the recent crisis and 1929-1933.) But by the same token it also weakens demand dynamics in the recovery; if a major margin on which households adjust to changes in incomes is changes in payments to creditors, rising incomes will do less to raise demand for current output.

The central importance of defaults in the deleveraging process to date also is a reminder of the importance of the terms on which debt can be discharged. Laws and norms that make default relatively easy can evidently serve as an escape valve that helps prevent the debt deflation process from taking hold.

Looking forward, this is further evidence of how difficult it is to reduce leverage just through lower expenditure. It's noteworthy here that since 2007, the household sector has had large primary surpluses (i.e. new borrowing is less than interest payments), but in the current environment of slow growth, relatively high real interest rates, and low inflation, this has not been sufficient on its own to produce any fall in leverage. So if lower debt-income ratios are a precondition for sustained growth, more systematic debt writedowns may be necessary. From the conclusion of Arjun's and my paper:
A recent IMF staff report (Gottschalk et al., 2010) notes that for public sector debt, defaults are most likely to lead a long-term improvement in the fiscal position (and have generally occurred historically) in countries with small primary deficits, or primary surpluses. In such cases unsustainable debt growth is driven by the interaction of high effective interest rates with a large existing debt stock; a one-time reduction in the debt stock can change an unsustainable path to a sustainable one, even if the interest rates on new borrowing rise as a result. A similar logic might apply to private sector debt. If so, some form of systematic debt forgiveness may be the logical, and eventually unavoidable, solution to the problem of excessive household leverage.
Finally, the importance of defaults over the past five years is a reminder that a crisis is precisely a situation when inconsistent expectations cannot be ignored. By definition, in a crisis not all contractual commitments can be fulfilled, and it's always ultimately a political question which are honored and which are not.

[1] The published report doesn't include writeoffs, only the fraction of debt that is currently delinquent. To get annual household debt writeoffs, we have to combine the report with the numbers reported by the New York Fed in its Liberty Street blog.


  1. This post and the previous one explain to me what you do mean as a "corridor" of different equilibria. However I have two problems with this:

    1) Why do you call the different leves at which employment may stabilize "equilibria"? It seems to me that a term like "employment inertia" would be better, because you speak of no feedback effect that keep the employment level on a certain equilibrium other than expectations.

    2) If I understand correctly this post, you imply that demand depends mostly on income spent on immediate consumption goods, and so when part of the income goes to pay interest of loans (or is diverted from it) this weakens the chain of transmission from wages to aggregate demand. But in this view, the ownership of stock doesn't count toward demand unless it limits the acquisition of goods, for example because the debt became too much; this IMHO leads to a "Steve Keen" style theory of permanent disequilibrium, that is somehow papered over by debt. Is this what you mean?

  2. (1) There's a well-established use of equilibrium in economics that means outcomes consistent with expectations, i.e. a situation in which the realized results don't lead economic decisionmakers to change their choices for the next period. In this sense, we can speak of an employment equilibrium as a situation in which firms find they have chosen the "right" level of output and employment given actual demand, and households find they have chosen the "right" level of expenditure given their income. There are many Keynesian models demonstrating in various ways that equilibria in this sense may not correspond to full utilization of the economy's productive resources; may change endogenously over time; and may not be unique given the economy's "fundamentals." An equilibrium may be stable or unstable, i.e. the relevant feedback mechanism may either dampen or amplify disturbances; or feedback may be stabilizing for some range of disturbances and destabilizing for others. That last thought was what I was trying to get at with the idea of concentric corridors at the end f the previous post.

    I think it's useful to keep using "equilibrium" in this sense, and reserve the term inertia to explain why dynamics (i.e. transitions from one state to another, regardless of whether those states are equilibria in the above sense) are (usually) fairly smooth rather than discontinuous jumps.

    (2) The important thing is how much expenditure (by both businesses and households, though here we are focused on households) changes in response to changes in income. Credit can go either way here -- if most units have unused borrowing capacity, credit will dampen the effect of changes in income on expenditure, but if many units face credit constraints that are linked to income (not implausible) then credit will be amplifying instead. The specific point I am making here is that defaults, like voluntary deleveraging, are clearly dampening, because a greater share of a fall in income goes to reduced payments to creditors, and conversely.

    I am very sympathetic to Keen's general perspective but I am afraid I have not found his work useful. I may revisit it at some point.

    1. Thanks for the explanation

  3. Thanks for the comment!

    i should add, the framing here is very much Keynesian. We could say the dominant paradigm in economic is the Walrasian, with the economy envisioned as a single, static "market" with the full set of transactions happening once and for all. The Keynesian paradigm sees a dynamic process of flows of income producing flows of expenditure which produce flows of income and so on -- thus the "hydraulic" economics of the postwar decades. The marxian paradigm, which I guess is where you are coming from, looks at the economy in terms of a different dynamic process -- the generation of surplus in the production process, its realization as profits in the market, and then reinvestment in production on an expanded scale.

    The Marxian and Keynesian frames each see the other as a a special case of itself. For Keynesians, profits and investment are just one example of linked income and expenditure flows. While for Marxists, limits on realization of profits by aggregate demand constraints are just one of various possible interruptions in the circuit of capital.

    In my opinion, neither of these perspectives is right or wrong (the mainstream Walrasian one is just wrong), both are useful for different kinds of questions. The Keynesian perspective is more useful for thinking about the quantitative evolution of capitalist economies -- question of growth, employment, booms and crises (in their strictly economic forms) -- while the Marxist perspective is more useful for thinking about the qualitative evolution of capitalism, and its relationship with (and transformations of) broader social reality.

    (And then we could add another paradigm, the Minskyan balance-sheet or money view...)

  4. I must admit that I don't really understand walrasian economics, though I suppose it is some kind of "neoclassical" idea of equilibrium.
    I often read right wing economists saying that, for example, stimulus would misallocate capital,or that the financial crisis was caused by "structural problems" such as too many housebuilders, too few nurses etc.
    Those economists apparently have this kind of equilibrium in mind:
    a) consumers buy stuff according to their subjective utilities;
    b) this means that stuff that is more useful and more rare will be more profitable, hence more capital and labour will be invested in producing it;
    c) this goes on until all stuff is equally profitable, that means that the stuff produced maximizes "aggregate utility"; this "equilibrium state" is also the "best possible world".
    On this basis financial market are supposed to be rational, because they lead to maximized aggregate utility (in theory).
    But this theory only says that the market is good at saying what share of capital and labour should go in each field of production, whereas those economists jump from this to the idea that the market also determines the optimal share of income that goes into capital and wages, and the optimal rate in unemployment.
    This would be something like saying:
    In a country there are many landlords and sharecroppers, who produce apples and oranges. The market determines the price of apples and oranges and influences how many apples or oranges are produced, in order to satisfy the tastes of the people, AND AS A CONSEQUENCE it determines how much of the product go to the sharecroppers and how much in rents, and how many sharecroppers will be unemployed, in the best possible way.
    Put this way, this kind of reasoning sounds obviously stupid, but apparently some economists really think in those terms.
    In fact they arrive to the point of saying that the value of an investiment at equilibrium is the discounted value of its future profits (throug an arbitrary discount rate); in doing this, they are impliy that the investors weight present utility whith a supposed future utility that for some reason is supposed to be discounted. This is also extremely unrealistic, I think it is much simplier to think that investors first choose how much money use in present consumption, and how much to invest in capital assets, and then choose the capital asset that promises the best rate of profit, with no link between this rate of profit and a supposed discount rate of future utility.
    This is something like saying that in the country of the above example, not only the relative price of apples and oranges is determined by tastes/quantities, but also the relative prices of apples/land depends on "time preferences" of sharecroppers VS landlords, with sharecroppers that only think to the present and landlords that are Calvinists that invest in the future.

    However if one doesn't go that far, and simply says that the market is good at saying how many apples VS oranges should be produced, the theory is reasonable.

    I would like a comprehensive theory that says that the "stuff equilibrium" [how much capital and labour is allocated to each field of production] is somehow neoclassical, but the "wage equilibrium" [level of wages, employment, profits etc.] is determined by more distributional phenomena, and is so to speak orthogonal to the stuff equilibrium. In this sense Keynes, Minsky and Marx are orthogonal, rather than opposed, to neoclassical economics (I have a very "distributional" reading of both Marx and Keynes).