Thursday, February 23, 2012

At Rortybomb: The Real Causes of Rising Debt

Last week I promised a discussion of my new paper with Arjun Jayadev on "Fisher dynamics" and the evolution of household debt. That discussion is now here, not here, but at Rortybomb, where Mike Konczal has graciously invited me to post a summary of the paper.

The summary of the summary is that the increase in household debt-to-income ratios over the past 30 years can be fully explained, in an accounting sense, y changes in growth, inflation, and interest rate. Except during the housing bubble period of 2000-2006, household spending relative to income has actually been lower in the post 1980 period than in preceding decades. If interest rates, inflation and growth had remained at their 1950-1980 average level, then the exact same household decisions about spending out of income would have left them with lower debt in 2010 than in 1980. And just as it wasn't more borrowing that got us higher debt, less borrowing almost certainly won't get us to lower debt. If household leverage is a problem, then the solution will have to be some mix of large-scale writedowns, higher inflation, and lower interest rates via financial repression.

But I encourage you to read the whole summary over at Rortybomb or, if you're really interested, the paper itself. Comments very welcome, there or here.


UPDATE: Now also at New Deal 2.0.

UPDATE 2: Responses by Kevin DrumKarl Smith, Merijn Knibbe, Reihan Salam, and The New Arthurian. There's some good discussion in comments at Mark Thoma's place. And a very interesting long comment by Steve Randy Waldman in comments right here.

10 comments:

  1. JW — First, thanks for doing this. It really is an interesting approach.

    I think there are a lot of conceptual complexities I've not worked through yet. Just as it's inappropriate to view government debt as analogous to household debt, it's inappropriate to view household sector leverage as analogous to a household's leverage. It has become trendy — and I think it is right as far as it goes — to call attention to measures of aggregate leverage and indebtedness as markers of financial fragility, and to mock the "orthodox" dodge of noting that one party's debt is another's asset so debt doesn't matter. But still, we have to resist the inclination of imagining household sector debt-to-income has implications similar to one household's debt to income. They are very different beasts. We have to think about who is lending to whom to generate this leverage. Is the household sector leveraged by virtue of claims against it by the business sector, or the government, or foreign creditors? Or is the leverage internal? If it is a matter of some households lending to others, and what are the implications of that?

    What is most striking to me about your wonderful Figure 6 is that it is both concealing and revealing a story of who. It is the interest line that really drives the tale, that exerts a terrible upward push on debt-to-income despite the fact that aggregate new borrowing is negative and neither inflation nor real growth seem outside of historical ranges.

    Superimpose your graph of the effect of interest on household-sector indebtedness with the "mountain graph" of interest rates that peaks around 1980 and the collapses to zero in the three decades since. "Benchmark" interest rates have fallen very dramatically, and yet interest payments as a share of household income have pushed upward relentlessly. You'd see a huge wedge between declining rates and growing payments. Part of that can be explained by the overall growth of indebtedness as a share of income, the base from which interest is computed is growing. But I suspect that not nearly all of the wedge can be explained by that. The rest of the difference is a credit spread, which I suspect you'd find has been expanding relentlessly over the period. (cont'd)

    ReplyDelete
  2. In other words, if my read of the tea leaves you've offered is right, you've documented a change in the character of borrowers (or the competitiveness of the institutions through which they borrow). In earlier periods, expansion of household indebtedness was mediated through creditworthy borrowers or against low-risk collateral at rates not distant from the benchmark rate. But over time, the expansion of household indebtedness undergoes a shift. It is mediated via unsecured or more speculatively secured debt, less creditworthy borrowers, and/or via less price-competitive intermediaries.

    I think this "anomoly" — the rising share of income devoted to interest payments despite rapidly declining rates — might rescue the traditional liberal story of income-stagnant consumers supplementing their purchasing power with borrowing. That story might well describe the behavior of the "median" household, or households outside the top quartile. Looking at aggregates you are implicitly looking at averages, and the apparent nonincrease in primary borrowing may be a matter of deleveraging at the top of the income distribution masking continual primary borrowing in the middle to bottom.

    One way to tell the story is that, in the past, before the "wedge" between interest rates and interest payments appeared, borrowing was the province of the rich and creditworthy, a matter of liquidity and convenience but with very limited credit risk. As the wealthy have got wealthier, they've been less frequently "asset rich, cash poor". They've come to have less need for liquidity and more need for remunerative destinations for capital. Meanwhile, stagnant wage income and deregulated lending made it both necessary and possible for more modest households to supplement their income with high interest borrowing. The aggregate effect on primary borrowing of this "rotation" was muted, but it's effect on interest payments was dramatic, despite the secular decline in benchmark rates.

    Anyway, thank you for an excellent and interesting review of household sector debt.

    [micronitpicks: the legend on Figure 6 is hard to follow. "i" could mean interest, income (growth), or inflation. b is debt/income and d is debt, that's not obvious. i'm perhaps more graph dyslexic than most, but i confused myself a few times before getting it down. in your conclusion you say "some combination of lower g, lower i and higher π will be necessary". I think you mean higher growth, although it would make the line 'g' on your graph go lower.]

    ReplyDelete
    Replies
    1. I agree that the study of aggregate data likely misses important parts of the story. This graph suggests that the bottom 90% of the income distribution - and particularly the bottom 20% - was probably adding new debt all along :

      http://static.seekingalpha.com/uploads/2009/12/3/saupload_debttoincome.jpg

      source article :

      http://seekingalpha.com/article/176462-debt-to-income-ratios-and-the-u-s-savings-rate

      Delete
  3. Steve-

    Thanks for these fantastic comments. I owe you a proper reply and will try to finish writing it soon. I admit, even though thoughtful, intelligent criticism that points the argument in a new direction is exactly what we all want, it's also the most demanding criticism to respond to, and I've been putting it off a bit. But will have some things to say in response tomorrow, I hope.

    ReplyDelete
    Replies
    1. Excellent paper, a synthesis of many crucial concepts that have been overlooked until recently.

      Some quibbles:

      1. You could acknowledge the work of Australian economist Steve Keen.

      2. The key question is why higher household/private debt levels do not circulate as higher incomes/GDP. The answer seems to be that targetting inflation in the price of goods and services restrains income inflation but not asset price inflation. The result is larger and larger pools of credit trapped in asset markets and not trickling down to raise incomes for the majority. It is fascinating to note the correlation between private debt:GDP and income inequality (gini index).

      Delete
  4. Finally responding to SRW:

    Just as it's inappropriate to view government debt as analogous to household debt, it's inappropriate to view household sector leverage as analogous to a household's leverage.

    Fair point, but i think that supports our methodology in that the sector as a whole is more analogous to the government than an individual household is. But we do need to think about this more. (Economic theory does not help us much since the representative household approach is almost universal.)

    It has become trendy — and I think it is right as far as it goes — to call attention to measures of aggregate leverage and indebtedness as markers of financial fragility, and to mock the "orthodox" dodge of noting that one party's debt is another's asset so debt doesn't matter.

    This is one trend I am happy to jump aboard. While the stick has been bent back a bit, I think it needs to be bent a lot further. I would add that while there is certainly an increasing interest in leverage, I don't think there has been much systematic attention to the mechanics of how leverage evolves over time.

    Is the household sector leveraged by virtue of claims against it by the business sector, or the government, or foreign creditors? Or is the leverage internal?

    Strictly speaking, households do not lend to other households. The immediate flows of new credit to households, and debt service payments from households, are overwhelmingly with the financial sector. Some consumer credit is extended by the non-financial business sector, as is trade credit to nonprofits serving households. But a good 95% of household liabilities are held by the financial sector. (In the FoF convention, the rest of the world does not lend to US households; local offices of foreign institutions are counted as part of the domestic financial sector.)

    Now in principle net interest flows from households to US banks should eventually show up as dividend payments from banks to (other) households. So in some abstract sense, it is households lending to other households. But I don't know that that changes the story in any fundamental way. Even in a world of identical households lending to each other, a larger ratio of total financial claims to income would lead to greater fragility.

    
"Benchmark" interest rates have fallen very dramatically, and yet interest payments as a share of household income have pushed upward relentlessly. ... Part of that can be explained by the overall growth of indebtedness as a share of income, the base from which interest is computed is growing. But I suspect that not nearly all of the wedge can be explained by that. The rest of the difference is a credit spread.

    This is a sharp observation. But note that our effective interest rate is the ratio of total household interest payments to the total stock of household debt, i.e. the actual average (not marginal) interest rate faced by households. So while spreads over the policy rate have increased, it is still true that, after all the changes inhe composition of debt and borrowers, the interest rate faced by households has declined. The story really does seem to be about the growth of the base as we try to show schematically in Figure 6.

    Also it's not the case that debt-to income ratios increased most at the low end of the income scale. The largest increases happened in the 40th through 80th percentiles, i.e. the middle and upper-middle parts of the distribution. So I don't think a story about borrower quality works. It really does seem that the interest rate plus disinflation shock of the early 1980s knocked households off of a sustainable debt path and they have not been able to get back onto it despite sizable reductions in spending and the eventual return of interest rates to even lower levels than pre-Volcker. If you read David Graeber (who really should cite), people have been falling into these sorts of debt traps periodically for the past 5,000 years.

    ReplyDelete
  5. JW — Thanks for the thoughtful and in-depth reply. I've very little to quarrel with in it, though I won't quite cede my spin on things. I'm with you that paying attention to financial fragility is a fine trend to jump on.

    I understand too that inter-household lending is intermediated by the business sector, but I'm still trying to work out whether that's an analytically important detail: do we lose very much if we imagine this leverage as being generated households lending to other households? Certainly, if what we are interested in is the financial system, we'd better capture the fact that lending is intermediated, and pay attention to the institutional details of the intermediation. But when we are tracking the evolution of household balance sheets, is it okay to look through financial intermediation? I don't know.

    I'm not so dissuaded from the "traditional liberal story" by the data point that leverage has been concentrated in the 40 to 80th income percentiles. I get in trouble for this sometimes, but those to me count as "people of modest means" for the purpose of the story. (Obviously, there are many contexts in which it would border on offensive to refer to this group as poor-ish.) But the "traditional liberal story" (to which I am attached, but I'm not irrevocably dogmatic) seems to me a story of a bifurcation of the middle and upper middle from the top. The poor, as they often are, are eclipsed, ignored and obscured by that story. But I think it is nevertheless real (and very salient in national debate because the middle and upper middle include people who are educated and culturally enfranchised but who are falling behind financially).

    Lending standards became loose, but not so loose as to direct large sums of borrowed money to people in the 10th percentile income percentile without assets. The very poor, recently as before, subsist on income, informal mutual insurance, government transfers, and forms of borrowing (pawn, payday lending) whose exaggerated collateralization usually limits the quantity borrowed. It was those in the middle who on the one hand have been "falling behind", and whose FICO scores and W2 history constituted adequate security from the perspective of backwards-looking credit modelers. This group also has had a cultural expectation of keeping up with an American "mainstream" that requires a tragically excessive affluence (even to cover "necessities" like health insurance and education). The story I'm perhaps overattached to is that this group of people substituted borrowing for wages, because they could, and because they had to in order not to fall out of what they (we) consider the middle class. So the 40 to 80 datapoint is consistent with this.

    We're in sync on the interest you show as reflecting an effective or average, not marginal, rate. And so some of the upward slope in interest/income that you document, besides reflecting a growing base, is just a matter of duration, the collapse of interest rates takes time to translate into new borrowing at lower rates. There are numbers that have to be run here. They could be run, and I could just be wrong. But I'd still place my money on the notion that you can't explain such a huge increase in interest/income on the growing base alone given collapsing benchmark rates, and that you'd have to posit implausibly long durations (remember, most consumer debt is floating rate or prepayable/refinanceable) to let the wedge between the cost of past debt and rates on new borrowing explain the growth. I still think you'll find you need a growing credit spread. But again, that's an empirical question, and I could very well be wrong.

    Thanks again for your excellent work, and for the detailed and interesting response.

    ReplyDelete
  6. Maybe this is OT, but I have this opinion about "middle class reckless spending":
    Many of the behaviours that are associated with "middle class overspending" are, IMHO, perceived as a form of thrift by those who spend.

    For example: Suppose that I live in a rented apartment, and I spend 400$ in rent. I will spend 400$ for all my life. But then I choose to buy the apartment with a mortgage: I spend a lot of money upfront, and have to pay 500$ in payments to the bank. Thus temporaneously my consumption is lower than what I could have just renting the apartment, but in a distant future I will own a capital good (the apartment) and have more disposable income. This is an "ant" behaviour, not a "grasshopper" behaviour, but it shows as overconsumption on the statistics.
    Another example could be student debt: In a world where it is increasingly hard to land a job, I try to enhance my future job opportunities by spending a lot in education. In this situation also I'm not presently overconsuming, I'm trying to accumulate assets to increase my disponsable income at a future date, and I actually constrain my consumption at present times, but it shows as present overconsumption in the statistics because I'm going into debt.

    So in the end I think that the concept of "overspending" in a financial logic is very decoupled from "overspending in an ethical/customs logic, and while the first might or might not have happened, the second didn't happen for sure.

    ReplyDelete
    Replies
    1. RL-

      I think the larger answer here is that there is not one appropriate definition of "saving," it really depends on what questions you are asking. On the one hand, we may be interested in, as you say "ant behavior" vs. "grasshopper behavior," in which case any expenditure today that improves the unit's income flows or standard of living in the future is saving, including reduced borrowing as well as e.g. purchase of housing or other long-lived assets. Or, we may be interested in behavior that frees up resources on an economy-wide level, in which case we should include reduced borrowing and purchases of financial assets, but not tangible assets. Or again, we may be interested in changes specifically in the unit's debt ratio, in which case we want to look only at changes in borrowing and not at the asset side at all (or at least, only at assets insofar as they produce reliable, short-term cash flows that can be used to service debt.) None of these measures is the "right" one in the abstract, they are each the right answer to a different question.

      Now, in the US, the national accounts are actually closed to what you call the ethical or customary logic, in that purchases of housing are counted as saving. This is slightly problematic in that in that it is inconsistent with the way other household purchases of durable goods are treated, and because it then requires imputing a flow of housing services to the household in a way that is not done for any other form of household production. But more importantly, for our purposes here, this treatment of housing is quite misleading from the point of view of assessing households' *financial* position. So e.g. n the 1980s there was a substantial fall in housing investment relative to income. This is an entirely rational response to higher interest rates. but because the national accounts treat housing purchases as a form of saving just like a reduction in borrowing, reduced housing purchases in response to higher interest rates looks like a perverse fall in savings rather than the rational response that it actually is. For this reason, part of what we did in this paper was to adjust the standard measures of household income and expenditure to treat housing purchases as consumption and remove the imputed flow of rental payments and housing services (about 3 percent of GDP) that households are supposed to be exchanging with themselves. From our point of view, a household that incurs a liability (i.e. a mortgage) to acquire a house has increased its leverage, but in the standard treatment in the national accounts its financial position has not changed.

      So there is definitely an important sense in which what you are saying is true. But, if we are interested in the *financial* position of households (or other economic units) specifically, we need to distinguish between different forms of saving.

      Delete