I’ve been reading Todd Zywicki’s paper An Economic Analysis Of The Consumer Bankruptcy Crisis (1Mb PDF). Zywickiâ€™s approach is to look at aggregate time-series data on a set of suggested causes of rising bankruptcy, suggest that the pattern for these time-series doesnâ€™t match the observed increase in bankruptcy, The main point is, as he says,
Static or declining variables, such as unemployment, divorce, or health care costs, cannot explain a variable that is increasing in value, such as bankruptcy filing rates.
Hence, he says, the ‘traditional model’ of bankruptcy as a “last resort” outcome of financial distress is no longer valid. He therefore falls back on the residual hypothesis of changes in consumer behavior in the form of an increased willingness to resort to bankruptcy, possibly due to the rise of impersonal modes of lending and the decline of moral sanctions. Zywicki doesn’t mention the other obvious residual possibility: an exogenous increase in willingness to lend to high-risk borrowers, but symmetry suggests he ought to.
I don’t think Zywicki’s is the ideal research strategy (see below) but it has the advantage that anyone can play, armed only with Google. So let me point to a variable that has risen in the right way and could reasonably be expected to lead to rising rates of bankruptcy. That variable is the volatility of individual income, or, in simpler terms, the economic risk faced by the average person.
What this means is that the bankruptcy ‘crisis’ is an outcome of the general changes in the US economy over the past 30 years or so. If it weren’t for expanded credit and increased reliance on bankruptcy, the distress caused by growing inequality and income volatility would have been substantially greater. If bankruptcy laws are tightened, distress will increase. To put it simply, bankruptcy is the lesser of two evils.
The link between income volatility and bankruptcy is straightforward. People enter into commitments, and establish consumption habits, when income is high, then run into financial trouble when their income falls.
If income volatility is mostly temporary (income returns to the previous level over a year or two) access to credit helps to smooth consumption. If income volatility is mostly permanent, then the best response is to adjust consumption immediately.
The trouble is, you canâ€™t tell the difference. So people increase their consumption commitments in response to an increase in income that turns out to be temporary or maintain consumption in response to a decline that turns out to be permanent. This is a rational response, but it leads to increased risk of bankruptcy.
The aggregate numbers are consistent with this. Aggregate consumption is more stable than aggregate income.
The obvious place to look for evidence on individual income volatility is the Panel Survey on Income Dynamics. A quick Google (I used â€œPSID income volatility consumption temporary permanent), confirms the broad outlines of the story above
* individual income volatility has increased fairly steadily since the 1970s, with increases in both temporary and permanent volatility
* income inequality has also increased
* consumption inequality and volatility have not increased nearly as much
* this can be explained by reliance on credit
This paper by Kreuger and Perri (0.5 MB PDF) is useful.
The aggregate trend approach Zywicki and I have used is not the ideal way to address the question. The best thing would be to look at individuals and see what kinds of things predict bankruptcy. There are a couple of ways of doing this.
The best, I think, would be to get access to the models used by credit providers to assess risk. The increase in bankruptcy might be explained by an increase in the proportion of borrowers with high risk characteristics. Alternatively, on Zywickiâ€™s account, there might be an exogenous increase in the default risk for borrowers with given characteristics. Iâ€™d expect a bit of both, since the two reinforce each other. Given a big increase in bankruptcy due to increased risk, the moral and legal sanctions against bankruptcy become less effective, which produces a second-round increase.
I donâ€™t suppose credit providers are going to part with their models. So, the next best thing would be to estimate a model of your own using a panel data set like PSID. Iâ€™m sure someone has done this, but a quick Google scan didnâ€™t turn them up, and Iâ€™m betting that one of my well-informed readers will be able to point me in the right direction.