Handicapping the Impact of the Bailout
A couple of days ago, I wrote a post that crudely estimated that if no material government action were taken to contain the current financial crisis, then the odds of a Great Depression level economic catastrophe were on the order of 1-in-4. I did this by using the natural experiment of the impact of Monday’s House vote on the Dow, in combination with a long-run view of the relationship between major downward movements in the Dow and subsequent economic results.
There is another way to try to get at a similar handicapping question. Remember that the hypothesized mechanism that would lead to a general economic collapse is a cascading series of defaults on debt that bring down numerous financial institutions. In this situation, the stock market is a bookie, but the debt markets are the action on the field.
Suddenly, numerous commentators are talking about something called the TED Spread as if they have been trading debt instruments in London their whole lives. There is a good reason for this. The TED Spread (which stands for Treasury-EuroDollar Spread) is the difference between the rate of interest paid on 30-day T-Bills (promises to pay by the U.S. government) and 30-day interbank loans (promises to pay by private banks). It is therefore a market measure of the likelihood of default by private banks.
Most of the time the risk that the U.S. government will default within the next month is considered to be almost zero, and the risk that institutions like Citibank will default within the next month is not considered to be much higher. For most of this decade, this difference has been a small fraction of 1% – something like 0.3%. About a year ago it rose suddenly to new level, and fluctuated between about 1% – 2% from September 2007 to the beginning of September 2008. On September 2, 2008, the TED Spread was 1.1%. On October 2, 2008, it was 3.6%. This increase in the cost of debt is now moving inexorably through the lending pipeline.
Econbrowser has an excellent decomposition of this change of the TED Spread into its component parts. It makes the point that the recent change in the TED Spread is dominated by some combination of fears on the part of lending banks: that the borrower (who remember is a big commercial bank) might default within the next 30 days, and that the lending institution might have a need for the money within the next 30 days.
A longer-term perspective, however, is useful. If we go back 20 or more years, a better long-term “typical” TED Spread is more like 0.5%. There have been a couple of other spikes that exceeded 1.5%. The most severe, by far, was at the time of the 1987 stock market crash when the TED Spread was about 2.7%. Previous spikes over the past 20 years (though none close to as severe as the current spike) have not been followed by Depressions.
All things considered, the difference in the general risk (or, more technically, the market price of risk) of interbank failure is therefore, as a rough guess, about 5 – 10 times its “normal” level right now.
There was one Depression in the 20th century. It seems to me that, therefore, a reasonable naïve Bayesian estimate of the odds of any randomly-chosen year being the year that a Depression starts should be about 1%. Obviously, if you think our process of government economic management is vastly better now than in the last century, you would have a lower estimate. I hope that most people who think this are a lot more humble about this belief than they were a year or two ago.
So, a crude estimate of the odds of this being the start of a Depression level economic catastrophe seem like they are something like 5 or 10 × 1% = 5% – 10%.
Now, to grossly simplify, traders here are handicapping: (i) the odds of a bailout happening, and then (ii) the potential range of outcomes, their odds, and their impacts under both the bailout and no-bailout scenarios. Given that the bailout seems very likely to pass (I’d guess the implicit market estimate for this is well above 90%), this is probably not too far from an estimate for what the odds of a catastrophe are even in the face of government action.
Combining the prior post with this one, that means that an unbelievably crude ballpark estimate for the market forecast for the impact of the bailout is to reduce the odds of a catastrophe from something like 25% to something like 5% – 10%. As per the previous post, this is really just guesswork, but I think if you play with these assumptions all you want, you will always come up with a directionally similar result.
If this is true, we are paying all this money and accepting all of these ideological costs in return for a 15 – 20 percentage point reduction in the odds of economic collapse. I think it’s clearly worth it, but it is an awfully unpleasant decision. Welcome to the very unpleasant economic world we’re likely to be living in for a while. Most Americans are discovering that they are not nearly as wealthy as they thought they were a year ago.
In the end, though, if this is a warning sign that stimulates a healthy reduction of debt in America without a massive contraction, it will be a positive step. The day you decide that you don’t want to deal with all this credit card debt, mortgage debt, car debt and so on, and take back control of your life is usually a good day. Even though you eat out less often and don’t buy that new, new flat screen TV or pair of shoes that you might previously have splurged on (with debt), you still are happier and more fulfilled. Let’s hope that’s where we’re moving.
It is very difficult to try to estimate the odds of a black swan event based on historical means, i.e. inductive reasoning. For example, the assumption that the annual risk of a Depression is 1%, based on 1 happening in the last century. Based on some quick calculations, an observed probability of 1% is consistent with an actual probability of ~0.1%-5%. Then, using the assumption that 5-10X higher TED spread means a 5-10X higher risk of a Depression, then the conceivable value is anywhere from 0.5-50%. I think my personal cutoff value for committing $700 billion to reduce the likelihood of a Depression is probably around 5% (assuming the bailout lowers the risk of depression by 50%), so I support it as well. Then again, others need a $6 million credit for wooden arrows to convince them…
— ol2102 · Oct 3, 07:18 PM · #
I agree with your point about the underlying philosophical problem of induction. Without writing a novel about it, the key way this manifests itself in a probability estimate is defining the relevant population of cases from which we draw examples.
To be practical in this case, while economic statistics get worse as you go back in time, so it’s hard to measure relative severity, by common attribution there were two Depressions in 19th century America: the Depression of 1807, and the so-called Long Depression of 1873-1896 (which affected most of the industrialized world). So I think it’s reasonable to assume a tigher distribution of outcomes than 0.1% (i.e., one Depression per thousand years) to 5% (i.e., one Depression per twenty years, which, if they last an average of about 10 years each, would mean spending half our time in Depressions).
All that said, these are total SWAGs that I would never even attempt unless I thought we were in a gun-to-your-head decision environment.
— Jim Manzi · Oct 3, 07:47 PM · #
This economic crisis has left me feeling pretty at sea, so I still don’t know what to make of the bailout and the rest. Nevertheless, yours was the most cogent explanation of the TED spread I’ve seen, and I thank you for at least giving me a little more of a handle on that particular acronym.
— Justin K. · Oct 4, 08:06 PM · #