Saturday, March 7, 2009

What a real stress test would look like

When Tim Geithner announced that banks would undergo a stress test, the press briefly took the view that he was being justifiably firm with the banks. If they could not perform under a dire economic scenario then they would be nationalized, the press figured. Financial bloggers, of course, were not fooled.

A few weeks later they announced the parameters of what such a stress test would look like. Here they are.

Lets focus on their more "adverse" scenario rather than their baseline scenario. The more adverse scenario is a GDP decline of -3.3% in 2009 and +0.5% in 2010. Average unemployment rate of 8.9% in 2009 and 10.3% in 2010. House price declines of -22% in 2009 and -7 in 2010.

First, GDP. Their adverse scenario is only one year of GDP decline. A -3.3% decline is about the same as the recessions in 1974 and 1982. Those were bad recessions but not once-every-50-years type events. For example in the great depression, the GDP decline in the four years following 1929 was -8.6%, -6.4%, -13%, -1.3% for a cumulative decline from 1929 to 1933 of -27%. The annualized GDP decline in the Q4 of 2008 alone was -6.3%. Q1 of 2009 is looking about the same.

Unemployment in the Great Depression reached 25%. Here is a plot of the unemployment rate back to 1948

The stress test is an average of 8.9% in 2009 and 10.3% in 2010. That would be about the same as 1982. It is 8.1% already in the beginning of March 2009.

Finally, lets look at house price declines of 22% in 2009 and 7% in 2010 which would be a 27% cumulative drop from here. This is about what I predict from extrapolating the Case-Shiller index.

So we can see that the adverse scenario is really not much worse than the 1982 or 1974 recessions with house price declines added on. Is this really a good representative of what the worse case scenario is going to be like? We have already shown that the Great Depression was far worse but that should not be surprising. Comparing to 1974 or 1982 is really not a good idea either. The 1974 recession was caused by the oil shock when OPEC raised oil prices considerably. It was alleviated when OPEC agreed to lower prices which lead to a quick recovery. The 1982 recession was engineered by Volcker's Federal Reserve in order to bring inflation down. In that case, ending the recession was easy. They just lowered interest rates and the economy came back strongly. Obviously we can't do that now since interest rates are already zero.

Is there any way that we can estimate what an adverse economic scenario would look like besides saying that it is likely to be worse than 1982 but better than the Great Depression? Well, actually we can. The key is focusing on other recessions in history that followed severe financial crises. Economists Reinhart and Rogoff wrote a paper on the economic aftermath of financial crises. They found about 20 cases and present statistical results that are relevant to what a stressed scenario would look like for the US over the next few years.

Here are some relevant statistics. The average house price decline from peak to trough was -35.5%. Ours seems likely to be slightly worse than that, maybe -40%. That is not surprising since housing was the main feature of our crisis. The worst, Hong Kong in 1997, was -53%.

The average GDP decline from peak to trough was -9.3%. The worst was the US in the Great Depression, about -27%, as we have said. The Fed's adverse estimate of -3.3% is only worse than 3 out of 15 (20%) of the historical examples. Seems hardly adverse. That is more like the rosey scenario.

The average INCREASE in the unemployment rate from peak to trough was 7%. Since we started at 5% unemployment, that would be 12% unemployment, worse than 1982 but not nearly as bad as the Great Depression. But that is just the averages not the adverse or worse than average case.

So now we can answer the question of what an adverse scenario would look like. Lets define that as roughly the 75 percentile (the average for the worst half) from this sample of post financial crises historical examples.

This would be (estimated from R&R's figures):

Real GDP peak to trough decline: -12% (3 years peak to trough)
Peak Unemployment: 16% (5 years peak to trough)
Real house price declines: -42% (7 years peak to trough)

Anyone think the banks would survive under this kind of scenario?