Reflecting on the Stress Tests

Most of the reporting on the stress test focused on the number of banks passing the test and the doable amount of capital the failing banks needed to raise. The tough question was not asked: if the stress test wasn’t really that stressful, why did a majority of the banks fail?

The Good News

  1. This is a move towards being more transparent. This is a good start.
  2. I was very concerned about the assumptions of the stress test. The news that leaked out before the release suggested the loss impairment assumptions were far from realistic. However, it turns out that all of the loss assumptions are for only a two year horizon. This is an important point. For instance, suppose you think that the total impairment over a four year horizon is 8%. You read in the WSJ that the government is using 4%. You think that is way too low. However, it is not as low if the horizon is only two years.
  3. The fact this is in the news is serves an educational function. People are getting a forced lesson in risk management. Whether they like it or not, this learning should decrease the chance of a systemic episode in the future. More people know the right questions to ask.

The Bad News

  1. As I have said before, the economic assumptions are problematic. The adverse scenario has 8.9% unemployment in 2009 (average). Today we learned that the unemployment rate rose to 8.9% in April. The adverse scenario has a rate of 10.3% in 2010. It is likely we will reach that rate by the end of the summer of 2009.
  2. The two year horizon essentially assumes that good times return by 2011. That is not obvious to me. We have not seen unemployment like this since 1948 and I project that we will blow by 1948 (adjusting for the size of the workforce) by the end of the summer. Employment lags the business cycle. This means there will be loan losses in 2011 and 2012. So rather than $599.2, given the assumptions on loss rates but simply extending out to 2012, a more realistic number is $1 trillion.
  3. I think the assumption on Prime mortgage losses is particularly problematic. The more adverse scenario assumes Prime losses of 3-4%. First lien mortgage category includes three components: Prime, Alt-A, and Subprime. Putting all these together the adverse scenario only assumes an 8.8% loss over two years. This includes 9.5-13% losses on Alt-A and 21-28% losses on Subprime. I focus on the Prime assumption because the market is so big. In the overall economy (not the 19 banks), we are talking about $6 trillion. So, for every extra 1.5% in losses we drop $100 billion in value.
  4. What about the other 8,000 banks? The two-year losses for the 19 are projected to be $599.2 billion in the stress test. These 19 have 2/3s of the assets and half of the loans.
  5. I reject the idea that nobody should fail. When anyone takes on risk, you reflect on the upside and the downside. You temper amount of risk you take because of the downside. But if you are always bailed out, there is no downside and you will (continue to) take excessive risk. In addition, it creates a double standard: financial institutions are bailed out but other industries or individuals are not (yes, there are some exceptions like the autos). There is also another double standard based on size: if you are big you are bailed out but if you are small you are allowed to fail. The only word to describe this is "unfair" This bailout mentality did not happen with the Savings and Loan Crisis in the early 1980s.


The Fed white paper on the stress test results is worth a read.

Follow me on Twitter.

See below my monthly employment graph that standardizes the job losses (based on the sized of the labor force) across different recessions.

This entry was posted in Finance and tagged , , . Bookmark the permalink.

One Response to Reflecting on the Stress Tests

  1. Pingback: Stress test of major US banks « Home page of Tony Bellotti

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.