Saturday, April 18, 2009

The 'Stress Tests' Are Really 'Fudge Tests'

Original Link: http://www.forbes.com/2009/04/15/gdp-stress-tests-unemployment-banks-home-prices-opinions-columnists-nouriel-roubini.html,

By Nouriel Roubini

The spin machine about the banks' stress test is already in full motion. Some banking regulators have already served up--to The New York Times--their spin that all 19 banks that are subject to the stress test will pass it. In other words, not one will fail.

But let's look at the actual data. The macro data for the first quarter on the three variables used in the stress tests--growth rate, unemployment rate and home-price depreciation--are already worse than those in the U.S. government baseline scenario for 2009. They are, in fact, even worse than those for the stressed scenario for 2009.

The government used assumptions for the macro variables in 2009 and 2010 that are so optimistic that the actual data for 2009 are already worse than the adverse scenario. As for some crucial variables, such as the unemployment rate--key to proper estimates of default and recovery rates for residential mortgages, commercial mortgages, credit cards, auto loans, student loans and other banks loans--the current trend shows that by the end of 2009 the unemployment rate will be higher than the average unemployment rate assumed in the more adverse scenario for 2010, not for 2009. Put plainly, the results of the stress test--even before they are published--are not worth the paper on which they are written.

Let us look at how the stress tests are done. According to the U.S. government, there are two scenarios: a more optimistic "baseline scenario" for 2009 and 2010 for the three macro variables (gross domestic product, unemployment and home prices); and a more pessimistic "alternative adverse scenario."

The baseline scenario assumes--based on the average of the forecasts by the consensus of macro forecasters at the time when the stress tests were announced--that GDP growth will be -2.1% in 2009 and 2% in 2010; that the unemployment rate will average 8.4% in 2009 and 8.8% in 2010; and that home prices will fall 14% in 2009 and 4% in 2010. In the alternative adverse scenario, GDP growth is assumed to be -3.3% in 2009 and 0.5% in 2010; the unemployment rate is assumed to average 8.9% in 2009 and 10.3% in 2010; and home prices are assumed to fall 20% in 2009 and 7% in 2010.

The description provided by the government of the stress test also shows graphs--but not actual figures--for the quarterly behavior of the three macro variables in 2009 and 2010 for both scenarios. Based on these quarterly graphs, in the first quarter of 2009 the unemployment rate would approximately average 7.7% in the baseline scenario and 7.8% in the adverse scenario; the GDP growth rate would be -1.9% in the baseline scenario and -2.1 in the adverse scenario; and home prices would fall 4% in the baseline scenario and by 7% in the adverse scenario.

How do these scenarios actually stack against actual figures for the first quarter of 2009, with current consensus forecasts and with current likely paths for these macro variables?

1. Take unemployment: In March of this year the actual unemployment rate was already--at 8.5%--much higher than the baseline scenario for the first quarter (7.7%) and higher even than the adverse scenario's 7.8%. Even the average unemployment rate in the first quarter--8.1%--is well above both the baseline and adverse scenarios.

Based on current trends in employment and initial claims, Ted Wieseman, U.S. fixed-income economist for Morgan Stanley and a very mainstream analyst, has recently argued that the unemployment rate could reach 10% by June of 2009. As he put it:

The key round of early March economic news was weak, with another disastrous employment report again standing out. Non-farm payrolls plunged 663,000 in March, bringing the average drop over the past five months to 667,000, an unprecedented run of job losses in absolute terms and the worst in percentage terms over such a period since the 1973-1975 recession ... The past week's jobless claims report showed substantial deterioration as we move toward the survey period for the April employment report, so at this point there is no reason to expect the extremely weak recent trend in the jobs report to show any improvement next month. Indeed, at the rate jobs are disappearing, the unemployment rate could be into double-digits as early as June (my emphasis).

Indeed with per-month job losses well above 600,000 for several months in a row, and with initial claims still averaging 650,000 per week for the last few weeks, there is no chance that job losses will be any less than 600,000 for the next two months.

Even if the economy were to turn to positive growth by the third quarter--as the consensus forecasts expect--the unemployment rate would rise for at least another 12 months, as job market data are lagging indicators of economic activity. For example, in 2001 the recession was short and shallow--only eight months long--and over by November, but job losses continued for another 19 months until August 2003. So based on current trends--and even heroically assuming that the economy recovers positive growth by the third quarter of 2009--it is almost certain that the unemployment rate will be 10.5% by December of this year (and would thus average about 9.5% for the year). In a more adverse--but more realistic--scenario, the unemployment rate would reach 11% by December of 2009 and average 9.8% for the year.

So based on current actual data, the unemployment rate is already well above the values that the U.S. government had for them in the first quarter of 2009 in both the baseline and stress scenarios. And given current trends in the unemployment rate and initial claims for unemployment benefits, the average unemployment rate in 2009 would be 9.5%. That is not only well above the baseline scenario of 8.4% for 2009, but also well above the average unemployment rate for 2010 in that baseline scenario. In fact, 9.5% would be even worse than the average unemployment rate (8.9%) that is assumed by the government for 2009 in the adverse scenario. (Actually, based on current data for initial claims, it is highly likely that by April 2009 (this month) the unemployment rate will reach 9% and thus be higher already at this early part of 2009 than the average of 8.9% that the government assumes it will be for the average of 2009.)

But, more important, the actual unemployment rate by the end of 2009 will be higher--at 10.5%--than the average unemployment rate assumed by the government in the adverse scenario for 2010, not 2009.

2. Now let's look at GDP. A similar analysis suggests that the U.S. government assumptions for GDP growth are already worse than the adverse scenario--let alone the baseline scenario--for the first quarter of 2009.

A first estimate of first-quarter 2009 GDP growth will be out only at the end of April 2009, but the current consensus is that it will be around -5% for the seasonally adjusted annual rate (SAAR) figure in the first quarter (i.e., the growth rate of the economy in Q1 2009 relative to Q4 of 2008, seasonally adjusted at the annual rate), with a number of reputable forecasters putting that figure at -6% or even worse.

Now, based on the government graphs, the GDP growth rate would be -1.9% in the baseline scenario and -2.1% in the adverse scenario. If we plug the current consensus estimate of Q1 GDP growth and then compute the year-on-year four-quarter percentage change, we get a figure of -2.3%. Thus, based on consensus estimates of GDP for Q1 2009, the four-quarter contraction of GDP will be--in the first quarter of this year--already worse than both the baseline scenario figure (-1.9%) and the more adverse scenario figure (-2.1%). And if, as is possible, the GDP contraction in the first quarter is -6% (SAAR), the Q1 2009 vs. Q1 2008 percentage change in GDP will be -2.5%, an even worse figure. So by the first quarter of this year, both the unemployment rate and the GDP growth rate are (or will likely be) worse than the government figures for both the baseline and more adverse scenarios.

Moreover, relative to the time in December when consensus forecasts for 2009 were used by the U.S. government to derive its baseline scenario for 2009, such consensus forecasts have significantly worsened. Based on the latest March figures, the consensus is now projecting that 2009 GDP growth will be -3.2%, rather than the -2% in the original government baseline. And mainstream sell-side research from Goldman Sachs, Merrill Lynch and Morgan Stanley--houses that have had a better-than-average track record in forecasting the current downturn--are now forecasting a 2009 GDP contraction of about 3.5% on average. Our forecast at Roubini Global Economics is even worse, at 3.7%.

Also, while the current consensus forecast for 2010 growth (2%) is practically identical to the baseline scenario for 2010 GDP growth (2.1%), a number of sources are predicting a much weaker scenario for 2010: Goldman Sachs, for example, has a current forecast of 1.2% for 2010 GDP growth, as opposed to the baseline scenario figure of 2%.

3. Let us now consider the outlook for home prices. Unfortunately, the Case-Shiller 10-City Composite figures are available only up to January 2009, with the February figures due to be published at the end of April. The government baseline assumes that home prices will fall--December 2009 vs. December 2008--by 14%, while its more adverse scenario sees home prices falling by 22% during 2009. The fall in home prices in January 2009 relative to December 2008 was 1.9%, a rate that--if continuing for all of 2009--compounds to an annual rate of 25.3%, well above the 14% of the baseline scenario and also above the adverse scenario of 22%.

One month does not make a trend, and the 12-month percentage change in home prices was -19% in January 2009 and December and November 2008, up from the 18% drop in August-September-October 2008 and the 17% drop in the May-June-July 2008 period. So in the last year the rate of home-price depreciation has accelerated from 17% to 18% to 19%--and, most recently, to 25%.

It is possible that the rate at which home prices will fall for the rest of 2009 will be lower than the latest 25% rate. But even the more recent optimistic data for the housing sector--a possible stabilization of housing starts, building permits, new home sales and existing home sales--do not imply that the rate of fall of home prices will be significantly lower in the months to come. The reason is that, while home sales and housing starts and completions are now showing signs of stabilization after falling from their peak by more than 70%, the excess supply of new and existing homes is still huge, both in absolute terms and as a ratio of sales, thus a significant downward pressure on prices.

Even if housing starts and completions were to go to zero--from their current 500,000 level--it would take about 11 months until the current supply of new homes is depleted by the current demand. And since a recovery of both supply and demand from current depressed levels would reduce the excess inventory only if new-home sales are well above new completions, there is no reason to believe that home prices will fall at a much slower rate for the rest of 2009. In other words, even if quantities in housing were to stabilize and then recover in the next few months, prices will keep falling for all of 2009 and 2010 at a very rapid rate.

In summary, home prices have been falling in recent months at a rate that is much higher than the 14% assumed in the government baseline for 2009. They are also running currently at an annual rate that is higher than the 22% in the more adverse scenario. Even considering actual figures for the last few months--that show an accelerated rate of fall in homes prices between the spring of 2008 and the most recent data--home prices have been falling in the last few months at rates of about 20% with an upward trend in the data. So the actual and trend figures are well above the baseline figure of 14% and closer to the 22% of more adverse scenario.

In conclusion, recent data and trends for the unemployment rate, GDP growth and home prices show that, as of Q1 of 2009, actual macro data are much worse than the baseline scenario of the stress tests and even worse than the more adverse scenario of the stress tests. So the results of the stress tests--even before they are published--are useless.

The stress-test exercise was already flawed from the start; it took as its baseline scenario the consensus forecast for the economy for 2009 and 2010, a consensus that, for the last two years, had gotten the growth rate of the economy totally wrong. And the more adverse scenario of the stress test was not really a truly adverse scenario: It assumed positive growth for 2010 (rather than additional negative growth) and assumed an average unemployment rate for 2010 that, based on current data and trends, is likely to be reached at the earliest by the third quarter, and at the latest by the fourth, of 2009. How can anyone take seriously stress tests that are blatantly rigged from the outset with scenarios that make little economic sense and that are already obsolete?

This financial crisis was due to opacity and lack of transparency in financial markets and regulators who were asleep at the wheel. But now the administration and the regulators have decided to add liberally to the fog of opacity. Why call them "stress tests"? "Fudge tests" would be a truer description.

No comments: