Wednesday, October 24, 2012

LIBOR 4: The Gift that Won't Stop Giving.

The search is on, or let's just make up news.

One would think that at some point the media would realize it has already beaten the dead horse to death.  However, even the WALL STREET JOURNAL is prone to falling into the trap of trying to make news out of a dead issue.  Granted, LIBOR is not a dead issue, but it's one that the media is unable to cover in a technically meaningful way.  As consequence, they end up saying the absolute silliest things.  On September 28, 2012, the WALL STREET JOURNAL gave major coverage to an article entitled "LIBOR Furor: Key Rate Gets New Scrutiny.”  But the subtitle for the article points out just how far the media is stretching to keep this story alive.  The subtitle was "Banks Often Don't Change the Quotes That They Submit."
My initial reaction was "Well da!  Hasn't the Fed been holding interest rates unchanged for some time now?  Why would anyone expect banks not to appreciate that the Fed will hold borrowing costs fairly stable?"  Given the silliness of the entire issue, the article becomes worth reading in order to understand how reporters get themselves so confused.

When one does start looking at the article, it becomes an even stranger report.  The first thing strange is the shifting and ill-defined timeframe that is the focus of their concern.  The article states:
"From the start of 2012 through the end of August, the 18 banks on the main U.S.-dollar LIBOR panel left their daily estimated rates unchanged on average 87% of the time, according to the analysis of data from Thomson Reuters Corp. and S&P Capital IQ.”
Between the start of 2012 and August, one might wonder when that 13% of the time when the estimates changed occurred.  Interest rates just did not fluctuate during that period.  So it's worth investigating how the reporter arrived at this twisted logic. It is a reputable paper so it's reasonable to assume that it wasn't just to fill space.  There had to be some logic, and the article goes on to explain that logic.  It goes on to state:

"For around a third of the banks, there was only a weak to moderate correlation—if any—between their LIBOR submissions since January 2009 and the rates on their one-year credit-default swaps, or CDS, which insure against the risk of the firms defaulting. Banks, for their part, argue that there are valid reasons why CDS prices may behave differently from LIBOR."
The first thing to note is that the timeframe has changed.  We're now talking about since January 2009.  However, what is stranger is that the reporter clearly believes that the CDS market is somehow more reflexive of what the market rate should be than Fed guidance.  Put bluntly, this either represents an absurd assumption about the efficiency of markets or reflects a significant distrust of Fed statements. 

Further, it ignores the institutional framework that surrounds the CDS market versus the actual lending market.  CDS’s can be used in a variety of ways, not the least of which is hedging the tail risk implied by actual loans.  One would not expect the return or payment from an individual insurance policy to represent the individual policy’s premium cost.  The purpose of insurance (such as CDS’s) isn't to get back your premiums.  It's to lay off certain risks. 
The conditions under which the underlying assumption of the logic that the CDS market and the actual lending market should be highly correlated require eliminating from consideration all use of CDS's other than as a way to make money.  If 1) the purpose of all CDS transactions and the purpose of all loans are to make money, and 2) people making the CDS transactions and the lending decisions have the same risk profiles, and 3) the leverage implications of the CDS are similar to those of the loan, then, and only then, would one expect a high correlation between the CDS market and the lending rates.

Clearly, not all CDS transactions are intended as a speculation that will make money.  Some are, in fact, insurance against a particular risk made with the hope that there will never be a payoff.  So, the first necessary condition is not met.  It would be incredibly naïve to assume that the risk tolerance of people buying and selling CDS's are the same as those taking out or making loans.  So, the second necessary condition for the whole argument to have any merit is violated.  Finally, anyone who has ever used an option knows that it has different leverage implications from an actual long or short position.  So, the third necessary condition, also, isn't satisfied.
It's important to remember that it is only for "around a third of the banks" that "there was only a weak to moderate correlation."  So, for about two thirds of the banks, there was more than a moderate correlation.  Given the multiple uses to which one could put a CDS if one were a bank, that seems like a very reasonable portion.

Most of the balance of the article consists of quotes from various authorities that a reporter feels are supportive of his thesis.  Consequently, the balance of this posting addresses each of those quotes in turn.  They can be skipped by anyone who can see the flaw in the logic of the article.
The first of those quotes is from Andrew Lo.

“There's a concern that if you're going to base financial decisions on a particular interest rate it should be a measure that responds to changes in market conditions, and that's not LIBOR.”
This quote is the reason that the article being discussed was written on September 28 and this posting isn't being made until late in October.  Andrew Lo is no lightweight when it comes to understanding financial markets.  The Hedged Economist has referenced his work as summarized in A RANDOM WALK DOWN WALL STREET.  One suspects that his quote has been taken out of context because the general thesis of A RANDOM WALK DOWN WALL STREET is that markets are not totally efficient in the sense of conforming to the stronger form of the efficient market hypothesis.  Thus, the low correlation between lending markets and credit default swaps should not be surprising. 

Nevertheless, the quote from Andrew Lo caused me to postpone posting this discussion until after I had been able to work through the mathematics and prove to myself that mathematically, there is no reason why credit default swaps and LIBOR rates should be correlated.  For them to have high correlation one would have to believe that the distribution of the probability of collecting on a CDS is the same as the potential loss from a default on the total loan to the banks involved. 
Even if one is willing to make the oversimplifying assumption that the probabilities associated with both are normally distributed, it seems unreasonable to assume that they have the same distribution.  Further, in fact, mathematically, the distributions are not normal.  Both distributions are leptokurtic with a negative skew.  No one could reasonably argue that the kurtosis and the skew are the same on both distributions.  Yet, mathematically those are the conditions that have to be met.  So, not only is there a logical institutional reason that one would not expect LIBOR and CDS markets to be perfectly correlated, mathematically, it can be shown that the conditions under which they would be highly correlated don't exist.

It took some time to work through the math, but I'm comfortable with the statement that it's totally unreasonable to expect a high correlation between LIBOR and CDS’s.  Andrew Lo’s comment, however, is just fine.  There is absolutely no reason that credit default swaps should be priced based upon LIBOR.  LIBOR and the credit default market are measuring different phenomena.
The next quote includes some statements that quite frankly don't make sense without substantial justification that isn't provided. Darrell Duffie, a finance professor at Stanford University, is quoted as follows: 

“One would expect LIBOR submissions to change frequently if the rate reflected accurately what it's supposed to measure.”
The problem is that Professor Duffie doesn't define what he thinks LIBOR is supposed to measure.  That isn't a difficult question to answer, and if one accurately answers it, his next statement doesn't make sense. 

“This year's frequent ups and downs in credit markets mean each bank's LIBOR submissions could be expected to change ‘most days.”
Again, Professor Duffie doesn't indicate what credit markets he's referring to.  If it is the credit market for interbank borrowing, which is what LIBOR supposed to represent, there clearly was an anchor that precluded frequent ups and downs in short run interest rates.  So, he's referring to a market other than the one LIBOR measures.  It is highly likely that he is focusing instead on LIBOR’s use in the CDS market.  But that's a use of the rate not what the rate is supposed to represent.  As Andrew Lo pointed out, the use may be inappropriate.  That does not mean that LIBOR was wrong.  It may have been wrong, but the data Professor Duffie sites is totally irrelevant to that issue.  He then goes on to point out the stability in the rates as if that were an indictment:

“Instead, banks leave their rates unchanged most of the time. The analysis shows the number of submissions that stay the same from day-to-day has shot up since the financial crisis, with some firms leaving their estimates unchanged for months on end.”
Curiously, he then points out why LIBOR shouldn't be expected to change much from day-to-day.

“There's not a great deal of new information for banks to use each day”
He then goes on to point out what he thinks is responsible for the absence of changes day-to-day even after he's presented the reason.

“Instead, their herd instincts take over and they tend to avoid changes that could make them stick out from the pack.”
The hollowness of that argument is reflected in the fact that, for the majority of banks, there was correlation between their LIBOR submissions and the credit default swap market.  As stated above in this posting, it's about the right portion.  But it's the correlation not the volatility that's relevant.

The next quote is interesting.  The periods being referenced change dramatically.  It involves a comparison of 2005 and this year.  In 2005 there was considerable uncertainty regarding what would be done with interest rates.  One might remember that there was discussion of the potential that interest rates needed to be raised in order to cool the housing bubble.  The discussion was public and often took place at the Federal Reserve Board meetings.  Given that one would expect LIBOR to fluctuate more in 2005 than it did in “the last year.”  That's exactly what the quote shows.
Banks on the three-month U.S.-dollar Libor panel on average left only 29% of their submissions unchanged in 2005, compared with 75% last year and 87% this year to the end August, the analysis shows.”

The final quote and associated comment are spot on.
“There are a number of factors that can affect movements in CDS prices, meaning they shouldn't necessarily go in exact lock step with LIBOR.

But Raghu Sundaram, a finance professor at New York University, said he ‘would expect a degree of co-movement’ between LIBOR submissions and CDS spreads, as stress on a bank should increase both.”
One can't argue with Professor Sundaram, but the issue is how much co-movement, and, as stated above in this posting, the co-movement seems to be about right given the fact that there are as many factors as there are that affect lending rates and CDS’s differently.

Interestingly, the article concludes with a response from HSBC and French banks that seems to indicate that they have a good understanding of how their borrowing rates interact with the CDS market in general.
What's really strange about this sorry effort to find malfunctioning in LIBOR where there wasn't any is that clearly, there were periods when LIBOR markets weren’t functioning.  Interbank lending was frozen along with much of the financial system.  However, as discussed in "LIBOR 1: Scandal!  Or Is It?” the problem with the period when LIBOR wasn't functioning is that it doesn't provide much support to the scandal mongers.

The September 28, 2012 article discussed above was accompanied by one entitled “U.K. Unveils Plans For LIBOR Overhaul.”  One might expect a lengthy discussion of their overhaul plans.  However, what they do is far less important than that they base it on a realistic assessment of the situation they're facing.  There is no doubt that if they can divorce themselves from the political and public relations impact of their decision, they are perfectly capable of developing effective changes. 
My only comment is that the discussion in the article doesn't provide an encouraging summary.  For example, reference to removing the British Banking Association’s role in setting LIBOR is presented as if it's a punishment for the role of the BBA in the scandal.  Since the BBA was one of the first organizations to express concern about LIBOR and to indicate that they felt it was too important to be left in the hands of the BBA, one suspects the whole issue is going to be played for the bleachers rather than to improve financial markets.

A few weeks after these articles appeared in the WALL STREET JOURNAL, there was another article that made reference to the fact that a banker, had pleaded guilty to encouraging his staff to make this a submittal that would reflect well on their back.  That represents effective reporting.  Trying to manufacture a scandal doesn't.

Tuesday, October 16, 2012

The Day the Data Died.

Be careful what you say.

After hearing the employment report for September, some of the words of Don McLean’s “American Pie” came to me in a slightly altered form.  This blog has only sparingly discussed employment reports.  The reason is quite simple and is summarized in the e-mail exchange mentioned below.
A former state treasurer forwarded the e-mail with this comment:  “[a friend of his] thinks the 7.8 is proof things are getting better. You may enjoy my reply, and perhaps use the explanation in your blog.”  His reply is shown below:

On Oct 7, 2012 11:15 AM, ….he wrote: 

“The number is Obama's sound and fury... CNN should have told you that... the number is meaningless.
Last night I went to sleep at 190 pounds. This morning I woke up at 189. Proof that my diet is working, or is it all sound and fury so I can have a donut for breakfast?” 

One doesn't even need to address the content of the employment report to realize that his response is correct.  The unemployment rate is much quoted, despite being the most meaningless statistic in the employment report.

That said, the unemployment rate isn't totally meaningless, but it can only be interpreted in the context of the other content of the total report.  Other data from the employer survey, including the change in employment, the change in hours worked, the wage changes reported, and the industry composition of the employment change, are far more important than anything resulting from the household survey. 
Looking just at the household survey, the unemployment figure is much less important than changes in the labor force participation rate, or for that matter, the number of people identifying themselves as employed.  Further, the household survey does not result in just an unemployment rate.  The unemployment rate is estimated in a number of ways.  Other than the so-called “headline number,” the most widely quoted number is the total unemployed and underemployed.  That is shown both as a number and as a rate.  Consistency within the household survey is an important criterion for judging how to interpret a headline unemployment rate.
However, there is one other result than can be derived from the combined employer survey and household survey.  In various postings The Hedged Economist has alluded to the fact that at some point in the recovery, the household survey should start providing employment growth numbers that seem inconsistent with the numbers generated from the employer survey.  The usual response at that point is to question the household survey.  That's exactly what is happening now.

The July 31, 2010 posting entitled “Unemployment map: The geography of a recession” recommended:
“My suggestion is ignore the unemployment rate and focus on the change in employment if you want to measure how an economy is doing. Notice the term was employment; not jobs. It is the failure to realize that employment and jobs aren't synonymous that is the Achilles heel of this recovery.”

This blog’s posting on August 4, 2010 in a posting defending an employment report entitled “An article about a fiction and the employment report,” discussed the issue when it was the employer survey that was being questioned.  At that point, someone was pointing to the employer survey which showed employment gains that were clearly not sustainable given the total employment report and the administrations claims that the recovery was gaining steam. The comment at that time was:
“Problem is he never bothers to show whether ‘companies continued to slash jobs and hold off hiring for months, even years, after profits returned’ is anything new. It’s not. Recoveries don’t come from having the same set of employers hire back people. Recoveries have always come mainly from new or different companies hiring. Interestingly, it tends to be smaller companies and often start-ups from the current or the last cycle….”
The posting went on to note that any talk of economic recovery at that point was extremely premature.  It stated:
“The picture isn’t pretty. The evidence that the needed churn isn’t happening is there. In addition to the behavior of the two different measures of employment, Friday’s report included other data with telling implications: namely, the revisions. The BLS is aware of the limitation of the employer survey (i.e., not adequately representing new and small businesses). It tries to adjust the data to correct for the undercount using what is called the business birth-death model. That adjustment has its primary impact on the last few months of data. The data is then revised when better, some would say real, data become available. Those were the months for which the employment data were revised down. The unavoidable implication is that fewer jobs are being created by new and small business than they expected. This absence of churn among employers deserves more attention.”

The September employment report gives every indication that the policy failures that were the reason the August 2010 report didn't signal the beginning of employment recovery are becoming less important. But, as economists are fond of saying," the recovery is quite fragile."  Put differently, the private sector has healed and is recovering and has reached a point where that recovery could easily become self-sustaining.  What is infinitely apparent is that there are new headwinds, or, I should say, a continuation of the headwinds that undermined the recovery in 2010.  One doesn't have to be a strong partisan to realize that the Obama administration has been anything but friendly to small business.

That said, judging just from the employment report, the recovery is continuing its anemic pace and is poised to accelerate.  That, however, is not a forecast.  This administration has repeatedly undermined the recovery.  To illustrate, if the data could be accepted at face value the employment report itself would quicken the pace of recovery.
It should be pointed out that the suspicion reflected in that last statement is not the normal posture of The Hedged Economist.  My normal posture is reflected in a January 5, 2011 posting entitled “Reality and the jobs numbers.”  It states my normal response to charges of intentional misrepresentation in the employment numbers:

“For those who see political conspiracies everywhere it is interesting to note that the undercount in December coincided with the hearings on extending the 99 week emergency unemployment benefits. I’m amused by the conspiracy theorists. They seem to be part of a massive delusion that government is always in control of its every activity right down to the details. I hate to pop anyone’s bubble, but governments make mistakes. It’s very likely someone just got the number wrong. That they were out of touch with the private economy around them shouldn’t be surprising: they’re government employees after all....This month the news in the employment report released on Friday will be in the sector detail and the household survey.”
That quote should make it quite apparent that in most instances, the best approach is to dismiss such charges as just political rhetoric.  It is, however, interesting how frequently they have arisen during this administration, especially given how inconsistent the employment recovery has been.  That in of itself would not be enough to justify questioning the number.  However, this administration has made it clear by its own statements that it considers generating phony government statistics a legitimate campaign tactic.  That justification comes from none other than Austan Goolsbee, former chairman of the Obama administration's Council of Economic Advisers. Back in 2003, Mr. Goolsbee himself, commenting on a Bush-era unemployment figure, wrote in a New York Times op-ed: "the government has cooked the books."  Clearly, he thinks cooking the books is a campaign tactic.

Even this administration's willingness to admit that it views fudging the numbers as a legitimate campaign tactic is not enough reason to dismiss the good news contained in the employment report.  This is especially true given that inconsistencies between the employer survey and the household survey are a cyclical phenomenon.  The more troubling problem with the employment report is the inconsistency within the household survey.  For the drop in the unemployment rate to represent the positive news on employment growth that will occur at some point during the recovery, it needs to be accompanied by a similar drop in the combined unemployed and underemployed.  It also should have been preceded, as opposed to accompanied by, an increase in the labor force participation rate.

The bottom line, however, is that the employment report can contain data that is just wrong.  It's much easier for one month’s report to contain a number that is wrong than for the report to show a false trend.  Thus, next month’s report will eliminate the ambiguity introduced to the interpretation of this month’s employment report.
It's worth noting in passing that accusations of cooking the books did not originate with Goolsbee's charges in 2003.   They played an important part in Clinton's ability to convince the public that we were not recovering from the recession during his campaign against the first President Bush.  In that instance, after the election the data confirmed that we were in fact recovering.  In this current instance, there is still another employment report before the election.

Thus, the report for October employment needs to confirm this report for September.  A perfectly logical question to ask is what would constitute confirmation.  Clearly, another drop in the unemployment rate without confirming data will only confirm that the administration is cooking the books.  Confirmation should come in the form of a pickup in employment in the employer survey, a reduction in the rate of combined unemployment and underemployment, and revision to this and previous month’s anemic showing in the employer survey.  Of less importance, but also telling, is whether the increases in employment in the employer survey disperses to include more industries.
Bottom line, Jack Welsh’s comments were justified.  The September employment numbers warrant questioning.  Whether they were fudged isn't a political question.  It is question of fact that will be proven over the next few employment reports.  It will be proven, regardless of who wins the election, and woe be to the incumbent if, in fact, they were fudged.  Perhaps the administration will be able to find someone to blame for the error.  Even if they do, the number of people they can fool will decrease.  They'll be undermining their party by their obvious ducking of responsibility for the data their administration generates.