BLS Hypothetical Job Growth is 2500% of Total!!

The attention-getting headline for today's article is true, but misleading.  It parallels the monthly assault on the BLS Birth/Death model from Barry Ritholtz and others.  In the most recent article, Barry incorrectly writes that the BLS has substituted modeling for measurement.

In fact, before the BLS adopted the birth/death adjustment they did not "measure" newly created jobs as Barry states.  They just treated those not responding to the survey the same as those who did.  This was clearly not correct, since actual state employment data proved that there was a sea change underneath the surface, something like 2.5 million jobs lost and created each month.

Briefly put, the "old method" assumed that job losses were offset 1 to 1 by newly created jobs.  These new jobs were not measured, and the "model" was an estimation.

Using the Ritholtz headline tactic, these are millions more "hypothetical jobs" than those imputed from the birth/death model.

When the BLS discovered that this old method was not capturing all of the new jobs, they created more sophisticated modeling techniques.  The birth/death adjustment has dramatically improved the performance of the job count, as demonstrated by actual state data.

We covered this on many occasions, and in particular last month.

Barry is followed by so many through his blog and media quotations, so he has a special responsibility to get this right.  He has criticized this method since its inception, as he himself notes.  Here are two simple questions for him: 

1. Can he find a single time period since the introduction of the birth death model where it made the estimate less accurate than would have been the case with the old method?  We cannot.  How can something that makes the job count more accurate be criticized?
2. What is so good about the "old" method of assuming that job creation exactly equaled job losses?

Comments

[Paddock]

"Use a mutual fund screener and you'll find out. Get off your own laziness. FYI, there are 7 managers above 15% for 10+ years annualized with 10+ years of tenure, in the domestic diversified equity space."

Except that includes results from before his funds inception.

"Hmm, is the risk/reward ratio not "favorable" in a five-year bull market run? I would think that would be part of the definition of a bull market, for an intelligent person ..."

Whether the insult was directed at Hussman or myself I am not sure. However, Hussman may be a bad manager, but certainly he is intelligent. The answer to your query is no. First of all he has not been fully hedged for the five year period. Second, risk does not mean down. It is about probability and the excess return one can get given the potential risk. The theory is that over time he will outperform, not any particular period. It also doesn't explain his top ranked stock selection either. He has also been right if you think about the reward part. This has been an abnormally tepid bull run so far. So, the risk of a downturn has been rewarded with small returns. Typical bull markets give you a much higher return for taking the risk of equities.

I am not sure what your hostility is about, I was asking a question. I guess at this point I can only assume you know of no domestic only managers who have returns higher than his over that time frame, you just assume they exist. So, if you are not impressed with the returns he has generated over that time period, I must assume it takes being above the top 1% of all managers to do so.

As for my own goals, I have done significantly better than Hussman over the last few years myself, more than quadrupling him YTD, I did even better than he did in November (when I made money) and I am very comfortable with that. Low risk, high return, that is the holy grail. What it does not mean is that Husmman's returns have been low over the fund's lifespan. They have been high.

Still, though you seem to take this personally, rather than in terms of discussion, I decided to look at exactly how many domestic oriented managers with more than 10 years of tenure and returns in excess of 15% there are. I assume given what you said above you would consider them high quality managers, the cream of the crop.

I was able to come up with 11 which were domestic, and not specialty funds. One we can just eliminate, because CGM focus is not really domestic. Great returns however.

So that leaves 10 managers out of the 25K+ in the Morningstar databse. Now, I only have data for this kind of comparison through Oct. However, including November would have helped him. From August 1st of 2000 until Oct 31st of this year, against this very select group of managers Hussman did quite well. He trailed five managers, four of them small or microcap managers. One was a midcap.

Given what he is doing(and three of them he may have pulled ahead of since then)trailing such funds given the style perfection they have enjoyed in the markets since then is pretty exceptional. He was ahead of all the others. So, even with that cream of the crop he was squarely in the middle of them. Even a 10% decline in the markets would likely put him ahead of all but the Bridgeway Ultra Small Co. fund and the Security Mid cap Value fund (which is pretty much a small cap fund despite the name.) Pretty elite company if you ask me.

If I widen the search to capture more mangers who may not have done as well as these, but might have done better during the time frame in question, he does similarly well. It is hard to construct a search (my screening tools will not allow me to search by the specific time frame) but from what I can see he finishes near the top of any search of funds I can run him against. Top 1 or 2% of all managers, and almost all of those ahead of him are small caps which have had a tremendous run. For funds with a more diverse or large cap orientation he beats them all. Every one.

Now I may be missing someone, as I said it is hard to do the search, but not the simulation, with the tools I have on my computer at the moment, but that is why I asked who you know that has done better. It was a legitimate question. Give me a few names that you think might have and I will gladly run the simulation. However, finding even 10 who have would put him at the top of his profession. Top 1 or 2% is pretty good in most careers.

[Bill aka NO DooDahs!]

"While you may be able to find three dozen diversified US managers who have outperformed him by some trivial amount since the inception of the fund, I would be curious who they are." Use a mutual fund screener and you'll find out. Get off your own laziness. FYI, there are 7 managers above 15% for 10+ years annualized with 10+ years of tenure, in the domestic diversified equity space. If one uses 5+ years tenure and 5 years of 15%+ then one finds 389 such funds. Hint, Hussy's fund (HSGFX) has five-year performance of 8.43%. I don't find 11.5% annualized over the last 7.5 years, or 8.5% over the last 5, impressive in the slightest. If you do find that impressive, it's likely you have set your personal investing goals much lower than I have.

"Nor is his fund supposed to be aggressive in a bull market. It is supposed to be more aggressive when he feels the risk reward ratio is favorable." Hmm, is the risk/reward ratio not "favorable" in a five-year bull market run? I would think that would be part of the definition of a bull market, for an intelligent person ...

[Paddock]

Whether he is overrated or not time will tell. I was just correcting your comment.

That being said, he does not run a bear market fund, but a strategic investment approach over time. That includes both bear and bull markets.

Over that time he has crushed the S&P 500. While you may be able to find three dozen diversified US managers who have outperformed him by some trivial amount since the inception of the fund, I would be curious who they are. Since inception he has averaged 11.42% annualized (far better than the S&P500 which is almost flat during that period and far better than the Russell 2000 as well.)

Nor is his fund supposed to be aggressive in a bull market. It is supposed to be more aggressive when he feels the risk reward ratio is favorable. For the last two years he has felt that the risk reward ratio is out of whack. It is not because he feels the market will necessarily go down.

Even if you could find a few funds which have outperformed over the life of his fund, their risk adjusted returns are almost certainly lower.

Of course how he will do in the future is unknowable, but if it is past returns you are basing it on, even allowing for the few funds which you claim have done better, and the far smaller number who have done better on a risk adjusted basis, past returns would put him at the very top of his profession.

Where you may have a reasonable point, is that given your appraisal of his approach his returns have been due to luck, that over most cycles it is unlikely to be as effective, the last few years of less exceptional performance are more indicative of his real skill level, etc. I disagree, but it is at least defensible.

As for the argument that he only outperforms in a bear market, I doubt that will be true over time, and it hasn't been true in at least one year, 2005. Still, since investments include both ups and downs, outperforming in down markets is a pretty nice thing to be able to do. It may not be what you want, but if it leaves one with more money I am not sure how it is a valid criticism.

It should also be noted that his stock picks, if you remove the effect of his hedges, would have outperformed both the S&P 500 and the Russell 2000, and put him at the top as well. The pattern of out performance would have changed, he would have done less well in 2000-2002 (but still at the top), but really well since. His total return however would have been lower. If you invested in his fund I am sure you would trade the (in relative terms year by year) better performance of his unhedged portfolio (he would have beaten most broad US index's in almost every year and over that time period has crushed the S&P 500 and the Russell 2000 with his unhedged picks) for the much less volatile fund that actually returned more money!

Once again, maybe not your style, but in return terms, pretty dang impressive.

[Bill aka NO DooDahs!]

Hussy needs a new model. His timing system is stuck on "bear" and he can't outperform in any other type of market. I've read his prospectus, and read some of weekly screeds, and "supposedly" his fund is supposed to be aggressive in a bull market.

The man's overrated. Severely overrated. I can find THREE DOZEN diversified (non-sector) domestic mutual funds with 10+ years of manager tenure and 10-year average annualized performance that outdoes the "Hussy." The only reason he's famous is the perma-bear sales job on his weekly "piss and moan."

[Bill aka NO DooDahs!]

Like I said, ping me when the "cycle is complete."

Looks like he missed on the 2005 chart, but if he's within a percent or so on the NAV, don't forget to include the divvy on the SPY, which is the alternate investment.

Your point about 2000-2002 is that he's running a bear market fund? LOL. Check BEARX if that's what turns you on.

[Paddock]

Bill,

"Last I checked, HSGFX was still on pace for a fifth straight year of underperformance relative to the SPX."

You checked wrong then, he didn't trail in 2005. He also didn't trail in 2000, 2001, or 2002 and is clobbering the S&P (about double) since inception. Also, he stands one down week by the S&P from not trailing this year. A small correction by the S&P would put him ahead since 2002 as well. So my guess is he will win from peak to peak, or trough to trough or however you wish to measure the cycle. His lead is too large or he trails by too little depending on which you choose.

[Bill aka NO DooDahs!]

Sell one, that's 1/5th the price of 1/2 the CDO. The rest is ...? Even knowing a price, the price is still a MODEL of valuation.

The point seems to me to be that one model replaced another, so it's not as if "reality" were replaced by a model ... considering that actually knowing the reality of how many jobs were created is both a logistical impossibility, and would be known (if at all) with certainty only many months after the fact, there's really no way of "knowing" whether the new model is better or worse at predicting "reality" than the old one was.

Send me an email when the "cycle completes." I'll be around.

[VennData]

A hundred million dollar CDO X has ten tranches. Each tranch will pull from the whole of the mortgages that continue to pay from the entire CDO. Each security from, say, the fifth tranch - down the waterfall - has an identical pro rata portion of all of cashflows as all of the other securities in that tranch, in other words, each security from the fifth tranch, 5X, is identical.

Sell one, that's the price. There's no model needed.

The main point here is about B/D... the real question is will the birth death have predictive value at large deltas in employment - up or down - over the non-adjusted (or prior "less-adjusted" methodology.) It will if the trend continues, it will not it the trend changes since more historical data is built into the "current" number. The annual adjustments noted by Ray may continue, and the real point by Barry R is that you don't get that adjustment until after it's too late.

Likewise, Hussman's underperformance for a given time frame doesn't prove he will underperform in the future. You can't yell scoreboard until the cycle completes.

[Ray]

Jeff, the Births/Deaths model replaced a different approach to minimizing non-sample error called the Bias Adjustment. The Bias adjustment was phrased out over a few year period ending June 2003. Effectively the bias adjustment took account of the difference between the BLS's estimate of payrolls and the underlying unemployment insurance related counts (the basis for the benchmark revisions). If the BLS was underestimating monthly payrolls the size of the additive bias adjustment was increased for subsequent periods. In a sense the bias adjustment was adaptive to the size of the benchmark revisions. The intuitive explanation of the Bias Adjustment was that it made some account of new business formations vs deaths of businesses. Thus, effectively a different version of the BDM was used before the current edition.

If I recall correctly the BLS made some attempt to add a cyclical element to the Bias Adjustment. This was a model based approach which incorporated the rate of change in GDP. I am not sure whether this improved performance or not. At the end of the day it is the size of the benchmark revisions that validate the performance of either the BDM or the old Bias Adjustment. The 2004 and 2005 benchmarks were relatively small, suggesting that the BDM had done a better job than the Bias Adjustment. the 2006 benchmark was a record size, and the preliminary 2007 is slightly on the high size of normal, suggesting that the jury remains out.

By the way i have visited your blog on a number of occasions, very well done.

[Bill aka NO DooDahs!]

The issue with CDOs is the vast amount of heterogeniety in the classes, combined with liquidity. To use a simpler model that focused more on transaction prices would be akin to using the last trade of Lennar to value your position in Centex.

CDOs can be relatively similar to each other, but are far from homogenous, and illiquid enough that finding a truly "comparable" trade is rare.

Last I checked, HSGFX was still on pace for a fifth straight year of underperformance relative to the SPX.

[oldprof]

Thanks to all for the comments. I wrote this from my hotel this morning, since I wanted to get something up before the payroll number comes out. I was hopeful of persuading Barry on this subject before his opinion gets its regular publicity splash.

Quite frankly, I have covered this so many times and so thoroughly, including commenting at Econobrowser on the cited charts, that I thought the issues were pretty clear.

I can see (Mike C is a good indicator)that I need to do something that reviews all of the details once again. The key point is that the "old method" inferred 2.5 million jobs each month. That is where my 2500% headline comes from -- 2.5 million versus the (average) 100K we have been seeing.

The main point is that there has ALWAYS been a birth/death model. The job creation was not "measured" or a product of statistical inference any more in the past than it is now.

Meanwhile, I asked two very simple questions of Barry. Since he has so many comments about the rest of the post it seems fair to request answers.

And by the way -- this is a straightforward question of methodology with a right answer and a wrong answer. It is not about mis-representation. My contention is that the BLS birth/death adjustment has improved the payroll jobs estimate.

Thanks to all for the spirited discussion, but I am really hoping to enlighten readers on this point.

I will try to take one more swing at this, since it is an important report and also a symptom of a lot of other things that happen in the interpretation of data.

Thanks again to readers and commenters.

Jeff

[Mike C]

"Take the brew-ha-ha over "mark to market" vs "mark to model" in CDOs. The simple fact is that a closing price or transaction price is nothing more than a MODEL of valuation, and not an actual valuation for any two participants. As a point of fact, if something traded for exactly $1,005.71, you can bet your asymptote that NEITHER party in the trade thought it worth exactly $1,005.71. After all, if you offered me exactly what I thought something was worth in today's time/place/circumstance, I wouldn't be willing to trade – I would be indifferent to the trade. A market price is BETWEEN two parties' estimations of value, which is why they trade – – and why the market price is a MODEL of valuation."

This seems to miss the point of the brouhaha (sp?). If one wants to rename the market price the "model" price, let's call it that. If we want, we could call it the "ledom" price. The point of the brouhaha is not what nomenclature we use, but that these CDOs are being valued on balance sheets at "mark to model" prices that probably aren't anywhere in the same ballpark of value that they would actually trade at between a buyer and seller in the market aka "model".

[Mike C]

"How you confused all these issues is, quite frankly, beyond me.

I'm sorry, but I am going to throw the yellow flag on this post, and move the ball back 30 yards for intentional misrepresentation.

This post was one of your weakest. I expect better of you. To be blunt, I am quite disappointed."

FWIW, I read the link provided for Barry's post, and it sure seems to me like he has some points here in terms of what he states above.

There was a mention elsewhere that Jeff was on the road but this post didn't state he sent this one in. Maybe it was written by an assistant? If not, not sure what to think. Usually all the posts here even when in disagreement with another blogger accurately represent what the other person stated. This seems like a departure from that norm.

"The market for busslhit is larger than the market for quality analysis"

Completely agree with this statement. Not too long ago, I read a blog post that analyzed the results of a particular mutual fund that I have a small position in, and the analysis can only be described as busslhit.

[Barry Ritholtz]

Speaking of playing fast and loose with the facts: I did NOT write "that the BLS has substituted modeling for measurement." That is a misstatement of what was discussed, and misses the key point of my post.

What I did write about BLS was "It has moved from a model highly reliant on measurement to a model highly reliant on more modeling;"

The accompanying chart (via Econbrowser) showed that 80% of recently reported new jobs attributable to B/D, as opposed to the establishment count. That compared with 20-25% in recent prior years.

I also make clear that in the BLS resolving one issue -- the missed jobs early in the cycle -- they created another problem -- overstating jobs late in the cycle.

Not only did you completely misread what I wrote, but you managed to disregard the chart that was referenced, ignored the phrase HIGHLY RELIANT ON MORE MODELING -- turning that into a completely different statement that I NEVER WROTE, and utterly failed reference the 80% datapoint.

Oh, and my headline was: NFP: Birth/Death Adjustments. That's neither inflammatory nor inaccurate.

Lastly, one math error: Hypothetical Job Growth (80%) versus measured (20%) is 2000%, not 2500% of the total.

How you confused all these issues is, quite frankly, beyond me.

I'm sorry, but I am going to throw the yellow flag on this post, and move the ball back 30 yards for intentional misrepresentation.

This post was one of your weakest. I expect better of you. To be blunt, I am quite disappointed.

[Bill aka NO DooDahs!]

Barry isn't exactly about accuracy in writing. Never has been. The market for busslhit is larger than the market for quality analysis, and Barry's proof of concept in that regard.

Many people take umbrage at the issue of using models, when in fact, they overlook the models currently in use. That is what is happening when pundits complain about the birth/death model, and that is the point of the post, in my opinion.

There are models in use all around us, and we constantly mistake them for truths, or overlook the fact that they ARE models in our everyday occurrence.

Take the brew-ha-ha over "mark to market" vs "mark to model" in CDOs. The simple fact is that a closing price or transaction price is nothing more than a MODEL of valuation, and not an actual valuation for any two participants. As a point of fact, if something traded for exactly $1,005.71, you can bet your asymptote that NEITHER party in the trade thought it worth exactly $1,005.71. After all, if you offered me exactly what I thought something was worth in today's time/place/circumstance, I wouldn't be willing to trade – I would be indifferent to the trade. A market price is BETWEEN two parties' estimations of value, which is why they trade – – and why the market price is a MODEL of valuation.

[Chris]

If Barry wanted to get it right, he would have fixed his errors after maybe the 10th time this blog pointed them out...

[Steve]

Barry is followed by so many through his blog and media quotations, so he has a special responsibility to get this right.

You're assuming he wants to get it right. Did you ever wonder why bearish blogs get so much more traffic on the web?