Requiem For A Princess

n Friday, March 2, Morgan Le Fay unexpectedly passed away in her sleep, the result of her advanced nine years. She went peacefully, having romped about with me just the day before. All who knew her will miss her, none more so than me and my son, Jonathan. Her ashes rest in a beautiful marble urn, engraved with her name, in my home study, a gift from Jon’s wonderful mom, Karen. And fortunately, the memory of Morgan will live in perpetuity through the Morgan Le Fay Dreams Foundation ¹, which I established and funded just last December, with a special focus on charities providing aid and services to children.

I was sad for a few days, but no longer: Morgan lived a wonderful life, brought me and my family great joy, and will live in our hearts through her continuing good works. It now brings me great joy when I sign checks embossed with her name. Happily, too, as if engineered by Morgan smiling through holes in the floor of heaven, a new bunny came into our lives only a week later – Bun Bun, who Jonnie and I have renamed Bunnicula. ²

A few days after Morgan’s passing, while waiting to get a sandwich at the deli downstairs in PIMCO’s building, I was flipping through the local newspaper and out jumped a big block ad, complete with picture, offering Bun Bun for adoption – by a retired couple, Al and Rhus. They had rescued her from a local park shortly after Easter two years ago, where she had been literally left to the wolves, presumably by parents who had misguidedly indulged a child’s request for a bunny.

Al and Rhus are moving to Arizona and wanted to find a home for Bun Bun, where she would be treasured as much as in their home. They interviewed me twice, once in their home and then in my home – they didn’t want just anybody to adopt Bun Bun, but a family that would love her. When I told them the story of Morgan and they saw her lovely "condo" outside all empty, the decision was obvious, to both them and me: the search for a new home for Bun Bun was over!

And she is an absolute delight, responding already to her new name Bunnicula, and insisting that she not spend all her time in the condo, but also inside with me in my home study, running about and tickling my feet as I read and write. You will get to know her better come December, when we plan to continue the annual tradition that Morgan and I shared of chin wagging on these pages about the year ahead. But that will be then. Right now, I’m much more concerned about getting the second half of 2007 right!

How Long Can the Fed Fight a Lagging Variable?

Net job growth is officially a coincident business cycle variable, as declared by the Department of Commerce. Note the word "net" at the beginning of that sentence, meaning the net of both job destruction and job creation. Job destruction is actually an official leading indicator, as

Thus, the featured statistics in the monthly Employment Situation Report, released (usually) on the first Friday of the month, are actually either coincident or lagging: net nonfarm payroll growth is coincident and average hourly earnings (half of the unit labor cost calculation) is lagging. The only thing that is actually and officially a leading variable in the report is the length of the manufacturing work week.  

None of this should come as a surprise to non-economists, who approach economic analysis more with common sense than with textbooks. And common sense tells you that employers do not preemptively hire and fire, but rather do so in reaction to something else. Similarly, merchants do not preemptively raise or lower prices, but rather do so in reaction to something else. And that something else is always and everywhere the state of order books, the stuff of future demand growth.  

Soaring order books are the fuel for both job creation and pricing power (inflation) and swooning order books are just the opposite. Always has been and always will be. Yes, I know that inflation is, as Milton Friedman famously declared, "always and everywhere a monetary phenomenon." But he also intoned that the lags between monetary policy action and inflation reaction are "long and variable." And in that context, it is useful to remember John Maynard Keynes’ wonderful commonsensical insight that in the long run, we are all dead. Thus, prudent monetary policy cannot be implemented in real time on the basis of Friedman’s doctrine, regardless of how true it may be in the long run.  

Prudent monetary policy must maximize a real-time utility function for society, otherwise known as cyclically fine tuning aggregate demand to aggregate supply potential, so as to minimize cyclical volatility in both unemployment and inflation. Or, if you prefer, prudent monetary policy must exploit and optimize the real-time trade-off between employment and inflation. Yes, for you economist readers, I know that there is no long-term trade off, that there is no long-term Phillips Curve ³.  

Well actually, I don’t know that, but will accept it as the gospel of our profession, because whether a long-term Phillips Curve exists or doesn’t exist doesn’t really matter for investment horizons that are relevant to what PIMCO does. The fact of the matter – both commonsensical and empirical – is that a cyclical Phillips Curve does exist, both because of price rigidities and frictions in the economy, and because as Alan Greenspan famously notes, human nature will never be repealed, and humans are inherently moody, both individually and collectively.  

Which is the proximate reason that asset markets are inherently given to cycles of boom and bust: human nature is inherently given to momentum investing, not value investing. Or, as Will Rogers famously said about stocks, "Take all your savings and buy some good stock and hold it till it goes up, then sell it. If it don’t go up, don’t buy it." Translating that advice to the bond market, I reckon he would say, "buy some duration and hold it ‘til the unemployment rate goes up. If it don’t go up, don’t buy it."  

But it would not necessarily always be good advice, because while the Fed rarely rewards duration with lower policy rates ‘til the unemployment rate goes up, the bond market does reward duration in anticipation of a rising unemployment rate, knowing that it will inevitably beget lower policy rates. Such has been the case since June 28 – yes, almost a year ago – when the 10-year Treasury yield peaked at 5.24%; today, it stands at 4.84%.  

Regrettably, however, it stood at 4.42% on December 4. So, while a long duration position has worked since the yield peak last June, it has most certainly not worked since the yield trough in December. Thus, a corollary of Mr. Rogers’ putative bond advice must be: once the duration market has discounted Fed easing, the only reason to hold it is if you expect the unemployment rate to go up. If you don’t, sell it.  

Here at PIMCO, we continue to expect the unemployment rate to go up. Thus, we are still (painfully, since December) long of duration, concentrated in the front end of the yield curve. And why are we still bearish on employment growth? 

First and foremost, unemployment is a lagging variable, notably of momentum in discretionary aggregate demand. And discretionary aggregate demand has been unambiguously decelerating in recent quarters, and not just in residential construction, as displayed in the chart below.

proxied by initial claims for unemployment insurance, a weekly series released every Thursday. In contrast, job creation is actually, and officially, a lagging indicator, as proxied by the average duration of unemployment (which is also proxied by the stock of continuing claims for unemployment insurance, a weekly series also released every Thursday). And finally, unit labor costs in manufacturing are also officially a lagging variable.  

The Great Puzzle

So why hasn’t the unemployment rate already risen? It’s the great puzzle, in the words of San Francisco Fed President Janet Yellen. The short answer to the puzzle is that the labor force participation rate has fallen, accounting fully for the drop from 4.7% to 4.5% for the unemployment rate over the last year. But this doesn’t make sense when you look at nonfarm payroll growth, which, again in the words of Ms. Yellen, has been gangbusters. The labor force participation rate is decidedly pro-cyclical, meaning that it goes up as tight labor markets induce new entrants into the labor market; and it goes down when soggy labor markets lead the discouraged unemployed to drop out of the labor force. So, the short answer to the puzzle is the right answer only if nonfarm payroll growth really ain’t gangbusters.  

And new research by both Ray Stone ⁴ of Stone and McCarthy and Sheryl King ⁵ of Merrill Lynch suggest this is indeed the case. Please refer directly to their research for the exhaustive details, but the bottom line is simple. Detailed data in the Bureau of Labor Statistics (BLS) Business Employment Dynamics (BED) release, which comes out with a two-quarter lag, show employment growth of only 19 thousand in 2006Q3, while the nonfarm payroll tally for that quarter was over 450 thousand. More recently, the BLS’s more timely Job Opening and Labor Turnover Survey (JOLTS) for April – last month! – showed job openings rose only 24 thousand, with this series essentially flat since last August. The JOLTS report also showed that new hires in March (this data subset is released with a one month lag) fell 29 thousand.  

Something smells more than fishy here. Not that I’m accusing the BLS of any skullduggery. None! Rather, it is a historical fact that nonfarm payrolls – before annual benchmark revisions, which continue for six years! – understate employment early in recoveries (leading to the inevitable contemporaneous label of "jobless recovery"), while they overstate employment late in expansions.  

And a key reason is that the BLS, while very good at counting heads at existing firms, must make an assumption, in real time, about the birth-death rate for firms, so as to estimate the net gain/loss in jobs as firms open and close, a never-ending feature of a capitalist economy. In the early years of expansions, the birth assumption systematically is too low and the death assumption is systematically too high, which results in "jobless recoveries," which turn out to be not-so-jobless recoveries upon revision. The exact opposite holds in the late years of expansions and particularly in recessions. Such is the case, it would appear, at present. 

Thus, in contrast to last August, when the job tally for the year ending March 2006 was revised up some 800 thousand, a stunningly large revision, the opposite is likely to unfold in this August’s benchmark revision for the year ending March 2007. Not to suggest, I hasten to add, that a downward revision equal to last year’s upward revision is in the cards. The honest answer is that we don’t know how big it will be. But available data, notably the BED and JOLTS data, point squarely to a downward revision. 

So what, you say. Economists always bellyache about the quality of the data when they go against their forecasts. This is true. It is also true, however, that poor data can make for poor policy making, if and when the data is taken to be religiously true. This is particularly the case if the data is known to be lagging data of the business cycle, as is the case with the unemployment rate. Acting on the data, or refusing to act because of it, is the stuff of policy mistakes, sometimes known as recessions. 

Bottom Line
Recent months have not been fun, either at home or the office. Morgan died and the markets have unwound Fed easing expectations, as the unemployment rate has defied PIMCO’s expectations of a climb.

But life goes on.  

The time for mourning of Morgan has morphed into the time for celebration of her life through her Foundation. And the case for Fed easing is actually more robust now than it was last fall, as inflation ebbs ever closer to 2% and job growth slows, even as the unemployment rate stays stubbornly low on the back of a falling labor force participation rate. 

In the fullness of time, there comes a time when time is full.

Paul McCulley
Managing Director
May 23, 2007
mcculley@pimco.com