check out the chart at the post Revisions to Payroll Employment at Economist's View of how payrolls were initially reported each month, and to what they've been revised to in successive months. Needless to say, downward revisions have been significant.
Canadian jobs data was released today, but I thought I'd take a closer look at last week's U.S. jobs data.
As bad as the official stats are, its clear that they don’t do the situation full justice. I’m not suggesting there’s an Obama administration conspiracy to make things look better than they really are (or not as bad as they really are). But, nonetheless, there have been changes made to BLS calculation methodologies over the years, going back to the Reagan and Clinton administrations, that do make some of the reported results a little rosier than they ought to be.
Take for instance one pet peeve of mine, the birth-death model, which the BLS includes in its establishment survey of payroll changes. Basically, what they do is survey a bunch of companies to find out how many jobs were gained or lost at each, and they add all that up. They then extrapolate this to the full work force. However, recognizing that such a survey is less than comprehensive, and, in particular, does not capture any information from brand new firms that have just started up, nor from dead firms that have just expired, the BLS has added a modeled-component to their payroll calculations. Basically, based on past trends, they guesstimate how many net new jobs (new jobs added less old jobs lost) were created each month from the birth and death of new and old firms.
In March, BLS assumed that 114k more new jobs were created than jobs lost from the birth and death of firms; in February, the assumption was a gain of 134k. These assumed numbers are pretty close to the numbers assumed in each of February and March of 2007 and 2008, and are in fact rather higher than the averages for those two months from 2000 through 2009. This doesn’t seem particularly realistic given the prevailing environment. Its not as if the BLS doesn’t admit that their model will be inaccurate at turning points in the economy – they do. But, nonetheless, there’s a clear bias introduced here into the results.
Despite that bias, the employment picture is rather grim. Over 7 million jobs (on a non-seasonally-adjusted basis) have been lost since the peak in November, 2007; the loss is 5 million on a seasonally-adjusted basis. The average monthly loss over the last six months is 619k and over the last year is 400k; if you back out the birth-death contributions, the average loss is 459k.
Chart 1: Monthly Payrolls Data, with 3-mth (blue) and 12-mth (red) moving averages, since 1998 (January was the worst month to date, with a loss of 741k jobs):
On a longer-term basis, job losses over the last year are as bad as they’ve been since the late 1950s.
Chart 2: Historical Non-farm Payrolls (seasonally-adjusted) Growth YoY (red line shows how current level of -3.7% compares historically; grey bars are recessions)
What you can note from the chart is that past recessions usually started as soon as YoY employment fell below 1%, but, more to the point, didn’t end until YoY employment headed upwards.
Clearly, the downward trend in payrolls doesn’t look like it will turn around anytime soon (not given the state of credit conditions, nor the increase in inventory-to-sales ratios, and the continuing climb in jobless claims). On a peak-to-trough basis, seasonally-adjusted payrolls are down 3.7% from the peak, which is also the worst result since the late 1950s.
Chart 3: Historical Non-farm Payrolls (SA) Growth, Decline from Peak
Of course, the above charts include the B/D contributions in the NFP results. And those have been systematically positive even in this terrible environment. (In the third chart below, the negative brown bars mean that annual job growth was negative while B/D contributions were positive; for the years 2003 to 2007, it seems that between one-third and all of job gains were derived from the model, i.e. from companies not surveyed.)
Chart 4: (a) Annual Monthly Change by year (blue - NSA; green - SA); (b) Birth-Death Model's Average Monthly Contribution by Year (peaked at 88k in 2006; was 60k in 2008); (c) Percentage Contribution of B/D Model to NSA NFP
Its interesting to note from the chart below that the seasonality of the B/D contribution has been pretty impervious to the deteriorating environment. So we can expect a nice contribution again in each of March, April and May before those contributions fall off for the rest of the year.
Chart 5: Monthly Birth-Death Model Contributions, since April 2000
Chart 6: Historical NFP Growth (NSA) (ex-B/D model contributions) (latest level is down 3.5%)
Another pet peeve of mine is the focus on the official (U-3) unemployment rate. At the current rate of 8.5%, it has more than doubled from the 2000 low of about 4%. As bad as that is, it still misses out on all the people who are no longer counted as part of the labour force, and also does not account for all the people who are still working but are now part-time or have otherwise had their hours cut back. The broadest measure of unemployment is the U-6 rate, and it has climbed to 15.6% (on a seasonally-adjusted basis; over 16% NSA), as shown below.
Chart 7: Unemployment Rates, Monthly (green U-3; red U-6, both seasonally adjusted)
Chart 8: Unemployment Rates, Annual, SA
For more perspective than the unemployment rate gives you, consider just how many jobs have been lost since the peak at the end of 2007. But, more than that, consider also the increase in the people not just who have been added to the rolls of the unemployed, but those who are no longer counted in the labour force (NILF: Not In the Labour Force).
Chart 9: Average Monthly Changes in Number of Employed and Number of Unemployed and NILF since December 2007 and in last six months
Chart 10: Monthly Number of Employed (lhs) and Unemployed/NILF (rhs)
Once again, though, even that portrays less than a complete picture due to under-employment; there are a number of people who have been fortunate to have retain some level of employment, but have been cut back from full-time to part-time. The number of full-time workers has fallen now by over 8 million, while the number who now work part-time has increased by about 3 million (for the net loss of jobs of about 5 million, on a seasonally-adjusted basis).
Chart 11: Full-Time Employment and Part-Time Employment
Labour data is considered a lagging indicator. While that surely remains true (i.e. the economy will show other signs of rebounding before labour does), labour trends impart their own impact on the economy, particularly in a recession that is being consumer-led, unlike some past inventory-cycle recessions. In this case, the consumer retrenchment can't help but be re-inforced by the employment trends, as overly-indebted consumers are now also increasingly income-constrained consumers.
p.s. sorry for the awful quality of the charts; I'll fix them as soon as I have a chance
No comments:
Post a Comment