Hussman's recession call is still not validated

- by New Deal democrat

Last week I wrote that both ECRI's initial recession call and also John Hussman's recession warning criteria had been invalidated. Oversimplifying somewhat, Hussman's 4 crieteria were: (1) credit spreads wider than 6 months before; (2) the S&P 500 lower than 6 months before; (3) the ISM manufacturing index under 54 simultaneously with less than 1.3% YoY employment growth; and (4) a yield curve of less than 2.5%. As of last Friday, not only was the S&P not lower than it was 6 months before, it was actually at a 6 month high! Further, not only was the ISM manufacturing index above 54, but employment growth was also more than 1.3% higher YoY. I've also pointed out that the yield curve element of Hussman's formula was in place for twenty years running during the 20th century, simultaneous with the strongest GDP growth in the last 100 years.

In closing, I said that Hussman should at least explain why he believed his recession call was still valid. Put another way, what is the "off" switch for the above criteria, if it is different from the "on" switch?

This week Hussman spent a large part of his weekly market comment defending that call. His defense rests, as I understand it, on two grounds: (1) one or more criteria was violated in 2008 and the recession warning, obviously, was still valid; and (2) there is no "off" switch for the criteria, but rather, once "on," a cornucopia of bearish evidence may be invoked, and entirely different criteria, e.g., a positive ECRI growth WLI, signals the end of recession.

Before I go further, I should emphasize that Hussman defended his metrics on their merits, with no ad hominem attack on me. For my part, although I still disagree with him, he did make some valid points, and none of what follows should be taken as a personal attack on him. In fact, I think his shorthand indicator briefly summarized above is very helpful.

That being said, idea of an indicator that switches "on" but never "off" strikes me as not intellectually rigorous. Beyond that, if the ECRI indexes are the determinant of an indicator, I should just go directly by them and cut out Hussman as the middleman. That in the interim a cornucopia of evidence may be selected (cherry-picked?) in support of a conclusion strikes me as inherently subjective and unreliable.

Since I wasn't satisfied with Hussman's explanation, I decided to examine the 3 non-yield curve criteria on my own to determine what should cause them to switch from "on" to "off." (With long term yields under 2.5%, that element is likely to remain in effect for a long time to come). The results indicate that if a recession were to happen now, it would still be unprecedented under Hussman's own criteria.

As an initial note, Hussman's claim that the S&P 500 criteria was violated in May 2008 is not correct. The closest it came was 1426.63 on May 19, 2008, only 7 points below its level of 1433.27 on November 19, 2007. That being said, my research indicates that it is not uncommon at all for that index to be higher than 6 months previously either shortly before or after the onset of a recession. Similarly, it is not uncommon for credit spreads to meander higher or lower than 6 months before during all but the most serious economic turns.

What is uncommon -- in fact, almost non-existent -- is for either the ISM manufacturing index criteria or the employment criteria to be violated. Generally speaking, once the ISM index falls below 54 in advance of a recession, it continues to fall under 50 and only rises back above 54 after the recession is over. Similarly, once job growth falls below 1.3% in advance of a recession, it typically only rises back above that level well after the recession has passed.

In fact, each has occurred only once, and not simultaneously. In particular, in all cases but one, the YoY employment percentage change was falling on a 6 month smoothed basis in advance of every recession since the second world war. The sole exception was in 1953. Similarly, the only time the ISM index fell below 54 within 1 year of the onset of recession but rebounded back above it was in 1959 for two months. These are shown in the graph below (in which the ISM index is normed so that a reading of 54 on the index = 0, and payroll YoY% growth is also normed so that 1.3% YoY growth = 0):



To show you the full record, here is the same graph for the 1970s and early 1980s recessions:



And here is the same graph from 1989 to the present:



Note that in every other case, not only did the ISM index continue to fall, but the YoY% change in employment was also declining going into a recession. There is simply no precedent for a recession occurring while both the YoY% change of employment is improving, and simultaneously the ISM index rebounding above 54. Put another way, whether a recession happens or not, under the set of criteria Hussman himself has established, a precedent will be set. Only in retrospect will we know the answer.

Read More

Is job growth getting ready to accelerate?

- by New Deal democrat

Just about the only drawback in Friday's employment report is that at the rate employment has been growing since it bottomed 2 years ago, it will take possibly a decade to make up all of the jobs lost plus population growth since 2007. Even job growth of 250,000 a month would be downright tepid for the 1980s or latter part of the 1990s. To substantially accelerate real, population adjusted job growth, we need monthly improvement of 300,000+ or even 400,000+ jobs.

And there are signs that we may be on the verge of achieving at least the lower part of that range in the very near future.

First of all, let's look at updated graphs showing the relationship of the rate of initial jobless claims with the unemployment rate. Here's the long term graph:



The rate of initial jobless claims has a long history of tightly leading the unemployment rate. That means that continued improvement of the former should lead to continued improvement in the latter.

Here's the close-up of the last several years (note: rate of initial jobless claims scaled on the right) after their divergence in 2009:



Note the several month lag of the unemployment rate. The continuing decline in initial claims means that we should expect a further decline in the unemployment rate in the next few months.

But not only is there good evidence to support the argument that the unemployment rate should decline, there is also evidence that the rate of employment growth should actually increase.

Initial jobless claims peaked at the end of March 2009. Since May of that year there has been a fairly constant ratio of changes in weekly initial claims to monthly job growth of roughly 1 to 2.75. That is, for every 10,000 drop in average weekly new jobless claims in a given month, the number of jobs added to the economy has improved by 27,500.



But as Jeff Miller of A Dash of Insight points out, monthly payroll growth is a function of both the number of layoffs and also the number of hires. If hiring accelerates, there will be accelerating job growth even at the same number of layoffs.

There are three reasons to believe that relationship might be shifting for the better.

First of all, during recessions there is typically a steeper loss in hours than in the number of jobs. During recoveries, initially hours increase faster until they make up the difference, and thereafter the two series move in tandem. Here's a longer term graph showing the relationship:



In earlier post-World War 2 recessions, aggregate hours made up all of their relative lost ground quickly, and then jobs and hours grew in tandem. Not so for the 1991 and 2001 recessions: it took until 2006 before aggregate hours made up all of their relative lost ground from the 2001 recession. Only in the last year or so of the last expansion did the two measures move in tandem.

This recession and recovery have been no different:



As the above graph shows, however, at the current rate aggregate hours will make up all of their relative losses by summer. In fact, measured strictly in private jobs, the measure has fully caught up as of this month. From that point forward job growth should accelerate to match growth in hours.

As an aside, remember that long leading indicators like housing permits, bond prices, and real money growth have been improving almost relentlessly for about a year now. Shorter leading indicators like stock prices, initial jobless claims, and durable goods, have joined them in showing improvement for the last few months. Real retail sales, which I have previously described as the "holy grail" single best predictor of future job growth, have also continued to improve, albeit slowly.

Payroll growth and aggregate hours, by contrast, are coincident measures of the economy. Thus with improving leading indicators, there is every reason to expect aggregate hours to continue their trend of improvement. That means job growth should start moving in tandem within the next 3-6 months.

Secondly, the household survey has been on a tear for the last seven months, adding 2.252 million jobs, or over 320,000 a month, during that time [Note: without the census adjustment, the numbers would be slightly lower, at 2.020 million and 290,000 a month]:



The household survey has a much smaller survey sample, and so any given month's change is inherently less reliable, but there is evidence that the household survey leads the establishment survey at inflection points, as indeed it did at the end of 2009. While the surge since last June might be noise, it looks very much like an inflection point that is beginning to show up in the establishment survey.

Third, there is one specific sector in which it appears significant improvement is finally taking place: construction. As the below graph shows, while manufacturing jobs picked up almost immediately (in fact, this is the best improvement in 20 years save for the tech boom), construction jobs continued to languish, finally hitting bottom 12 months ago:



There are excellent reasons to believe that construction employment will improve, and improve at an accelerating rate.

This next graph shows the relationship between housing permits (blue), private residential construction (red), and construction jobs (green):



Note the sequential relationship: housing permits lead residential construction, which in turn leads construction employment.

Now let's look at a close-up of the last several years:



With the exception of the distortion from the $8000 housing credit, the relationship has continued. Permits bottomed in early 2009, then construction spending, and finally construction jobs. More importantly, for the last year housing permits have continued to increase. Private construction spending began to follow several months later. Construction jobs began to turn up more meaningfully several months ago (+14,000 in January). There is every reason to believe that construction spending will increase, and that jobs in construction will continue to increase, at an increasing rate, as well.

Just like the tsunami in Japan last year, there are any number of factors which could derail this analysis. But since the unemployment rate is likely to decline further, job growth is likely to accelerate to the trend of increasing aggregate hours, the household survey suggests acceleration is already occurring, and the construction sector is finally participating in meaningful job growth, there is every reason to believe that the rate of job growth is going to accelerate in the coming months.

Read More

1953 Employment and Wages

The recession that started mid-year was (obviously) a very negative development for employment.

However, the unemployment rate didn't start moving higher until the fourth quarter, with a big spike in both November and December.

The big reason for the drop is the huge cratering of goods-producing jobs, which declined by more than 700,000 over the course of the year.

Service producing industries followed s somewhat different trajectory.  First, we see increases from January until October, but a pretty sharp drop in November and December.

And overall, we see a drop in government employment, largely as a result of the drop in defense spending related to the Korean war cease-fire.

As a result of the weakening employment picture, real DPI took a hit.

On a continuously compounded annual rate of change basis, real DPI contracted in the third and fourth quarter of 1953

Which we all see in absolute values above.

Below are the graphs from the Economic Report to the President

Read More

1951, Employment and Income

 

Thanks to the economy being on a war footing, unemployment was incredibly low - coming in below the 5% most economists consider to be full employment.

 Total non-farm employment grew by over 1 million jobs during the year.

However, despite the need for durable goods, total goods producing actually added less than 200,000 total jobs during the year.

The real job gains occurred in the service industries, which saw about 900,000 jobs added, and

government employment, which saw large increases (around 300,00).

Thanks to near full employment, disposable personal income increased:

On a continuously compounded annual rate of change basis,we see a slight gain in the first quarter, a big gain in the second and then more moderate increases in the third and fourth. 

And on a percentage change from the previous year, we see strong increases in the second, third and fourth quarter.

The following charts are from the 1951 Economic  Report to the President, and are included because I think they're really interesting.

Read More

1950: Employment and Income

1950 was a very good year for employment.  Consider the following charts:

The unemployment rate dropped over 2% - moving from 6.5% to a little over 4% by the end of the years.  And the employment growth was split between manufacturing and employment:

Manufacturing employment increased over 2 million while

service sector growth increased over 1 million.

Government employment also increase about about 400,000.

As a result of this activity, we see real disposable personal income increase on a YOY basis of over 9%.

Below are some charts on wages and savings from the 1951 Economic Report to the President that show the above information, but in 1950s econ style.

Read More

Notes on Friday's employment report: positive trends intact, actual good news in household survey

- by New Deal democrat

Friday's employment report continued the trends of employment as continued to initial jobless claims that were established in March 2009.

First of all, here is the updated graph comparing population-adjusted initial jobless claims with payrolls:



This relationship was historically very close beginning in 1967 but broke down in 2009. For the last two years, however, the new relationship established then has continued. A change in the population adjusted initial jobless claims rate is mirrored in the unemployment rate, frequently with a 1 or 2 month delay:



I've also noted that if we were to be heading towards a "double-dip," we should expect the relationship between the number of initial jobless claims and employment growth to change, whereby in the net many more jobs are lost per any given rate of initial claims. Although there may be some very slight drift in that direction since April (brown points), it hasn't happened in any meaningful way:



The longer term leading relationship of initial jobless claims to the unemployment rate (comparing the last 6 months' average with the average over the same six month period one year prior) as measured by Thumbcharts is also intact:



Finally, back in December 2009, Prof. Paul Krugman said:

I thought it might be useful to create a sort of benchmark for the level of job growth that would really count as good news.…
[L]et’s set a … modest goal: return to more or less full employment in 5 years –which means seven lean years of depressed employment. To keep up with population growth over those 7 years, the United States would have had to add 84 times 127,000 or 10.668 million jobs. (If that sounds high, bear in mind that we added more than 20 million jobs over the 8 Clinton years). Add in the need to make up lost ground, and we’re at around 18 million jobs over the next five years — or 300,000 a month.

So that’s a useful benchmark.

(emphasis mine)

With that in mind, here's a graph of monthly gains in civilian employment from the Household Survey since its bottom:



No, it isn't the more generally reported Establishment Survey data, but the Household Survey does tend to lead at inflection points, and for the last 4 months, the average employment gain in the Household Survey has been 321,000/month, with the lowest being 277,000. I'm not nitpicking the Professor, but it's worth noting that this does meet Prof. Krugman's threshold criteria for actual good employment news.

Read More

Putting a stake through the heart of an employment canard

- by New Deal democrat

Since the site's occasional Doomoron troll has posted one of his usual central falsehoods, claiming "Bonddad and NDD were dead wrong in predicting the recovery of employment," it's time to set the record straight.

Bonddad is a big boy and can speak for himself, but as for me , in September 2009, as the economy was still shedding over 200,000 jobs a month, I wrote a 6 part analysis looking at indicators which in the past had led the jobs turnaround and concluded:

Based on my analysis above, November or December are when I believe that turning point will be reached, plus or minus one month in either direction. Let me be the first to acknowledge that this is not a scientific truth or certainty, but a best estimate based on a logical review of existing data with a long history that accommodates both traditional and "jobless" recoveries. Nevertheless, at least in terms of payroll growth, the analysis in these six installments cause me to predict that this will not be a "jobless recovery" for long.

So what are the facts? Here's how private jobs look beginning in August 2009:



As of the most recent revisions, the turning point was in February/March, only 2 months later than my window. I was off by a grand total of 63,000 jobs lost in January and February. In fact, before the most recent revisions, the 2010 current data showed my prediction to be spot on. By the way, the data looks the same for all jobs if you take out census hires and fires, but there's no graph for that.

Since then, every single month has shown job growth - clearly not enough compared with the 8 million lost jobs from 2007, but consistent growth nevertheless. I'll let you decide if that means I was "dead wrong in predicting the recovery of employment" or not.

As for our troll's hero, the Pied Piper of Doom, on January 13, 2010 he claimed that the only job growth in 2010 would be that of Census hiring. OOPS!

And while I was writing the above in 2009, he said:

overall unemployment rates are expected to enter into the double digits (and I'm only talking about the Bureau of Labor Statistics' U.3 index, not the more accurate, and much greater numbers posted in their U.6 index), and remain there throughout 2010

In fact, the highest unemployment reached in 2010 was 9.8% OOPS!! OOPS!!

He also predicted in March 2010 that the unemployment rate would not fall below 9.7% at any point in 2010.

In fact, unemployment fell below 9.7% in May and remained below 9.7% for all but two of the next 7 months. OOPS!!! OOPS!!! OOPS!!!

Facts are very inconvenient things. The above are just 3 of the over 100 false prophecies made by the Pied Piper of Doom about which that I have kept book. His acolytes have to have very big memory holes down which to pour and forget about all of those false prophecies.

By no means do I claim infallibility. But by now the facts have put a stake through the heart of the canard repeated by our troll.

From Bonddad:

On August 22, I wrote the following:

Let's tie all of this information about unemployment together.

1.) The trend of initial unemployment claims is down. We see this in the 4-week moving average, along with other data such as the Challenger job cut report, the decrease in the number of seasonally adjusted mass lay-off events and the decrease in the number of seasonally adjusted people laid-off in mass lay-offs. Because the number of initial claimants is decreasing, we can expect a slow improvement in the number of people unemployed for various lengths of time.

2.) The structure of unemployment is showing improvement. The number of people unemployed for less than 5 weeks has been decreasing since the beginning of the year. The number of people unemployed for 5-14 weeks and 15-26 weeks dropped in the latest jobs report. The longer term unemployed are still increasing, but given the drops in the other metrics this number should start to show a decrease within the next 4-6 months.

Now -- that leads to the final question: when will the jobs come back? In order for that to happen we need to see at least one quarter of a positive GDP and probably two. That means the soonest we can expect a drop in the unemployment rate would be the a few months from now and that is only under the rosiest of scenarios. The most likely possibility is it won't be until the end of the first quarter of next year before we start to see a ticking down (and that assumes a stronger rate of growth than I think is going to happen).

So, in the meantime what should we do?

1.) Make sure the longer-term unemployed are given benefits and other help to ease their suffering.

2.) Ask ourselves is there a way we can better implement the stimulus money to increase the rate of GDP growth and thereby see a faster drop in GDP?

As NDD has pointed out above, we started to see job increases in the first quarter of 2010.

In addition, if you look at the way the recovery has unfolded, it has done so more or less in line with what I wrote in the First and Starts Expansion. And, I should also add that both NDD and I have advocated for the creation of a new WPA, along with highlighting the importance of maintaining our infrastructure. See here, here, here, here, here and here.

None of this will of course appease the trolls, but then nothing ever will.

Read More

Pre-BLS Numbers, Pt. 2

Earlier today NDD emailed me and asked me to add some charts to the story below (he was away from his computer). I thought the story was fine, but he pressed on. So, I pulled up the ADP employment report which I had never really looked at. So first, let's look at the ADP/BLS comparison.

It's important to remember one big difference between ADP and the BLS: ADP does not include government employees. But with that in mind, notice that in general ADP tracks the BLS' numbers. They seem to jockey for overall position, but in general move in the same direction.

Let's look at ADP in more detail.



First, notice the ADP and BLS reports generally track each other.



The BLS has service employment increasing last month by 58,000. In the latest ADP survey we see an increase of 12,000. So these series agree that service industries are starting to hire again.

As NDD points out, this is in contrast to the ISM service report which had this to say in its latest report:

Employment activity in the non-manufacturing sector contracted in December for the 23rd time in the last 24 months. ISM's Non-Manufacturing Employment Index for December registered 44 percent. This reflects an increase of 2.4 percentage points when compared to the 41.6 percent registered in November. Four industries reported increased employment, 12 industries reported decreased employment, and two industries reported unchanged employment compared to November. Comments from respondents include: "Decrease in occupancy equals decrease in employment" and "Continuing to aggressively manage costs down."

The industries reporting an increase in employment in December are: Other Services; Retail Trade; Finance & Insurance; and Public Administration. The industries reporting a reduction in employment in December — listed in order — are: Arts, Entertainment & Recreation; Real Estate, Rental & Leasing; Mining; Management of Companies & Support Services; Information; Wholesale Trade; Transportation & Warehousing; Accommodation & Food Services; Construction; Utilities; Health Care & Social Assistance; and Professional, Scientific & Technical Services.

The logical conclusion from the ISM report in relation to the BLS/ADP report is the 4 industries that are hiring did so in greater numbers than the 12 that were reporting a decrease in employment. Note that this ISM report is for December (the holiday season) and one of the areas of hiring was retail employment. However, also note that finance and insurance were hiring in the ISM survey and that "professional and business service" employment increased by 86,000 in the BLS survey. In short, it appears as though the professional services sector is very important right now.

In contrast, manufacturing employment is still in the dumps:



But it's important to place this number in historical perspective with two other data points. First, here is a chart of total manufacturing jobs going back to 1979:


Notice that after the last recession we lost about 2.5 million manufacturing jobs that never came back. However --


Overall production (the red line with the arrow underneath it) continued to increase during the 2001 expansion. In other words, the US was making more "stuff," only with fewer people. I would expect that trend to continue after this expansion as well. This means we have two choices regarding the displaced manufacturing workers. Either get more manufacturing business going or retrain them.

However, the BLS and ADP data is in contrast with the ISM Manufacturing survey:

ISM's Employment Index registered 52 percent in December, which is 1.2 percentage points higher than the 50.8 percent reported in November. This is the third month of growth in manufacturing employment, following 14 consecutive months of decline. An Employment Index above 49.7 percent, over time, is generally consistent with an increase in the Bureau of Labor Statistics (BLS) data on manufacturing employment.

Seven of the 18 manufacturing industries reported growth in employment in December in the following order: Apparel, Leather & Allied Products; Printing & Related Support Activities; Petroleum & Coal Products; Paper Products; Transportation Equipment; Machinery; and Electrical Equipment, Appliances & Components. The eight industries that reported decreases in employment during December — listed in order — are: Wood Products; Textile Mills; Miscellaneous Manufacturing; Nonmetallic Mineral Products; Plastics & Rubber Products; Food, Beverage & Tobacco Products; Chemical Products; and Fabricated Metal Products.

In order to coordinate this data with the BLS/ADP data, the logical conclusion is that the number of employers shedding jobs (8 industries) are doing so at a faster rate than those that are adding jobs (7 industries).

Finally, we have initial unemployment claims:

Back in the late spring/early summer, this was one of the first economic numbers that got me thinking the worst was over. Well, it is still moving in the right direction.

Taking all the data into consideration we get the following points:

1.) Service are employment is increasing, despite the numbers from the ISM non-manufacturing survey.

2.) Manufacturing employment is probably going through another round of attrition. That is,, manufacturers are using the current downturn to again shed workers.

3.) Initial unemployment claims continue to move lower.

Read More

Initial Jobless Claims Continue To Move Lower

From the Department of Labor:

In the week ending Nov. 7, the advance figure for seasonally adjusted initial claims was 502,000, a decrease of 12,000 from the previous week's revised figure of 514,000. The 4-week moving average was 519,750, a decrease of 4,500 from the previous week's revised average of 524,250.

Here is the relevant chart:



Here is a chart from the St. Louis Federal Reserve that shows the long-term trend for the 4-week average. This chart has not been updated with the latest reading:





This was the first piece of data that clued me into the fact that the recession was ending. The first time I noticed the continued decline in the 4-week moving average was maybe June(?) (there was a Barron's article that highlighted the trend). Since then the 4-week moving average has continued to move lower. Notice that the 4-week moving average is a good indicator for the end of a recession as evidenced from the previous recessions.

From NDD:

The 4 week moving average, at 519,750, is the lowest reading in a year, since the equal number on November 19, 2008. It appears the weekly jobless claims data are getting reading to test my hypothesis that net new jobs will be added if the jobless claims simply stay at this level for a few more weeks. In that regard, at the time of the 501,250 new jobless claims reading in 1990, payrolls lost 160,000 that month and 211,000 the next. In 2001, the new jobless claims high of 489,250 coincided with payroll losses of 325,000 that month and 292,000 the next. This year, we have already seen in August new jobless claims in the 560,000-570,000 range coinciding with a payroll loss of 151,000.

I have also surmised that, while Calculated Risk is correct that there will be little seasonal hiring this year, there will also be little seasonal firing this year either (believe it or not, layoffs increase dramatically as you move from September through New Year’s Day). Here is how the UNadjusted jobless claims look from their September lows through the first full week of November in the last 4 years:

2006: +86,500 weekly new claims

2007: +80,300 weekly new claims

2008: +203,100 weekly new claims

2009: +118,300 weekly new claims

In other words, so far my hypothesis is largely correct.

Read More