The bifurcated jobs recovery

- by New Deal democrat

Karl Smith at Modeled Behavior yesterday wrote a couple of excellent posts wherein he divided the US workforce into two parts: (1) "goods and government" vs. (2) private service jobs. He pointed out that:

Goods and Government are what we might have thought about as backbone jobs. These are police officers, fire fighters, school teachers, factory workers, construction workers. When you think of a stereotypical 1950s American, they are doing one of these jobs.

He noted that it is this group - public employees, construction workers, and manufacturing workers - who have seen little or no improvement since the bottom of the recession. Here's his graph, showing the number of "goods and government" worker jobs in red (left scale), and private service jobs in blue (right scale):



That got me thinking. One of the signature differences between the "Great Recession" and other post-WW2 recessions is how much it impacted private service jobs. To compare, here are the same two categories of jobs, "goods and government" (red) and private service jobs (blue) showing the relative impact of the two previous worst recessions since WW2 - 1974 and 1981-82 - on those categories, by showing the number of jobs in each gained or lost since their peaks just before the 1974 recession:



You can see that private service jobs were barely scratched in 1981,and actually increased during the 1974 recession!

Now let's look at job losses and recoveries for the two sectors in the Great Recession. The difference between this graph and Karl's is that both categories are shown on the same scale, and show the number of jobs lost since their respective peaks immediately prior to this recession:



Unlike every other post-WW2 recession, private service jobs were slammed in this last recession. As many private service jobs were lost - about 4.5 million - as were lost in the goods and government sectors. Further, the private service sector is having close to a V-shaped recovery. Three-fourths of all private service jobs lost during the recession have already been made up. If the current rate continues, by the end of this year there will be no remaining job losses in that sector (although it will not have caught up with population growth).

Virtually the entire 5,000,000+ continuing job losses are in manufacturing, construction, and government. Manufacturing has long-standing issues of automation and offshoring, and it can be debated how much of its long-term losses are due to each factor.

Construction and government job losses, however, are another matter entirely, reflecting a colossal and unnecessary waste due to ideological political failure. We need something like $3 trillion in infrasctructure repairs to bridges, sewer and water mains, and the electric grid. We have several million idle construction workers. And we could finance the infrastructure repairs that would make this country so much more competitive and pay for itself multiple times over (see, e.g., Erie Canal and interstate highway system) by issuing debt at 2% in the open market. Further, whatever one's opinion regarding government inefficiency, police officers, firefighters, and teachers are not by any stretch its primary source -- and yet those are the government workers who have been most impacted.

If it were not for this utterly unnecessary waste, we would not have to pay for so many food stamps, we would not have to fund so much extended unemployment compensation, we would have the benefit of the professional and constructed infrastructure, and there would be probably a million or more persons employed in Karl Smith's "backbone jobs." Those employees would be supporting their families, and paying withholding, property and income taxes, with all of the resulting positive multiplier effects throughout the economy.

Instead here we are, almost three years since the bottom of the recession, having made up less than half of the total jobs lost, tolerating a bifurcated jobs recovery.

Read More

Getting deep into the weeds about Jobs, Jobs, Jobs

- by New Deal democrat

It's been quite a while since I looked at detailed metrics of future job growth, something I devoted a lot of time to a couple of years ago. With a divergeance in forecasts between recession vs. continued growth, this is a good time to take another look. Some of the metrics have performed better than others. Many continue to support optimism, but at least one is downright ominous and may be telling us that revised jobs data will show substantial job losses from earlier this year.

1. V shaped real retail sales and industrial production recoveries vs. jobs:

One point I frequently made is that this is a "bifurcated recovery", where manufacturing and sales are performing much better than job and income growth. Although we've had a recent slowdown in some ISM series, the description of a "bifurcated recovery" is still valid.

Real retail sales and Industrial production are still in V shaped recovery mode. Real retail sales have recovered 80% from their trough, and industrial production two-thirds:



Comparing those with private jobs (red) and total payrolls (green), we can see that the percentage losses in sales and production were steeper, and have made up nearly or more than all of their ground compared with jobs. Meanwhile, private jobs have regained only slightly over 30% of their losses. When government employment is added for the total jobs picture, less than 25% of the losses have been regained:



2. Comparing improvements in aggregate hours and jobs:

Another point I have frequently made is that aggregate hours worked are recovering faster than new jobs. Since more hours were lost than jobs during the recession, if past was prologue then we would have to wait for aggregate hours to regain their lost comparative ground before job growth would match the growth in hours. This is still the case:



If the trend continues then by about next summer aggegate hours (red) will have made up all of their comparative losses with jobs (blue) and we can start to expect job growth to fully reflect growth in hours.

3. Comparing real retail sales with jobs:

The closest thing I found to a "holy grail" leading indicator for future job growth when I took a thorough look over 2 years ago was real retail sales. Real retail sales tended to lead turning points in jobs by about 4 to 8 months. Since sales are still rising, we should expect continued job growth over the next few months as well. (This metric has also recently been touted by Prof. Karl Smith of UNC - Chapel Hill at Modeled Behavior).

I also found that over the last 40 years, the YoY% growth in real retail sales, divided by two, gave a reasonably close forecast to YoY% growth in jobs, at least over a longer horizon if not every month. Since real retail sales were averaging 6% YoY growth at the end of 2009, this led me to expect strong job growth in 2010. It didn't happen, although measuring by private jobs, the metric is at the moment almost in perfect alignment:



When we add government jobs into the mix, and compare real retail sales with total jobs (green), the metric still falls considerably short:



This is yet another indication of just how significant government job losses have been to the relatively poor jobs recovery. At the same time, because real retail sales are a leading indicator for jobs, this reinforces that we should expect to continue to see positive job growth in the economy, with private jobs at least being added at something like a 2% YoY rate.

4. Comparing initial jobless claims with jobs added:

In 2010 I thoroughly debunked the idea that we needed 400,000 or less in initial jobless claims to be consistent with job growth. It simply makes a lot of difference how deep the recession is, and also the pattern declining into a recession is quite different from the pattern during a recovery. I pointed out half a year ago that if we were to descend into a "double dip", I would expect to see a break in trend in the scatter graph comparing these two series, with a new trend line to the left of the recovery trend line developing. Here is the updated graph (using private jobs vs. all jobs to avoid the 2010 census distortions) , with the last 6 months' data in brown:



No break in trend happened. Since this scatter graph ends with October, it doesn't show the decline in the 4 week moving average below 400,000 in the last couple of weeks. Should that continue through the end of November, I would expect a very good November employment report, with something like 175,000 private jobs being added.

5. Okun's Law

Okun's law is actually a rule of thumb that holds that for every 2% YoY increase in GDP, there should be a 1% decline in the unemployment rate. Generally speaking, 2% YoY GDP growth equals no change in unemployment. A 4% GDP increase gives you a 1% decrease in unemployment. Contrarily a 0% YoY change in GDP gives you a 1% increase in unemployment.

I make use of a corollary, which is the YoY% growth in GDP minus 2% approximately equals the YoY% change in job growth 3 to 6 months later. Here is the graph of this relationship going back 65 years, and it has ominous implications:



Since 1948 there has never been a period of 2 or more quarters where YoY GDP% growth was under 2%, that has not equated actual YoY job losses in the next few months. If this relationship holds true now, then contra all of the other above data, we should have already been seeing outright job losses, and they could continue through the winter.

As I said, this contradicts virtually all of the previous indicators we have discussed. A possible explanation comes via Jeff Miller of A Dash of Insight, who informed us yesterday that the BLS's Dynamic Business Report of actual job data collected from the states showed that in the first quarter of this year only 250,000 jobs were created, rather than the 500,000 previously reported. If this revision is applied to all of the 2011, it would mean that the pathetic job gains of this past summer turn into outright significant losses. Not only would this tend to vindicate Okun's law, it would affect all of the data sets above. For example, the scatterplot graph above would probably then show that we did indeed break trend in the direction of a "double-dip."

6. Forecasts of the unemployment rate:

Finally, let's update a few metrics forecasting the unemployment rate. The premise here is that initial jobless claims are a leading indicator of the unemployment rate. The best way to measure initial claims, however, is to adjust for population, which is done in this first graph:



So adjusted, the recent initial claims levels aren't so bad. In fact, they are better than most readings during the last 50 years.

This metric had an excellent record for predicting the unemployment rate several months out -- until this recovery. It predicted an unemployment rate of under 7%, and needless to say were are far above that:



Taking a closer view of the last several years, it appears that the big disconnect occurred in 2009, when initial claims steeply declined, yet unemployment remained stubbornly high. Since then, the two series have tracked one another rather well. This suggests we should see the unemployment rate drop slightly to about 8.8% in the next few months:



Finally, here is a slightly different metric from Thumbcharts. This compares the last six month period with the same six month period one year before. In this longer term metric, initial claims also have an excellent record predicting the unemployment rate -- although like the metric above it shows that the YoY decline in initial claims considerably outpaced the decline in the unemployment rate a year ago:



This metric likewise predicts a continued decline in the unemployment rate over the next few months.

Summary

Continued job losses in government continue to have a depressing impact on job growth during this recovery, causing distinctly subpar growth compared with previous recoveries. Undoubtedly as I have pointed out just a few days ago, that housing until recently did not participate in the jobs recovery also has had an effect.

At the same time, most of the above metrics suggest that we should see continued job growth in the months ahead, and a continuing decline in the unemployment rate compared with a year ago. If present underlying economic trends continue (as to which there is obviously no guarantee), then by next summer or so we may see stronger job growth as the deficit in aggregate hours is completely made up.

The contrary indicator is Okun's law, which suggests we should already be in the throes of actual job losses. It is possible that we will find when the jobs data is revised that we did lose a significant number of jobs earlier this year, but that the situation will improve going forward from here, which would be more consistent with all of the data sets above.

Read More

The Jobs Report big picture in 5 graphs

- by New Deal democrat

On Friday I showed that the trend of increasing jobs reports was intact, when one bears in mind that for 12 of the last 13 months, the final revision was on average about +50,000 higher than the initial report. Even after the yearly benchmark changes that was still true for 9 of the twelve months of 2010.

Today let me update 5 graphs that I have been running over the last few months in some cases, and for over a year in others, to show some "big picture" correlations with the jobs report. All of these show data that has reliably led the jobs picture by weeks (the first graph) or months (the next two), or has led the unemployment rate by months (the last two).

First of all, here is the scatter graph of initial jobless claims (left) and monthly jobs gained/lost (bottom) for the last two years. Blue is all private jobs, red includes government jobs as well (except I have excluded the March through August 2010 period which was distorted by census hiring and firing):



As you can see, there is a nearly linear relationship. So long as initial claims continue under 400,000, we should expect robust monthly jobs reports (with a very big +/-125,000 variance, however). It is also important to note that we have been shedding government jobs for the entire two year period, without any "double-dip." While laying off government workers now is insane policy, and will be a drag on the economy, it is safe to say that it alone will not be enough now or during this summer to cause a renewed recession.

Next, let's look at jobs vs. GDP. This is a series I have been running for well over a year. Despite the complaints that GDP does not measure the happiness of Main Street, the fact remains that it is an excellent way to predict YoY jobs over the next 6-12 months. Subtract 2% from the YoY percentage growth of GDP, and you will be very close to the YoY% of jobs growth a few months later. Here is that graph now:



As you can see, it remains an excellent predictive tool.

Third, let's look at real retail sales vs. jobs. This is another series I have been running for well over a year. It is updated monthly, and so gives earlier signals of turning points than GDP. While it remains an excellent tool for the direction of YoY job growth, it has not performed so well in the last year:



My suspicion is that it is not performing so well this time around because the retail sales do not include those sales associated with new home sales and home improvements, compared with other recoveries. Nevertheless, if real retail sales continue to predict 3% annual job growth for an extended time, I expect job growth to trend in that direction.

In summary, the above graphs strongly suggest that we will continue to see jobs reports that at very least keep up with population growth in the coming three to six months.

Now let's look at two graphs comparing initial jobless claims with the unemployment rate. First, courtesy of Thumbcharts, is a graph comparing the average of the last last 6 months of initial claims (blue) and the unemployment rate (red) over the same period, with the equivalent 6 months the year before. With the exception of the double-dip 1980-1982 recessions, this has an excellent track record.



As the readings from last August and September disappear, the initial claims average will continue to decline YoY, and this predicts that there will continue to be a substantially lower unemployment rate than the same period a year ago. While this might not mean further decline in the unemployment rate from here, it very strongly argues against any significant increase back above 9% in that rate.

Finally, let's look at initial jobless claims as a percentage of the population vs. the unemployment rate. This is the graph I began running about 3 months ago when I was totally surprised by the close and long-lasting fit (although there has been a slight drift upward over the long term in the unemployment rate vs. claims):



This graph argued 3 months ago that the unemployment rate was far too high compared with initial claims. Since then the unemployment rate has declined 0.9%!!! As population-adjusted initial claims has consistently led the unemployment rate for almost 50 years, this graph suggests that further declines in the unemployment rate in the coming months are likely. If so, the dramatic drop in the unemployment rate could be the surprise economic story of 2011.

The question remains, however, how much of the decline in the unemployment rate is an unalloyed good and how much has to do with participants dropping out of the work force. I will get down into the weeds on that subject in my next post.

P.S. One note of final emphasis: I'm not just running the above graphs because of their performance in the past. Not infrequently in the last couple of years at DK I would get comments on the order of, "Pretty graphs. But they have nothing to do with reality. Everything sucks, and it's going to get worse." And yet, the "pretty graphs," which are just easy to digest representations of complex data, showing how A has relentlessly led to B over a very long time, have prevailed. If the pretty graphs kept kicking my ass, I would seriously reconsider my position. In summary, they still lead now, and are giving us a good look at what we should expect from the monthly jobs reports in the next few months.

Read More

Construction jobs in recoveries - a comparison

- by New Deal democrat

In my last couple of posts, I discussed how the recovery in manufacturing is giving us very few jobs relative to the strength of thats recovery, compared with 1946-1983 recoveries, to the tune of close to 1,000,000 jobs.

In contrast, let's look at construction. There first couple of graphs look at housing starts (blue) vs. construction jobs (both residential and non-residential), in the 1948-1983 period:



and for the last two recessions and recoveries:



Note that construction jobs lag housing starts by about 1 year, a pattern that continues now. (Also note, in the above graphs I divided construction jobs by population, to give the percentage of people working in construction, for ease in viewing purposes only).

Now let's take out housing starts, and look at the change in raw numbers in construction jobs during recoveries. Here are the post-World War 2 recoveries:



And here are the last two recessions and recoveries:



In the post-war period, about 200,000 construction jobs were added per year once growth started. None at all were added after the dot-com bubble until the housing bubble caused about 250,000 a year to be added.

By contrast, with the crash in the housing market, that has been limping along a bottom for 2 years, construction jobs are still being lost, albeit very slowly.

So, not only is the current boom in manufacturing failing to create many jobs in comparison with past booms, causing a population-adjusted shortfall of perhaps 1,000,000 jobs; but the continuing housing bust is costing another shortfall of 200,000 or more jobs.

That leaves the private and government service sectors to pick up the slack. We know the public sector is bad. I'll discuss the ISM non manufacturing report later.

Read More