Oops...so much for that gap down. Pictures and discussion below the fold.
First, here are the unemployment durations. Kind of a bump up across the board. This looks like a bit of noise to me, for several reasons. The rise from 6.1% - 6.2% can generally be attributed to a rise in the 0-4 week category, and, if anything, signs of recently initiated unemployment have been very low. If this isn't noise, it could be from an increase in Quits, which would be a positive sign. But, I don't think a monthly change in Quits would have much effect on unemployment. Quits usually move quickly back into employment.
Here are unemployment flows. Whereas the very positive April unemployment report had the fortune of all six net flows moving in its advantage, this month had five of the six moving against it. These series reflect some of the noise in the monthly employment data. Looking at reported unemployment compared to the forecast unemployment rate coming from unemployment insurance claims, in April, the forecast called for a fall of 0.2%, but instead there was a fall of 0.4%. In July, the model called for a fall of 0.1%, but instead there was a rise of 0.1%. I think the flows data suggests that a lot of this variation is noise. I continue to see the declining insured unemployment level as a signal of declining total unemployment.
Here are a couple other graphs comparing unemployment insurance and total unemployment.
The first one is one I post frequently. Again, this month looks like an anomaly, which is part of the reason I expected a better number this month. Either this month is an anomaly, or all of the past five months were an anomaly. Obviously, the likelihood is that we see a snap back in the next couple of months which takes us under 6.0%.
The other graph compares the unemployment rate and the ratio of continued unemployment insurance claims to initial unemployment insurance claims. This ratio gives us a measure of how difficult it is for job losers to become re-employed. By this measure, the labor market has rarely been more robust. To think we have politicians still grandstanding about extended unemployment insurance. Another interesting issue here is that this measure indicates short unemployment durations. But, average total unemployment durations are still somewhat above previous lows. I attribute this to demographics. But, it's interesting that this doesn't show up in the unemployment insurance data. I think the interpretation this would have to lead to is that the longer unemployment durations that we see among older workers normally come from workers that quit or are fired, but that older workers who are laid off don't tend to have longer unemployment durations.
More Data on EUI
I have discussed extended unemployment insurance before as a suboptimal policy because with each added week, it is likely to be covering older and older workers who would normally have higher durations, partly because they have more savings and more options. But, what this graph suggests is that the older workers who do qualify for unemployment insurance might not normally have this characteristic. I've been characterizing EUI as a policy that extended the unemployment durations of workers who would normally have had somewhat long durations to begin with. But, what generous EUI might have done is enticed older workers who would normally not have had especially long durations, but given the economic downturn, and the availability of EUI, they were enticed into extending their unemployment durations much as older workers who are fired or who quit do. Look how high this ratio got during this downturn. And, look how the bottom dropped out when EUI was ended. Keep in mind, this ratio is just measuring regular insured workers who have been unemployed less than 26 weeks, so this is just capturing the slower reemployment happening during the first 26 weeks of unemployment. You can see the tremendous number of insured unemployed in 2009, which drops off quickly. EUI led to the persistence of unemployment, but if workers were tactically using EUI, you would see a similar drop in unemployment in 2011, when that large cohort of unemployed workers from 2009 would have timed out of EUI. But, total unemployment doesn't mirror the number of unemployed workers under 26 weeks. This is another signal of the workers that may have been enticed by EUI into an unhelpful situation. They weren't able to quickly move back into the labor market when EUI ran out, and their exit rate from unemployment is now much, much slower than newly unemployed workers.
We are now far enough out from the EUI termination to also get some confirmation of its age skew. This first graph shows how older age groups had uncharacteristically high unemployment during the EUI period. As the graph here shows, in the 7 months since the program ended, unemployment among 45-54 year olds has now dropped back down to normal range, coincident with the end of the program.
In the next graph, I created a trend unemployment rate for each age group, based on a regression with the 25-34 year old unemployment rate from 1990 to 2007. The trends move around zero with noise until the introduction of EUI. The 35-44 year old group has intermittently high unemployment until 2013. The 45+ age groups tended to remain elevated 0.5% to 1% for the duration of the program. 45-54 year olds' unemployment has quickly dropped back to trend since the end of EUI and 55+ year olds are declining, but still slightly elevated.
This suggests at least two things. (1) There were an unusual amount of older workers who were incentivized into long term unemployment. They would have amounted to over 500,000 workers, or about 0.3% of the total labor force. And (2) since the unusual level of unemployment among the older workers has mostly disappeared after the program ended, this suggests that the approx. 1.2 million very long term unemployed workers who had timed out of the program and are now only very slowly re-entering employment do not skew older.
More on the state of employment
The measure of 3 month exit rates for workers unemployed for longer than 14 weeks took a step back this month, from last month's very high 43% back to just under 37% this week. The actual rate is probably somewhere between these two numbers, and is rising smartly.
Finally, here is the graph of the average unemployment duration for all workers, from the BLS (in blue), and the average duration for workers unemployed for less than 26 weeks, which I have estimated from BLS data. Both of these measures continued to show a decline that started at the end of 2013, when EUI was terminated. The average duration number is mostly high due to the remaining approx. 1.2 million very long term unemployed workers, who account for most of the remaining reported excess unemployment. This should slowly continue to decline. The average duration under 26 weeks is a measure that looks at the normal part of the labor market, including workers who might be entering unemployment today. It is slightly higher than it was in 2007, but I think demographics is responsible for most of that shift. Older workers have less employment churn and longer durations. The under 26 week average duration is probably near its cyclical low, and I expect it to level out soon.
Generally, a model of unemployment that combines the exit rates of recent unemployment with a continued linear decline in very long term unemployed continues to point to an unemployment rate in December of around 5.5%. So, I expect to see a rebound from the stated rate this month. The gap down will come....
Wages have finally dropped to appropriate levels
One final graph is a graph I have been following that compares changes in real hourly wages to the unemployment rate. If I use average hourly earnings of production and non-supervisory workers, adjusted with the Core PCE inflation rate, I get a decent long term relationship. I have included the graph with the axes both ways. The first one is the way I have tended to draw it, but the second one might be easier in terms of imagining the relationship.
During this cycle, there was a very unusual and persistent deviation from the relationship. Wages have been way too high since 2008. Some of this can be attributed to the sticky wage problem that comes from the low inflation environment and also to pro-cyclical federal policies that would tend to increase average wages, including (broken record warning) EUI and minimum wage increases.
This graph is only updated through June, because that is the most recent PCEPI reading, but nominal wages were pretty flat this month, so this measure should continue to moderate.
Wages were way out of whack in the early part of the crisis. By the end of 2011, wage growth started to track the normal relationship, except that wage growth was persistently too high. Beginning in late 2013, that gap began to disappear, and with this month, YOY wage growth should cross the long-term trend. I suspect if you are a regular reader, you might be wondering what change there could have been in labor markets at the end of 2013 that would have caused average wages to moderate.....
So, the good news is, we are back in normal territory. We should start to see real wages grow as unemployment continues to fall.
One final note on thinking about Jobs, Wages, and Economic Growth:
I was listening to NPR today. They were talking about the employment report. And, one of their experts said that one detail in the report was that employment growth had slowed in the hospitality area. She mentioned the recent trend of reducing wait staff by putting tablets on the table. She said this was a concern. A concern, she said.
The dynamism of markets is so difficult to conceptualize. The tendency to think of things as creating or destroying jobs is so debilitating to an understanding of markets. First, think of it this way. We currently have about 43 million jobs for workers with a high school education or less and about 3 million of these workers who are unemployed. How did we luck out? Why don't we only have 20 million jobs for this education level? Or, on the other hand, what if we had 60 million of these jobs? What would we do? How is it that the amount of jobs we have for each type of worker is always within a few million of the number of workers we happen to have? Even when education trends change drastically, or whole genders up and join the labor force over a generation or two, there always just happens to roughly be the right number of jobs out there. Millions more people go to college than have ever before. How in the world did we think jobs were just going to appear for all these college graduates? But they did.
There are so many moving parts in a complex economy, and so many dynamic processes that find an equilibrium between the kinds of people who want jobs, the kinds of services that people demand, and the kinds of jobs that will match those two pools, that thinking of an economy as a ticker tape of jobs lost or created is a fool's game. It would be like measuring the health of a rain forest by counting the number of flowers. You'd see orchid petals wilting, and wonder how long the forest could stay healthy at that rate of decay.
The only reason that this development would be a concern is if frictions in the labor market make it difficult for displaced workers to become re-employed. Sticky wages are one factor. This is where employer mandates are problematic. It's not a matter of gaining or losing jobs. It's a matter of attaching a new cost to that worker. In a number of scenarios, this will tend to lead to disemployment. Stickiness in nominal prices, for a number of reasons, may mean that it temporarily becomes more difficult for those workers to be re-employed. (There is much more going on there than just an unemployed person being stubborn. It is very hard to conceptualize all of the dynamics of markets.) So, inertia and inefficiencies in the labor market might mean that the new obligation increases the total value to the employee while the market is in disequilibrium. It will also lead to some disemployment. But, eventually, as equilibrium is re-established, wages will be the equilibrating factor. So, eventually, as obligations are increased, wages will become a smaller portion of the payment made for the employee. Maybe your paradigm calls for employees to have some services paid in kind instead simply being paid in cash. But, in the long run, it is clearly the employee that is paying for the services. This should be conceptually obvious. Note that corporate profits don't seem to have declined in the face of decades of increasing non-cash obligations.
So, it is wrong to speak of these things as creating or destroying jobs. The number of jobs, in the end, will be roughly equal to the number of workers. Now, if eventually, we decide that for every hour of work, $20 of incense has to be brought to an altar to appease the gods, workers who are capable of producing $25 of value will probably decide that it's not worth the trouble for the $5 they are able to keep. So, on the margin, there will be changes in labor force participation. But, these changes are subtle. And, the subtlety of these wage and employment effects make it difficult to identify them in real-time data.
But, I'm getting off track...
What if there is no policy change, and the market wage for that job is simply rising? Let's imagine that this is the case with the use of tablets in restaurants. There will always be some frictions in the labor market. Although, I will note that unemployment durations for young workers tend to be much lower than for older workers. In any case, progress requires some dislocation.
But, this is exactly the kind of process that brings about our interdependent outcomes in the economy. A healthy economy - an economy with few frictions and high utilization - means that wages are growing. It means that revenues and profits are high.
The end of the story is that the restaurant has tablets and fewer employees, and those workers that used to wait tables are now doing something else that nobody was doing before, because we needed people to wait tables. So, now, the economy is producing something new. How do we know those new jobs will appear? For the same reason there aren't just 20 million jobs available for those workers today. (This is only true if the dislocation comes from growth. There may not be acceptable jobs for workers who are carrying around mandates and regulatory obligations that reduce the equilibrating wage. I wonder if the reporter who was concerned about the tablets is also concerned about health insurance mandates and minimum wages.)
It's wrong to think of rising wages as a product of bargaining power. It's wrong to think of the investment in those tablets as a response to rising wage costs. It's certainly wrong to think of new forms of efficiency in a growing economy as a concern. (Or, at least it is wrong to think of it that way in an economy unburdened by fresh frictions in the labor market.)
In a healthy economy, the causation isn't linear. We focus too much on temporary disequilibria - the temporary boost the employer gets when they are the first restaurant in town to install the tablets; the temporary pockets of unemployment; the difficulty an employer has finding qualified workers because they are reasonably wary of raising wages at the first sign of difficulty. Much like the CEO who thinks he is moving production to Malaysia because the wages are low, we look at all of these situations and assume the causation started at the agent we happen to see in front of us. But the causation is a mess. But, the good news is that in a functioning economy, it's mostly good news. Generally, everybody has more. Of course, sometimes disequilibria are so strong that, say, an automotive worker in Detroit who is riding a wave high above equilibrium from geographic and path dependent barriers to competition can find himself later sitting way below equilibrium in a city hollowed out by decades of rent seeking made possible by those barriers, with no quick solution for undoing the damage. But, in many cases, potential losers are overestimated in a far too common static conception of the economy. And we end up swatting at bees while we prop up wilted flower petals.
We are the 100%.