More on Unemployment - Analyst Blog

The picture on the duration of unemployment showed a few hopeful signs in the July report, after a long series of unrelentingly dismal reports. We actually had a very significant drop in the number of mid duration unemployed this month, while the number of long term unemployed continued to soar. There was a minor increase in the number of short term unemployed.

The first graph (green) shows the history of unemployment by the duration of being out of work. This tracks people through the unemployment process since the people who are unemployed for under five weeks this month will either find new jobs or move into the 5- to 14-week group next week, and then in a few months move into the 15- to 26-week group, and then finally move to the ranks of the long term (>26 week) group. All but the short term group remain well above the highs seen in previous downturns.

The declines in the middle two groups was pretty dramatic, with the 5- to 14-week group falling by 12.5% or 509,000 and the 15- to 26-week group falling by 15.5% or 536,000. The long term unemployed continue to be a very significant problem, rising by 13.3% or 584,000 in one month alone. However, this rate of increase is likely to slow in future months given the lower numbers of unemployed in the pipeline.

It has to be emphasized that being out of work for more than six months is a very different experience than being out of work for less than a month, both financially and psychologically. If you are able to find work in a couple of weeks, it is more like an unplanned vacation, and the odds are that the job you get is at something like your old salary. Being out of work for over six months will sap a person’s sense of self esteem, not to mention wiping out any accumulated savings they may have had. When these people do find work, it is usually at salary levels well below what they were previously earning.

It is important to keep in mind that the number of short term unemployed does not really change that much with the economic cycle, it is the number of medium and long term unemployed that define the employment aspect to recessions.

The number of unemployed for less than 15 weeks tends to start to decline pretty much right around the time a recession officially ends. The number of people unemployed for more than 15 weeks, and particularly those who are out of work for more than 26 weeks, tends to continue to rise long after the recession is over.

Indeed, in the last two recessions, the bulk of the increase in long term (26+) unemployment cam AFTER the recession had ended. One could look at this two ways. The optimist would say that the decline in the shorter term unemployed (despite a small blip up in the under 5 week group this month) could be indicating that the recession is over, and that the continued rise in long term unemployed is irrelevant to the dating of recessions. Those inclined to see the glass half empty would point to lots more pain to come in the form of more long term unemployed.

The next graph presents the same data as the contributions to total unemployment, with short term unemployment on the bottom and long term unemployment on top. Note that total number of unemployed dwarfs that of any of the last three recessions, but then again the population is bigger.

The current problem with long term unemployment is quite literally unprecedented (with the possible exception of the Great Depression, but we don’t have good stats for back then). The second graph looks at the ratio of short term to long term unemployment and at long term unemployment as a percentage of total unemployment. By both measures, things have never been worse in terms of long term unemployment (also see the graph in POST LINK TO 1ST post on UE, for yet another measure of this).

Since 1978, long term unemployment has averaged 43% of long term unemployment, and prior to this cycle, the highest it had ever gotten was 78% (3/83). It now stands at 154%, up from 137% in June and 121% in May. As a percentage of total unemployed, long term unemployed have averaged 16% since 1978 (and even lower if we look at the all the data back to 1947) and, prior to this cycle, the highest it had ever gotten to was 26% (6/83). Today it stands at 33.8%, up from 29.0% in June and 27.0% in May.

The decline of the number in the middle of the unemployment duration spectrum is responsible for the divergent behavior of the mean and median duration of unemployment as shown in the final (red) graph. Both remain far above where they reached in previous cycles, but the median reversed sharply this month, dropping to 15.7 weeks from 17.9 weeks in June. That is still the second highest level on record (May was the third highest on record at 14.9 weeks) but it is a sharp break in the trend.

Since it is impossible to be unemployed for less than 0 weeks, the average duration of unemployment is always going to be higher than the median. However, it is unusual to see the two measures head in opposite directions. Note that the average person who is unemployed has been out of work now for 25 weeks, and that regular state unemployment benefits run out after 26 weeks.

Another thing to note about the graph is that the average duration of unemployment has been rising even in the good times. If we treat the double dip recession of the early 1980s as a single episode, then each expansion since the 1960s has failed to bring the average duration of unemployment down below that of the previous expansion. This was particularly apparent in the last expansion, where the best that was registered was worse than the worst aftermath of the mid 1970s recession. A similar, but less distinct pattern can be seen in the median numbers as well.

Long term unemployment is as significant as total unemployment, or can be seen as the most significant part of unemployment. When people are first laid off, they may cut back their spending somewhat, but they still have an income from unemployment insurance, and they can either dip into savings or use the credit card.

As the time wears on, the savings get depleted and the cards get maxed out. If they reach the point where they don’t have unemployment benefits, then it becomes almost a forgone conclusion that they will default on their credit cards, which is not good for firms like Capitol One (COF) or American Express (AXP).

If their house is underwater, they will either mail the keys in or simply stop paying the mortgage and wait for the sheriff to show up (in areas with lots of underwater houses, that can be years). That substantially reduces the cash flow for banks with big exposures to the mortgage market, like Bank of America (BAC), and is a disaster for Fannie (FNM) and Freddie (FRE), which are 100% exposed to residential mortgages. Given our 80% ownership stake in the GSEs, it is not good for the taxpayer either. But most of all it is bad for the unemployed person; he or she is highly likely to fall out of the middle class and never find his way back there.

Another way of looking at the jobs data is not by the percentage of people in the labor force who are unemployed, but by the percentage of people who actually have jobs. Not everyone is in the labor force since we don’t expect two-year olds to bring home a W-2, and most people still plan (?) to retire after they reach 65. Thus the jobs picture becomes a question of what percentage of the population is actually in the labor force, which is largely a question of demographics and social norms (do wives get paid employment or do they stay home with the kids?).

It was the combination of the baby boomers entering the labor force, and women working outside the home, that drove a huge increase in the participation rate from the mid 1960s to the end of the century. Since then that trend has started to reverse. The change in the participation rate is not particularly tied to the economic cycle, other than some minor flattening during bad times.

On the other hand, the percentage of people who are working is very much tied to the economy. Part of the reason that unemployment got so high in the Reagan recession was that it happened in the middle of the steep uptrend in the participation rate. In other words, the economic treadmill was running faster back then. The economy had to produce bigger percentage gains in employment just to keep the unemployment rate steady.

Note that the normal cyclical rise in the percentage of people in the economy was extremely anemic in the last expansion, and that a good deal of the improvement in the unemployment rate was due to a decline in the participation rate (mostly demographics). The decline in the percentage of people working has been particularly steep this time.

In this cycle the percentage of people working topped out at 63.4% (12/06), far below the all time peak reached at the height of the Clinton expansion (64.7%, 4/00). As recently as a year ago it stood at 62.3%. Since then it has declined to 59.4%, its lowest level since April of 1984, and the same level it was at in August 1978. The 4.0% decline from the cycle peak is huge.

By contrast in the double dip recession of the early 1980s, the percentage peaked out at 60.1% (2/79) and took four years to hit its low of 57.1% – a decline of just 3.0%. We have seen almost that big a decline in just the last year. Over the next several years, demographics (retiring boomers) are likely to push the participation rate down (but the wealth destruction from the housing collapse will keep many of them working far longer than they had planned, thus mitigating this). This will tend to bring the unemployment rate down.

However, unless the reduction in the unemployment rate is accompanied by a rise in the percentage of people working, then it is simply a case of the treadmill running slower, not of making forward progress.