Thursday, November 13, 2008

How high will the unemployment rate go?

The Bureau of Labor Statistics reported that there is a lot of slack building in the economy: the October unemployment rate rose to 6.5% and the total number of jobs lost in 2008 grew to 1.2 million. The question is, for how long will the labor market contract (unemployment rises)? If the contraction is anything like the contraction in 1957-1958, the unemployment rate could rise to 9.5%. You thought that I would say 1982, right? Well, read on.

I look to previous cycles for a clue as to how severe the 2008-2009 labor market contraction could be. This cycle does not have to be like previous cycles, but here we go.

The chart lists the historical unemployment rate spanning the years 1951-2008. The noticeable peak occurred in 1982, where the unemployment rate hit 10.8%. This is certainly the highest compared to all labor market contractions since 1951, but the unemployment rate was already above 7% when the contraction started. So it’s a little unfair to compare the current 6.5% to 10.8% when the unemployment rate was just 4.8% in February 2008.

This Table (click to enlarge) should be familiar to my readers, as I used a similar version in this post. The 1981-1982 contraction produced the longest consecutive monthly decline in nonfarm payroll, 17 months (column 3), and the most job loss, 2.8 million (column 4). Furthermore, the 1957-1958 and 2001-2002 labor market contractions each saw around 2.2-2.3 million jobs lost over 10 and 15 consecutive months. Finally, the 1953-1954, 1960-1961, and 1990-1991 contractions were benign in comparison of accumulated job loss.

Let’s put the job loss in relative standards since the labor force grew over the 1953-2008 period. The better comparison across time would be the accumulated job loss as a percentage of the payroll (column 6), rather than the accumulated job loss in levels (column 4).

Column 6 lists the implied job loss for the 2008 contraction when the accumulated job loss for each contraction is compared to the magnitude of the nonfarm payroll payroll at the time (calculation is column 5 times the current size of the payroll in 2008). The most severe labor contraction occurred in 1957-1958 with an implied 5.9 million jobs lost in just 10 short months. The second largest payroll decline occurred in 1953-1954 with an implied 4.8 million jobs lost; the third is 1981-1982 with an implied 4.3 million jobs lost, and followed by the 1960-1961 contraction, the 2001-2002 contraction, and the 1990-1991 contraction respectively.

The chart above shows the implied 2008-2009 unemployment rates across the same five cycles; the implied nonfarm job loss (column 6 in the table) is used to extract the unemployment rate for each cycle. Using the method highlighted below (Appendix below my name), the peak 2008-2009 unemployment rate implied by the previous cycles would be: 9.5% for the 1957-1958 contraction, 8.7% for the 1953-1954 contraction, 8.3% for the 1981-1982 contraction, 7.7% for the 1960-1961 contraction, 7.3% for the 2001-2002 contraction, and 7.0% for the 1990-1991 contraction.

I expect that the unemployment rate will rise to 7.8-8.3%: not as bad as the job destruction seen in the 1957-1958 contraction, but somewhere between that seen in the 2001-2002 and 1981-1982 contractions.

Rebecca Wilder

Appendix: How I calculated the unemployment rate going forward. I use the average monthly labor force growth rate over the last 3 years to calculate the new labor force in each month. Furthermore, I allow for a 17 month contraction (the longest consecutive labor contraction in column 3 of the table) to calculate the number of people that are unemployed each month. Going forward (post October), the number of people that are unemployed is a simple average over 7 months (the remaining months in the forecast) of the difference between the implied job loss for each cycle (column 6) and the 1.2 million jobs that have already been lost to date. The unemployment rate for each month is the percentage of the labor force that is unemployed.


  1. Therefore, the end of recessions can be forecast when the unemployment numbers start to fall? Does it occur to you that unemployment is a cyclical way of getting rid of the deadwood in the economy through loss of marginal companies? Its not nice for the good workers who lose their jobs but forces them to retrain in an industry of more use to the economy going forward. Love the lists of so called "hot" jobs that appear occasionally. ITs are still on that list.

  2. Hi Janie,

    The unemployment rate can rise well after the recession ends - and it usually does - forcing workers to re-train (as you suggest) even after the economy starts to expand again. You can see that in the first chart, let's use the 2001 recession as an example, where the unemployment rate rose until 2003, two years after the recession "ended". So, no, the end of the recession is unlikely to correlate with the end of the labor contraction; economists refer to this as the "jobless recovery".

    And as always, thanks for your insight!


  3. Forgot about that jobless recovery - ot looking at the chart properly - THANKS, aj

  4. Just looked at the chart again - I'm a masochist - if you look at the recessions before 1993, umemployment does peak very shortly after the end. Hope we can do that this time but I have a feeling it will act like 1993 and 2001. Oh, well.

  5. Some recessions were "boom" (inflation) "bust" (recession) and others just contracted. A flawed monetary policy was responsible for all the ups & downs.

  6. Hi Flow5,

    Boom and Bust is most certainly one prime contributor of this labor cycle, but what about 2001-2003? The Fed was tightening going that cycle.

    As always, thank you for your comments!