by Calculated Risk on 12/04/2009 12:02:00 PM

In August, when it was reported that the July unemployment rate dipped slightly to 9.4% from 9.5% in June, I pointed out that the dip in unemployment was just monthly noise: Jobs and the Unemployment Rate

FAQ: How can the unemployment rate fall if the economy is losing net jobs, especially since the population is growing?

This data comes from two separate surveys. The unemployment Rate comes from the Current Population Survey (CPS: commonly called the household survey), a monthly survey of about 60,000 households.

The jobs number comes from Current Employment Statistics (CES: payroll survey), a sample of approximately 400,000 business establishments nationwide.

These are very different surveys: the CPS gives the total number of employed (and unemployed including the alternative measures), and the CES gives the total number of positions (excluding some categories like the self-employed, and a person working two jobs counts as two positions).
...
[T]he jobs and unemployment rate come from two different surveys and are different measurements (one for positions, the other for people). Some months the numbers may not seem to make sense (lost jobs and falling unemployment rate), but over time the numbers will work out.

Here are a couple of scatter graphs to illustrate this point ...

The first graph shows the monthly change in net jobs (on the x-axis) as a percentage of the payroll employment, and the change in the unemployment rate on the y-axis.

The data is for the last 40 years: 1969 through July 2009.

Click on graph for large image.

Although these surveys are different measures of employment - there is still a correlation - in general, the more payroll jobs added (further right on the x-axis), the more the unemployment rate declines (y-axis). And generally the more jobs lost, the more the unemployment rate increases.

But the graph sure is noisy on a monthly basis.

Look at the two red triangles - those are the data points for the last two months.

Notice that the increase in the October unemployment rate was much higher than expected based on the number of payroll jobs lost. And the opposite was true for November (the unemployment rate fell even though payroll employment declined slightly).

The second graph covers the same period but uses a two month rolling average:

Now we see a much sharper correlation.

The red triangles are the for the last two data points, and the Sept-Oct point is above the curve, whereas the Oct-Nov point is on the curve. All this means is the jump in the unemployment rate in October was higher than expected, and the decline in November balanced it out.

This also suggests the economy needs to be adding about 0.13 percent of payroll employment per month to keep the unemployment rate from rising. That is about 170 thousand net jobs per month - this accounts for both population growth and an expected increase in the employment-population ratio.

Note that the trend line is a 3rd order polynomial (equation on graph). When the economy starts to add jobs, more people start looking for work - and the relationship between net jobs and the unemployment rate is not linear. (see next graph).

If we use a six month rolling average for the above graphs, R-squared rises to 0.8.

This graph show the employment-population ratio; this is the ratio of employed Americans to the adult population.

Note: the graph doesn't start at zero to better show the change.

This measure was flat in November at 58.5%, the lowest level since the early '80s. However once the economy starts adding jobs, more people will be looking for work, and the employment-population ratio will start to increase. This means the stronger the economy, the more net jobs required each quarter to lower the unemployment rate by the same amount.

The bottom line is the decline in the unemployment rate this month was noise, and the unemployment rate will probably increase further. If the economy adds about 2 million payroll jobs next year, we'd expect the unemployment rate to still be at about 10% at the end of the year.