The United States is the 3rd largest country in the world by population. With a population of about 300 million people it is only slightly smaller than the populations of Germany, France, the United Kingdom, Italy and Spain combined, which according to the Wikipedia article above, is about 314 million. Because of this, whenever news reports on the unemployment rate it seems as odd as reporting on the unemployment rate for North America or the European Union, when a somewhat more useful number would be the unemployment rate of Mexico or Germany, or your state. In this series of articles we’ll examine the national unemployment rate in a little more detail.
Perhaps the most important unemployment rate number is your unemployment rate, whether you’re unemployed. But regardless of your employment status, the unemployment rate should give you some sense of how hard it will be to find a job (if you’re unemployed) or how secure your job is (if you’re employed). All by itself any unemployment number doesn’t provide much info, but one can get a sense of these answers by examining the unemployment rate over time. The chart below goes back to just before the last major recession in the early 1980s. It also shows 2 different aspects of the unemployment rate, seasonally adjusted rate and the unadjusted rate. In the interest of space I’ve combined these 2 graphs, but for now either rate will do.
The first thing that stands out is that the unemployment rate from the 1982-83 recession was much higher than the unemployment rate is now. Of course both the 1982 rate and the rate today are much higher than the intermediate recessions, but at least our current recession is not (yet?) as bad as the 1982 recession.
There are 2 other oddities that the careful viewer may notice. The first is how much the red “unadjusted” line moves around the blue “seasonally adjusted” line. The second is harder to see as it’s at the extreme right hand edge of the figure. The figure below shows this area in detail. In particular it shows that the “seasonally adjusted” rate is significantly higher than the unadjusted rate. Since we’re all looking for good news here one has to wonder if the seasonal adjustment is correct.
The Bureau of Labor Statistics describes both the motivation for using seasonally adjusted data as well as an overview of the method by which they calculate it. We won’t delve into the mathematics here, but rather just show it in action. To start, examine the first figure above again and notice that, in a fairly regular pattern, the unadjusted rate is higher than the adjusted rate at the beginning and middle of the year. This pattern occurs year after year. The BLS website states that the seasonal adjustment process smooths out predictable variations due to weather, holidays and other factors.
What sort of jobs would be most affected by these patterns? Farming is perhaps the most obvious, but I’m going to ignore farming because if farming were the ONLY seasonal job sector then the adjustment process would be pretty straight forward. Construction is also seasonal. One might expect more people to be doing construction during the warm/dry months and fewer during the cold/wet months. Retail employment might also be seasonal as we’re used to hearing of increased hiring for the Christmas shopping season. What sort of jobs would be the least affected by seasonal variations? Healthcare is one that comes to mind.
The BLS website provides both employment and unemployment data on these and other job sectors. Below I’ve plotted the EMPLOYMENT data for these 3 sectors over the past 5 years. Note that this is employment data, so the higher the line the more people are employed. Also, these are “absolute” numbers meaning the numbers at the right of the figure are the actual number of people with jobs, not a percentage. Thus in mid 2006 the actual number of people employed in the construction industry was about 8 million. Now, looking at the graph as a whole, we see that our intuition about these sectors was indeed correct. Construction employment is high during the late spring to late fall months, while retail employment peaks in the late fall, and there is practically no seasonal variation in the health care sector.
But as far as the validity of the adjustment process goes, the question you need to ask yourself is does the smoothing of the red line(s) into the blue line(s) look right? If all you wanted to do was to compare May of one year to May of another year then the unadjusted number would work as any seasonal variation during the year would be negated because the seasonality in May of one year would also be the seasonality in May of another year. But if you wanted to compare May to June or try to get a sense more frequently than once a year as to whether things are getting better or worse then you’ll need to smooth out the seasonal variations. To drive this point home look at the construction graph for Jan 2009 to May 2009. Even though we see the red unadjusted number growing in May the blue adjusted number is still falling. An “eyeball” inspection of the growth of the unadjusted number in May seems to be due entirely to the normal Spring increase but at that even the seasonal increase seems to be less than the normal seasonal increase, hence the adjusted number falls; even in an “up” month we’re increasing less than normal.
Conclusion (to part 1)
So now you should have a better sense of how the seasonal adjustment is both used and how it’s determined. Next time we’ll examine the unemployment rates in various states in the US as well as the unemployment rate across various demographic sectors. For that we’ll be using seasonally adjusted numbers hence the detour into it here.
Regarding the info presented so far there are couple of final points. First it’s noteworthy that healthcare employment continues to grow even during the current economic downturn. Next, while both the retail and construction sectors showed decreases, the drop in jobs in the construction sector is much more severe. Just eyeballing the data it looks like more people lost jobs in construction than in retail, and since the total number of people employed in construction is smaller than in retail, from a relative perspective the percentage of job lost in the construction sector is MUCH worse than in retail. (For example, 1000 jobs lost in an area that normally employs 10,000 people is less severe than 1000 jobs being lost in an area that normally employs 5000 people). This is consistent with what we’ve been hearing in the news about the special weakness in the housing sector.