>On Fri, 20 Sep 2002, Christian Gregory wrote:
>
>> > Question: "Seasonally adjusted" means ... ?
>>
>> Taking into account the fluctuations of both the labor force and employment
>> rate due to the time of year.
>>
>> Christian
>
>I'd gathered it meant something along these lines. But I guess my question
>is really how these things are determined, what they consist of, etc. I
>understand it can be a fairly involved topic. Any explanatory websites
>that break it down (outside of the monstrously unhelpful DoL pages, that
>is) that you know of?
I find the BLS web pages very helpful, but maybe I'm so immersed in the stuff I can't tell when it's incomprehensible to nonspecialists.
Most economic series can be thought of as being determined by four sets of forces: secular trend (long-term, usually up), cyclical (boom or bust), seasonality (relatively predictable fluctuations because of the calendar), and noise (random fluctuations that have little meaning over time). Seasonal adjustment attempts to take out the calendar effects so that cyclical movements can be made more clearly visible.
A common example of a seasonal fluctuation in employment is what happens around Christmas. Retailers hire people in November and December and lay them off in January. Over time, the pattern is recognizable and fairly predictable. So, seasonal adjustment compensates for the effect of this, so that you can compare the changes in employment between October and Feburary with more confidence. Say retailers hire 100,000 in a normal November. If they hire 100,000, after adjustment the employment change will be 0 - it's just what's seasonally expected. If they hire 150,000, though, that would mean a gain of 50,000 - a strong Xmas season. If they hire 75,000, that would become a loss of 25,000 - a weak Xmas. Other seasonal features of the employment calendar include hiring of new graduates in June, or the return of teachers to work in September, or layoffs in the auto industry between changes for yearly new models.
That's the concept. The way it's done statistically is more complex, but it's basically a high-tech version of a moving average taken over several years - that is, months (or quarters) are examined for consistent departures from average movements in surrounding months (the burst in retail employment in November, for example).
Conspiracy types (e.g. NY Post columnist John Crudele) often treat seasonal adjustment as evidence the government is cooking the books. That's stupid. There are some quirks in seasonal adjustment, but it's a reasonable technique. And over the year, the seasonal adjustment factors cancel each other out, as high months are offset by low months.
Doug