>Hope this hasn't already been discussed on list, but
>I heard that starting in January there was a change in the
>way the unemployment rate is calculated. Does anyone know
>what the new method us, and what the effect is likely to be
>on the putative rate? Without knowing anything about it,
>and given that the Bush administration already stopped issuing
>"mass layoff" stats, my guess is that the new method will
>make the economic outlook seem rosier. FWIW, the official
>January un rate was 5.7%, down .3% from December.
A couple of things happened (which I wrote about here the other day). First, the BLS introduced new population weights, derived from the 2000 Census; previously they'd been using 1990 Census data. These are used to translate the results from the monthly sample of some 60,000 households into national data. The effects of this were fairly minor. Second, they introduced new seasonal adjustment factors and a new technique as well. The new factors are applied every year, based on recent experience; when they're introduced, they revise the older figures going back three years. Seasonal adjustment compensates for regularly occuring things like pre-Christmas hiring and post-Christmas layoffs in the retail industry, summer break for teachers, and the like; the intent is to separate those recurring developments from the underlying economic trend. (If, as recently happened, Christmas retail hiring is weaker than usual, and post-Christmas layoffs are weaker as weel, this will be seasonally adjusted into a decline in retail employment in December and a rise in January.) And, as if that all weren't enough, there were technical complications with the January data (that I don't understand, not that I've tried) that affect only Jan's numbers; things will be back to normal with Feb's release.
That unemployment decline was mostly the result of seasonal adjustment quirks; it'll be very interesting to see what happens with the Feb numbers.
There should be no longer-term effect on the unemployment rate from any of these changes - they're purely routine and technical. Over the course of a year, all the seasonal factors average out to zero; they're meant to facilitate short-term comparisons, like over one to six months. But if you average the adjusted and unadjusted figures over the course of a year, the two annual averages will be equal.
Doug