The BLS talks about the birth/death model - which models jobs added or lost (in theory) due to new businesses starting up and old businesses failing or closing - and how tends to lag behind turning points in the economy - it'll keep adding jobs for a while as jobs are actually lost, then more accurately reflect reality as job losses accelerate in the depths of the recession, then lose jobs when jobs are starting to be added in real life, then add jobs as the recovery picks up steam and real job additions are increasing. But beyond that, they refuse to disclose how it works, in any way. It's a black box. At this point, of course, everyone who follows it just assumes it's a number-massaging tool to make things look better than they are. I certainly do.
But it occurred to me: what if it's not? Let's pretend, for a moment, that the BD model is honest, and just broken, and is lagging economic reality, as is stated. Just as a thought experiment.
That would imply that we're still in stage one - adding jobs as jobs are lost early in the recession. Before job losses get really bad and the recession really starts to grind home. In the last recession, you had a substantial streak of at least well-mixed negative/positive birth/dead adds and losses - say, from 2002, from July 2002 to February 2003, you had adds and losses that came to a total of -151,000. More importantly - because of January - you had birth/death model job losses four months out of eight, or, backing up out of 2003, three months out of six.
That's what the "bottoming out" period looks like in this model, when the worst of the recession is upon us. We don't have that yet. What does that, in turn, imply?
Taking data again from the previous recession - which is unfortunately the only one using this model - it implies substantial continuing job losses into 2010 and quite possibly 2011, and I suspect that's where the "green shoots"/"recovery" combined with "continuing rise in unemployment" mixed message comes from.2 It implies a target U-3 (headline number) unemployment rate of somewhere between 11% and 13.5%, and a U-6 (under- and unemployment) in the 23%ish-range - something comparable to the Great Depression. (And if you didn't read footnote 1 when it was referenced above, please go do so.)
Of course, that assumes it's not just a numbers-manipulation tool. It's probably that. But if it's not - if we take them at their word - then you have this message from their model: unemployment is going to get substantially worse before anything gets better.
1: From BDA historical model: Jan08: -378,000. Jan07: -175,000. Jan06: -193,000. Jan05: -280,000. Jan04: -321,000. Jan03: -391,000. In other words, the January 09 subtraction of -356,000 number is less than the numbers shown at the worst of the previous recession, in the tech crash, and not hugely out of line in comparison to non-recessionary periods such as January 04. It has functionally never turned negative in this recession.
2: And if the "green shoots" terminology doesn't remind you of Hoover administration propaganda from 1931, it should. It really, really, should.