Thursday, April 19, 2012

Don't trust the government

Is the Labor Department playing politics with job numbers?

Whenever a situation appears to be a conspiracy, the explanation is likely to be happenstance. But, looking at how often (and how much) BLS has revised its unemployment numbers, we have a strange trend that is, at the very least, worth noting.


More Americans "than forecasted" filed applications for unemployment benefits this week -- first we heard that the jobless claims fell by 2,000 but now the revised numbers show 6,000 above the initial forecast. As Ed Morrissey notes, "That number got revised this week, but the real story is in the 4-week rolling average. Just three or four weeks ago, that number was in the 360K range. Now it's close to 375K, roughly the same level as last spring’s stagnant economic conditions."


Without even taking into account the incredible shrinking job force, this portends bad economic news. Yet, the bad news always seems to be tempered by the Labor Department.


According to Bloomberg:


Revisions to previous data have been larger than normal and the government is trying to determine the cause, a Labor Department spokesman said as the figures were released to the press.


A lot more than normal, actually. According to Dow Jones, the Labor Department had revised its estimate of seasonally adjusted jobless claims upward in 56 of the past 57 weeks. That's unprecedented.


Fact is, initial estimates draw the most reaction from media, while the revisions are treated less newsworthy, despite their relative significance. How beneficial it is to muddy this information is debatable, I suppose, but the trend certainly warrants some explanation. The media and the public pay especially close attention to unemployment rates. Having more favorable numbers blasted out by newspapers and television before hard reality hits can't hurt the president.


It's what Zero Hedge calls "Initial Claims Propaganda 101":


Instead of printing at the expected 370K, an improvement from last week's already big miss of 380K, this week came at a whopping 386K, the worst standalone print in 4 months. Well, until last week's revision that is: instead of the 380K print that stunned everyone, last week's number has now been revised to a massive 388K. Why? So that mainstream media can declare, with a straight face, that this week saw the number of initial claims decline!


I emailed Dr. Anthony Sanders, a Professor of Finance at George Mason University, wondering why he thought this was happening -- is it a problem with methadology or just wishful thinking? He writes: "The stock answer is that the methodology is flawed (surveys) and better data comes in after the fact. But 56 out of 57 weeks indicates that the BLS is playing games with the data. Fool me once shame on you, fool me twice shame on me, fool me 57 times shame of the BLS... The problem with BLS numbers are the same ones as the inflation numbers. The collecting and processing of the data is in the hands of those who gain or lose from the numbers. Massive moral hazard risk."


Gathering weekly, or even monthly, data of this scale is no doubt exceedingly complex so Americans can forgive revisions. (Though, wouldn’t it be nice to outsource this work to private firms that have reputations to protect, rather than some bureaucratic entity? Naturally, a non-starter.)


The problem is that even other government reports that matter to the market do not suffer from similar consistency problems. Sure there are revisions, but they move up and down with nearly the same regularity. Why is it that unemployment numbers always seem to shift in the same direction?


Well, I guess a report on the situation is forthcoming... you know, at some point.


No comments: