Standard rant on standard errors

September 2nd, 2015 at 7:00 am

I keep hearing this strange refrain about this Friday’s jobs numbers and it’s freaking me out a little. Important people pulling important levers seem to have decided that August’s job growth is really decisive in some cosmic way that’s going to resolve any lack of clarity regarding job market slack and the Fed’s plans for rates.

I yield to no one on my obsession with the monthly jobs numbers. Once, when I was in China adopting one of my daughters, I stopped the proceedings to check the release (and I named her U-6, in honor of that important indicator of underemployment–JK! Who would name their kid after a slack measure? Maybe “Wage Growth” or “Full-timer”). But I well know that one month does not a trend make and that given the volatility and revisions history of a high-frequency indicator like this, you should never weight a monthly result too heavily.

Then there’s the standard error. The confidence interval for the monthly change in payrolls is plus or minus 105,000 jobs. Let’s say we get another on-trend report of 220K net jobs added in August. The 90-percent confidence interval on that change ranges from 115K to 325K, i.e., there’s a 90-percent chance that the true August change is within this interval. Think a bit about how these different extremes would change the discussion on Friday.

Don’t get me wrong–there’s important info in the jobs report, which is why I and others blather on about it on jobs day. And if August is another month in what’s been a pretty steady string of solid reports, that will legitimately signal that the job market appears to remain on track. But the idea that it’s a deal maker or breaker for big decisions like the Fed’s liftoff strikes me as downright weird.


Print Friendly, PDF & Email

3 comments in reply to "Standard rant on standard errors"

  1. Gauss says:

    You said “The 90-percent confidence interval on that change ranges from 115K to 325K, i.e., there’s a 90-percent chance that the true August change is within this interval.”

    But this is a misinterpretation of the meaning of a confidence interval. The true August change either is or is not within the CI. The correct interpretation is that for every 100 CIs, 90 contain the true value and 10 do not.

    • Jared Bernstein says:

      The BLS statisticians, who know there stuff, describe it the same way I do, I think: “These figures do not mean that the sample results are off by these magnitudes, but rather that there is about a 90-percent chance that the true over-the-month change lies within this interval.” [my bold:

  2. Unhappy Guy says:

    In my opinion, statistics are totally misunderstood my most people. Someday I’m going to put this into better words, probably.

    I believe the BLS is using proper statistical means in this discussion. The line is sometimes pushed by profit-driven things like banks.

    This is one of the most personal and important posts by Jared ever. You adopted a Chinese girl…

    I’m not going to say anything about it. I always new Jared was a good guy, and I always knew we was not gunning to hurt China in any way.

    How did I know that about Jared? Well, I think I’m a good judge of character.