Fun with Numbers

October 22nd, 2012 at 12:31 am

Zach Goldfarb has a thoughtful column on how economic numbers are used and abused.  Just a few points I’d add.  One could easily come away from such a column concluding that you just can’t trust the numbers that partisans use to make their cases.  And while one would often be right, one would just as often be wrong.

This is an argument worthy of more ink (bytes) than I have time for now, but a few rules of thumb that I guarantee will serve you well:

–No question, as Goldfarb stresses, that the most common game the politicians and the economists play is cherry-picking the start date from which they begin counting jobs, growth, or whatever.  Do so, and you can almost always get any result you want.  But what good economists and journalists need to do in such cases is either point out the rationale for that start date or explain why it’s bogus.

–EG, Presidential cycles—counting from when a president took office—just don’t make sense.  It’s a date related neither to the business cycle nor the president’s policies, none of which take effect until well after they take office.

–On the other hand, and here I disagree with the piece, there’s nothing wrong with counting job growth, as team Obama is wont to do, from the employment trough, as long as you’re explicit about it.  “Since we’ve been adding jobs, we’ve added X million…” is legit.

Goldfarb says you should start counting from the GDP trough, not the jobs trough.  But I don’t see why that makes more sense, again, as long as you’re explicit about when you’re starting your count.  In fact, I think it makes more sense in an apples-to-apples sense to say “since we’ve started adding jobs, we’ve added X million,” than to say “since GDP started growing, we’ve added X million jobs.”  This is particularly the case in an era of “jobless recoveries” where the GDP trough has preceded the employment trough by many months.

Of course, linking that to your policies, or providing relevant context (how does X million compare with past employment expansions?) raise other relevant questions, but if your goal is to get politicians to disavow their role in good things that occurred during their watch…well, good luck with that.*

–The trend is what matters.  Obviously, one month’s data from one report doesn’t mean squat, especially if it’s inconsistent with the underlying trend or other data sources.  But once you’ve got enough data telling the same story, that’s likely the right story.  The fact that many different housing market indicators—price indexes from different surveys, construction numbers, mortgage activity, builder confidence—are all pointing in the same direction and have been for a few months now provides ample evidence that the housing market is finally starting to climb back.

–This trend result is particularly important with the monthly jobs numbers.  Goldfarb underscores a point I often raise here myself: we give much too much policy and political weight to these monthly reports, especially given their sizable margins or error (100K on the payroll employment delta; 400K (!) on the household employment change) and revisions.  I vividly remember as a White House economist how after a strong jobs report, the media, if not the nation, pretty much judged your policies to be working, only to decide next month, after a lousy report with negative revisions, your policies stunk.

So, my idea, which has never gotten the slightest bit of traction, is to switch from monthly jobs reports to quarterly ones, like we do with GDP.  That would mean larger samples, less statistical noise, a more reliable read on the job market, less policy, political, and market lurches based on monthly data, and smaller revisions.  It might even mean we’d hear less from Jack Welch, which for many of us, would be more than worth the price of admission.

*My favorite example of this phenomenon is in Ohio right now, which was doing worse the national economy in unemployment terms and is now doing considerably better.  Obama says “it’s my stuff that did it!” and with the auto rescue, he’s got a point.  But Republican Gov Kasich says “it’s not him…it’s me!”  Meanwhile, Ohio Senator Rob Portman, a Romney surrogate, is going around saying “hey Ohio, things really aren’t that great at all!”

Print Friendly, PDF & Email

One comment in reply to "Fun with Numbers"

  1. Tom in MN says:

    This is why I like plots and not single numbers. I want to see the trends and context. You asked previously about blogs and evidence. Well the blogs I like (yours, Krugman, Duy, Delong, Thoma) all present the data they are using as plots (from reputable sources) and the reader gets to analyze the data too.

    But I think the problem is that you and I can look at a plot and immediately seen trends, think about rates, etc. The average voter, unfortunately (it’s all math skills!), probably has not dealt with plots since school. Thus people tend to give them processed numbers and the average person may not even know enough to realize all the issues with single numbers. My largest complaint with the “News” is when they spit out single numbers with no context: “Oh, 10 billion dollars — that’s huge.” Well compared to what?