Ethical standards in different data communities

I opened the paper today and saw this from Paul Krugman, on

Jack Welch, the former chairman of General Electric, who posted an assertion on Twitter that the [recent unemployment data] had been cooked to help President Obama’s re-election campaign. His claim was quickly picked up by right-wing pundits and media personalities.

It was nonsense, of course. Job numbers are prepared by professional civil servants, at an agency that currently has no political appointees. But then maybe Mr. Welch — under whose leadership G.E. reported remarkably smooth earnings growth, with none of the short-term fluctuations you might have expected (fluctuations that reappeared under his successor) — doesn’t know how hard it would be to cook the jobs data.

I was curious so I googled *general electric historical earnings*. It was surprisingly difficult to find the numbers! Most of the links just went back to 2011, or to 2008. Eventually I came across this blog by Barry Ritholtz that showed this graph:

That looks pretty fishy, indeed. Also a link to a news article from 2002 (shortly after Welch stepped down from running GE) that said:

GE used to feature on university courses as a model of probity. These days, it crops up in the seminars about earnings-manipulation. Everyone agrees that GE practises one form of earnings management: it times one-off asset sales to coincide with one-off write-downs or restructurings. . . . Beyond this, the amount of profits-smoothing that GE indulges in is a matter of speculation. GE also manages expectations about its earnings by managing its analysts. . . . managers who are in the habit of smoothing earnings have an especially strong motive to keep the good news coming, whether or not the business warrants it.

Yup. Also this from Marie Leone and Tim Reason:

[In 2009,] after a four-year investigation, GE settled accounting fraud charges with the SEC for allegedly misleading investors with improper hedge accounting and revenue recognition schemes. Specifically, GE was charged with violating accounting rules when it changed its original hedge documentation to avoid recording fluctuations in the fair value of interest rates swaps, which would have dragged down the company’s reported earnings-per-share estimates.

In addition, the SEC charged GE with concocting schemes to accelerate the recognition of revenue from its locomotive and aircraft spare parts business, to make the company’s financial results appear healthier than they actually were.

Without admitting or denying guilt, GE paid a fine of $50 million, and agreed to remedial action related to internal control enhancements. “GE bent the accounting rules beyond the breaking point,” noted Robert Khuzami, director of the SEC’s Division of Enforcement, in a statement.

OK, fine. This isn’t my area of expertise and, in any case, our circulation is on the order of 1/10,000th of Krugman’s, so why report it here?

What’s interesting to me are the different attitudes on statistical manipulation. The Bureau of Labor Statistics, Census Bureau, etc., take their data pretty seriously and I agree with Krugman that it’s hard for me to imagine them manipulating the numbers in any way. For one thing, don’t have much incentive to do so: as Krugman notes, they are civil service workers, not political appointees. And it’s not like better numbers would increase their budget line. (In contrast, I can understand the motivation for those military guys who faked the data on missile tests: success can lead to more funding.) Beyond this, it just doesn’t seem that this sort of fraud is part of the culture of government statistics in the United States.

In contrast, Leone and Reason report:

The SEC complaint relates several instances of round-robin email discussions among GE accountants, internal auditors, executives, and the company’s external auditor, KPMG, debating whether aggressive accounting would past muster with regulators.

So, it’s not about Welch being some sort of data sociopath; rather, data manipulation is part of corporate culture. And, indeed, these guys have lots of motivation to fake the numbers (i.e., “aggressive accounting”). The executives and accountants personally make millions of dollars from it. Millions of dollars in win, very little personal risk if you get caught (it was GE that paid the fine, right? Jack Welch is still at loose on Twitter): that’s what I call an incentive.

One reason this interests me is the connection to ethics in the scientific literature. Jack Welch has experience in data manipulation and so, when he sees a number he doesn’t like, he suspects it’s been manipulated. In academia, Steven Levitt has seen economists “work behind the scenes constantly trying to undermine each other” and been involved with a journal that suppressed unwelcome research results and so, when he has a paper rejected (by a different journal), he suspects the rejection was for illegitimate reasons. I don’t think such a thing would be done in the field of statistics, because I think we have more of a tradition of going with the data.

26 thoughts on “Ethical standards in different data communities

  1. IMHO, the right deals very heavily in freudian projection. They think about what they’d do in a situation, and accuse others of doing this. In this case, Jack Welch knew what those numbers would mean at GE (the CEO picked up the phone and ordered some good numbers), and assumed that the rest of the world worked the same way.

  2. It would be pretty easy to find historical earnings in a proprietary database like Bloomberg or CapitalIQ, which probably relates to why it’s difficult to find them for free.

  3. If Krugman thinks that civil servants never play partisan politics, he’s much more naive than I thought. I was a GS-15 for several years, and I saw firsthand that Federal civil servants play partisan politics.

    • Nothing like this, though, right?

      “several instances of round-robin email discussions among GE accountants, internal auditors, executives, and the company’s external auditor, KPMG, debating whether aggressive accounting would past muster with regulators.”

  4. … I have a very low opinion of Welch, Krugman and BLS bureaucrats, but the fundamental issue remains– is the BLS unemployment data accurate or not ?

    BLS flatly stated the September 2012 U.S. Unemployment Rate as a concise “7.8 %” , with no margin of error specified nor even vaguely referenced in the official pronouncement.

    Would most statisticians readily accept that the BLS methodology could measure unemployment in the vast U.S. economy to an accuracy of one tenth of a percent, each month ?

    Matthew Yglesias ( says: “BLS reports today that the economy added 114,000 new jobs (with, I remind you, a +/- 100,000 confidence interval) in September.”

    • Jklein:

      I think that referring to government employees as “bureaucrats” is about as helpful as referring to corporate employees as “suits.” Like it or not, “bureaucrat” is generally taken as an insult, and I don’t think it’s appropriate to insult someone just because they work for the government. Somebody has to prepare these numbers. What’s next: insulting garbage collectors because the trash smells?

        • It’s rather amusing that it originally had positive connotations. Consider Max Weber and the Preussian model of bureaucracy.

        • I’m sure the good folks at the BLS are gloriously free of the foibles of other government workers. The atmosphere there no doubt resembles a silicon valley start-up more than a bureaucracy.

          Seriously though, terms like “bureaucrats”, “suits”, “ambulance chasers”, “jarheads”, “liars and crooks”(for politicians), and similar descriptions are a fine set of prejudices for people to have. Their use should be encouraged at least until either human nature changes or we cease to be a free people.

          (P.S. “garbage collector” is insulting. They like to be called “sanitation engineers”)

        • Also, I have no problem mocking people if they deserve to be mocked (for example, if they repeatedly plagiarize and never apologize). But I don’t see the point of mocking someone just because they work for the Bureau of Labor Statistics. As a statistician, I happen to think that working for the Bureau of Labor Statistics is an honorable calling.

        • honorable != free from the slovenly tendencies of bureaucracies

          I actually interview and considered working for BLS before joining the Marines. I also think there is precisely zero possibly they cooked the numbers for political gain.

          Nevertheless, they are a bureaucracy and they are bureaucratic. They should be mocked mercilessly for it as the price they pay for job security and avoiding the rigors of a real competitive environment.

        • Seriously, Entosphy? BLS employees should be “mocked mercilessly […] as the price they pay for job security and avoiding the rigors of a real competitive environment”? This is your argument as a Marine, a federal employee with no private-sector competition? I can think of some names with negative connotations that get thrown at Marines, too, and I hope we can both agree they would be unacceptable.

        • All the names thrown at Marines are perfectly acceptable. Most like “jarhead” and “devil dog” are badges of honor. It might surprise you to learn that Marines don’t partake of the cult of hypersensitivity and don’t actually cry much when people call then names. In fact, about the only real insult you can throw at a Marine is to call them “Soldiers”.

          Having said all that, I’m not currently a federal employee. I joined to deploy and after that I left. My entire time was spent training or deploying. I wasn’t interested in joining any peacetime bureaucracy.

          So yes, seriously, go ahead and mock the time-servers at the BLS and every other bureau. Mock them mercilessly!

    • They actually provide a rather detailed discussion of sampling uncertainty as part of their release. It is on page 6 out of 38 on the pdf version of the release so it is relatively close to the front of the document.

      I do not work at BLS but the folks I know who do have solid backgrounds. I am not sure what has led you to have such a negative opinion of them.

  5. An interesting post re Welsh. That graph looks very interesting.

    On-the-other hand it is a bit diffucult to believe that the US Bureau of Labor Statistics is that corrupt. I don’t hear any outcry from Statistics Canada or the European stats organizations questioning the data.

    Of course, most of the world sees the US Republians as batshit crazy,anyway and totally conspiricy oriented in any case. An interesting example is John Quiggin’s blog post Republican conspiracy theory update

    As David Frumm,a noted conservative, speaking of US conservatives and presumably most Republicans puts it
    Backed by their own wing of the book-publishing industry and supported by think tanks that increasingly function as public-relations agencies, conservatives have built a whole alternative knowledge system, with its own facts, its own history, its own laws of economics.

  6. My former CFO at SpeechWorks “settled” his fraud cases for a measly $100K and lifetime ban from being a corporate officer, with the CEO and others also being fined, but even less:

    The big speech reco company before SpeechWorks and Nuance was Lernout and Hauspie. They bought up all kinds of other companies, including Dragon, before they were thrown in jail for yet more accounting fraud:

    As a result, I have serious trust issues with “C”-level executives.

    Hmm. Maybe Andrew wants to do another “how many X do you know” study.

  7. “the household survey — the survey that the unemployment rate comes from — showed that the number of people with jobs rose 873,000 in September, though the gain had averaged 164,000 each month earlier this year.
    These numbers are always tremendously volatile, but the reasons are statistical, not political. The numbers come from a tiny survey with a margin of error of 400,000”.

  8. Pingback: “The treacherous are ever distrustful…” (Gandalf to Saruman at Orthanc) « Another Word For It

  9. Before responding to anything with political aromas, I state my non-partisanship. This is a nonpartisan opinion and assessment not eluding to anything other than concern over unemployment numbers dating back much further than this quarter.

    When the employment numbers declaring a .3% drop were presented this week, my skepticism drove a private email conversation with my friends. This was based on job creation over the past month not being enough to maintain 8.1% let alone reduce unemployment.

    A few points:

    The people who compile unemployment data are just people.

    The U3, which is the published unemployment count, does not tell the complete unemployment picture. Rather than take my word it, read this explanation of U6, the six levels of unemployment measurement, and then look at the U6 stats on the same page.

    This Huffington Post article claims both sides are right and discusses the U6:

    “The more interesting aspect of the latest LBS data is this: Even if the 114,000 new jobs figure is correct, it is below the level required to match new entries into the labor force. In other words, the U3 (and U6) rate should have risen instead of declined.” 

  10. It will be very interesting to watch how the numbers get revised after the election. If the economy really is doing so much better, then I suppose the revisions will likely be upward. If Welch is correct, then I suppose that after the election is out of the way, the BLS will try to correct the number.

  11. Interesting post! I fear I am not as confident as you that

    I don’t think such a thing would be done in the field of statistics, because I think we have more of a tradition of going with the data.

    You and I (academic statisticians) do indeed have nothing to gain and everything to lose from faking stuff, AND we know how hard it would really be. But a lot of qualified professional statisticians are working as one operator on the conveyor belt of data, with speed being valued as highly as accuracy, and not enough information to understand the whole picture. When pressure is applied to massage the results in a certain direction, they may not have the time and energy to ask why. And they do have something to gain from pleasing the boss. They’re not wicked people, they just don’t have the flexibility to stop the conveyor belt and ask what’s going on.

  12. Pingback: Links 10/13/12 | Mike the Mad Biologist

  13. Pingback: Politics as an escape hatch « Statistical Modeling, Causal Inference, and Social Science

Comments are closed.