While most analysts who cover military-industrial issues will spend the next several weeks explaining what this or that result for the American presidential election might mean for this or that procurement program, I will offer something different. That is because I am struck by two seemingly unrelated news stories this week, which both address the problems of what people like to call data, or worse, big data:
Aaron Zitner, "What Went Wrong With the Polls This Year?" Wall Street Journal, 4 November 2020. Actually, he means what went wrong again.
Jason Sherman, "DOD moves Google-like tool for searching U.S. military weapon spending behind firewall," Inside Defense, 3 November 2020.
The polls, Donald Trump has been telling everyone for months, have been fake. This was predicted by some to be a blowout, and it was anything but. So maybe not fake those polls, say some pollsters, but certainly fundamentally flawed. In a dynamic dating at least to John Major's seemingly improbable victory in Britain in 1992, we suspect that Trump supporters have tended to lie to telephone pollsters, because they don't care for the torrent of abuse they get from the indignant. Or, maybe they're actually embarrassed. As pollster Robert Cahaly put it in an earlier WSJ article (Barton Swaim, "The Pollster Who Thinks Trump Will Win," 29 October 2020),
Take any retirement community in America. Poll how many people watch ‘The Jerry Springer Show,’ and you’re gonna get one number. Contact the cable company and find out how many people at the same retirement community actually watch ‘The Jerry Springer Show,’ and you’re gonna get another number.
Well, there it is. All this money spent on polling, and most pollsters are totally confused by something that people can easily understand. Meanwhile, the Defense Department is trying to hide its spending data. What's the connection?
Earlier this year I wrote a paper for a conference at the Naval Postgraduate School, later republished as a working paper by our Center for Government Contracting at George Mason University, on "The Value of Intellectual Property in Government Procurement Auctions: The Case of Military Trucks in the United States after the Cold War" (July 2020). One of my findings in the course of the research was that the Defense Department's accounting data are actually worse than I imagined. Consult three DoD sources, and you can get three totally different sets of figures on how much money was spent in a given year on a given systems. Writing that paper was challenging—it's no wonder that this agency can't pass an audit.
Now, I have heard people claim that's somehow not a problem. Why hold government to the standards of business? Accounting information is famously aggregated and misleading anyway, right? Well, if the Defense Department can't understand how much something cost, to any metric recognized by students of accounting, then what hope does that outfit have to leverage big data, machine learning, predictive analytics, and any of the other fashionable buzzwords in making better decisions? Sure, the costs analysts will claim otherwise, because they have their data! But their data are rubbish. So, while Sherman reasonably points out that no one should have confidence in the Defense Department's attempt to obfuscate its records, we may not be missing much.
In short, if you want to work on military-industrial matters, don't rely on government figures. Consult the firms' figures, and then build your own databases.
James Hasik is a senior fellow in the Center for Government Contracting at George Mason University, and a senior fellow in the Scowcroft Center on Strategy and Security at the Atlantic Council.
Comments