Gene Epstein, Econospinning: How to Read Between the Lines When the Media Manipulate the Numbers. Hoboken, NJ: Wiley, 2006
“Early in life I had noticed that no event is ever correctly reported in a newspaper.” —George Orwell
Economists are familiar with the cliché “lies, damned lies and statistics,” which puts statistics at the top of the pyramid of lies. ESPN sports radio personality Colin Cowherd, on the other hand, insists, “People lie, the numbers don’t.”
Since people create the numbers the line between liars and bad numbers may be less than bright and clear, but Gene Epstein — economics columnist for Barron’s magazine and author of Econospinning — essentially sides with Cowherd. Epstein finds little fault with government’s economic numbers and plenty of fault with the reporters and pundits who use those numbers.
Perhaps surprisingly for a pro-market, anti-statist economist, I agree with Epstein’s assessment. Most of the numbers, especially the labor data produced by the Bureau of Labor Statistics upon which Epstein concentrates, are tolerably reliable and accurate, provided the user takes the time to learn their limitations. In 2001–2 I was chief economist at the labor department, a Bush-Cheney political appointee of all things, and was quickly impressed with the thoroughness and professionalism of BLS personnel and their numerical output.
Epstein defines “econospinning” as the “sort of economic journalism that shapes data around a predetermined story rather than the story around the discoverable data.” It is all familiar enough, now isn’t it? Rather than asking “What happened?,” too many observers push their own agenda, fixing the intelligence around the policy as the Downing Street memo on Bush’s Iraq invasion put it.
Epstein concedes that no one is completely immune from bias, but given all the scandalous media behavior he documents, he goes rather easy on the sinners. As for media motivation, he is content to merely list the “usual reasons for bad journalism” like a preference for “sizzle” over substance, sloth, ignorance, and the fact that economic reporting “tends to be more complicated than most other forms of journalism.” Epstein is too much of a gentleman to point out the general pattern he uncovers, namely, a preference for governmental intervention in economic affairs seems to motivate many pundits and reporters and thereby silently shapes their use of the numbers.
While Epstein ignores the truly big picture, he is great at the applied level he chooses to write at. Virtually every observation he makes is spot-on. Markets have an “insatiable and irrational need for timely data,” consistently failing to recognize the weak signal-to-noise ratio in weekly and monthly data releases despite the fact that the numbers are often substantially revised later. I saw this almost daily while I served as chief labor economist but especially on the first Friday of most months when the monthly employment report was released. Epstein shows that “even the three-month and six-month trends are so rife with false signals that they obscure more than they clarify.” Epstein asks whether the monthly employment report moves the bond market in the right way and answers, “Probably not.” That ought to tell us something.
Most chapters are based on a clear explanation about important data and these numbers are contrasted with the distorted media coverage. Chapter one, for example, discusses the nonpartisan Congressional Budget Office’s (CBO) projections of Social Security, Medicare, and Medicaid spending, forecast to rise from a combined 8 percent to 19 percent of GDP in 2050, so-called “eldercare.” These are reasonable projections, given the baby boomers and increasing real value of benefits.
The US taxpayer, not the US Treasury, is the ultimate resource, of course, to pay for these benefits, but propagandists like Paul Krugman, the famous New York Times columnist and John Bates Clark award winner in economics, heap confusion on the issue by pointing at the Bush tax cuts as the financial culprit, along with other distractions. Epstein, by contrast, is superb in clarifying the exact nature of the looming financial crisis in this collectivist swamp. Krugman, meanwhile, rattles on about the glories of a “single-payer” for medical care (Who? the Easter bunny? Saudi Arabia?), something he labels “Medicare for all.”
Chapter two gets into the heart of the book: labor data. There are two ways to survey employment: ask people or ask employers. The BLS does both, of course, in the household survey each month, a sample of 60,000, and the so-called establishment survey of 400,000 job sites. Yet writers mistakenly mix the surveys up too often to nonsensical effect.
Because of little or no job growth during much of Baby Bush’s first term, pro-administration pundits claimed the establishment survey was no good because it was missing growth in self-employment and undercounting start-up businesses. Yet the BLS all along counted limited liability corporations (LLC) workers as wage-and-salary workers and the BLS monthly adjustment for start-ups showed it actually had been overcorrecting for missed start-ups. The BLS finds this out annually because it has access to the “universe” count of employment via the mandatory unemployment insurance tax system and applies that data to “benchmark” or revise its monthly data retrospectively. The propaganda attack on the establishment data was wrong again. Epstein has a gift for explaining the complex in clear and simple terms and the annual “benchmarking” is a shining example.
Another attack on BLS data came from left-wingers like Paul Krugman who insisted that the unemployment data understate the real scope of unemployment. Yet Epstein shows convincingly that the unemployment rate is “roughly accurate,” all we can expect of such a series. There was no “phantom dropout” from the labor market. Epstein challenges Krugman rhetorically, “Why not look up the data?” The BLS has many variations on the basic unemployment rate to capture the “discouraged” and other people weakly attached or unattached to the job market. These series all move together. Epstein skillfully disposes of related claims about changes in the labor force participation rates, employment/population ratios, etc., supposedly exposing hidden labor market troubles.
While Epstein is a keen student and defender of BLS data, he has sensible suggestions for reform. Average hourly earnings have long outlived their usefulness and the BLS plans to discontinue the series by “early 2010.” Their weakness hinges on the fact that most establishments have no clear meaning for “nonsupervisory” and “production workers,” and even if they did, do not keep separate records on hours and pay for this subgroup. As a result, most do not answer these questions so the BLS resorts to “special methods.”
Average hourly earnings adjusted for price inflation are important data because they have shown little or no growth over time and have little or no correlation with aggregate productivity growth, apparently contrary to economic theory. But these are the wrong labor earnings data, especially since better-paying firms have been underrepresented and the underrepresentation has worsened over time. The “crabbed” data on hourly earnings should be replaced with data on all employees and include all wage/salary/benefit disbursements. Once that is done, Epstein shows that average pay has increased and remains highly correlated with GDP productivity change.
Epstein corrects another GDP-related issue, namely, “record” corporate profits as a share of GDP (BEA’s National Income and Product Accounts are spongier data than BLS labor data). This “profit orgy” is a false alarm because the corporate share of GDP has risen and the traditionally highly profitable financial sector has become a bigger share of the corporate sector. Hence, corporate profit as a share of GDP is higher while no higher as a share of corporate GDP. Corporate profit as a share of corporate GDP, for example, in the “Republican year” of 2004 was less than the “Democratic year” of 1997.
A most worthy BLS reform suggested by Epstein is to “end the monthly madness,” whereby statistically meaningless changes in the monthly payroll employment numbers set pundits and markets a-twitter. Epstein reports that the media cite the employment report more than all other economic reports combined. In an economy of 145 million employed, any employment change less than 400,000 (a fraction of one percent) is statistically meaningless.
What is important is the trend, so the BLS should report year-over-year percentage changes in employment, according to Epstein. That puts one year of data between the new numbers and the old, drastically reducing volatility and related problems. It would tame most of the problems stemming from sampling error, benchmarking, and the enormous seasonal adjustments the BLS uses. The last would be obsolete with 12-month trends performing the work instead.
Epstein recommends only a 6-month trend for the unemployment rate because it is less volatile and already is a ratio rather than an absolute number. Among other blessings, this change would end fussing over 0.1 or 0.2 changes in the monthly unemployment rate, say, from 4.7 to 4.9, which are statistically meaningless. The media attach an unwarranted “concreteness” to sample estimates out of proportion to their real status, probably out of sheer ignorance.
The last four chapters are great fun, exposing luminaries Alan Greenspan, economist Steven Levitt of best-selling Freakonomics fame, sociologist Barbara Ehrenreich of best-selling Nickel and Dimed fame, and CNN’s Lou Dobbs for the error-strewn economists they are. I used to call Greenspan “Mr. Magoo” around the labor department office while the press idolized the fool. Epstein shows how mistaken the media were, proving Greenspan had all the perspicacity of Magoo as an observer, forecaster, and decision maker.
But the most delicious criticisms Epstein levies are those directed at Levitt, and Epstein minces few words here, declaring Levitt’s celebrated research as “truly of the emperor-has-no-clothes” variety. Levitt (and co-researcher John Donahue) claimed to find that legalized abortion was substantially responsible for the falling crime rate. But Donahue and Levitt failed to “show an actual decrease in births of crime-prone offspring, only that there might have been.”
Epstein looks at the actual data and finds crime-prone population increased as a share of overall population, and so abortion accounted for zero percent of the 1990s drop in crime. That would suggest the vast increase in law enforcement spending and imprisonment as the likely explanation, a theme I harped on for a decade.
Epstein shows how a “more pedestrian arithmetical exercise than any they engaged in … has the virtue of addressing the issue head on.” Epstein gently says Levitt and Donahue were “snookered by their own heavy-duty statistical techniques.” Epstein is nearly as good in eviscerating Levitt’s fancy work on asymmetric information and real estate agents.
There is lots more to Epstein’s book and it is all of the highest quality. I wish I had this book when I stepped into the job as chief economist at the labor department. Toward the end, Epstein says the whole issue of income distribution and inequality deserves a book unto itself. Let us hope he writes it.
