Statistical Gnosticism on the Right

Folks, it’s time we had a serious conversation about statistical gnosticism. I talk a lot here at American Compass about the right-of-center’s problem with market fundamentalism—the tendency to say things like, “markets do not fail us, we fail markets” or “as we are dealing with changes in our economy, tax cuts are always a good idea” or “profits are the mother’s milk of prosperity.” But a related problem concerns the absolute conviction on the part of some right-of-center analysts that basic statistics describing some of America’s challenges are obviously wrong, everyone knows this, and with a sufficient set of biased and values-based corrections one can show that, to paraphrase the coffee-sipping dog in the flame-engulfed kitchen, everything is fine.

Statistical gnosticism holds, for instance, that the federal government’s Bureau of Labor Statistics adjusts incorrectly for inflation (using a measure called CPI) when reporting slow to non-existent wage gains for production and non-supervisory workers over the past 50 years, and we should instead use an alternative inflation measure (called PCE) that says inflation has been lower and thus, voila, wages have risen higher. The claimed reason for this change is that PCE better accounts for “substitution,” when consumers switch from more- to less-expensive goods as prices rise. That’s not quite right. As I showed in a 2020 report for the Manhattan Institute (see figure 2), PCE’s accounting for substitution does give a lower estimate for inflation and thus make wage gains look better, but a number of offsetting methodological changes cancel that effect almost entirely over the past couple of decades. The actual reason PCE makes inflation look much lower is that it badly underweights the rise in housing costs. The argument is essentially, “the typical earner is doing well if you ignore how much housing costs have risen,” which is not a very good argument.

To be sure, measuring how social and economic changes interact with price levels to affect the cost of living and thus the budgetary concerns and well-being of the typical family is a complicated endeavor. PCE is useful. But, as I argue in my Manhattan Institute paper, the traditional measures of inflation used to detect changes in economy-wide price levels badly understate the financial pressure on households. Others have argued that those same measures overstate price increases because they fail to account for quality improvements in what we buy. Both things can be true! iPhones are really amazing and it’s harder to make ends meet. Untrue, though, is the belief that all this has been quietly settled in an arcane matter that conveniently validates the preferences of a small set of right-leaning scholars; to the contrary, from the outside such a belief looks faintly ridiculous.

I encountered another bout of statistical gnosticism recently in the replies to my essay on The Rise of Wall Street and the Fall of American Investment. The essay reviews the data on declining business investment in America, presents original analysis of how publicly-traded firms behave, and makes the case that the financial sector has become a drag on the real economy and must be curbed with policy reforms. There is, to be sure, much to disagree with there, and I hoped it might spur an interesting debate. The response, instead, was that actually business investment has not weakened.

This claim is bizarre, and flies in the face of a broad consensus among economists and commentators. For instance, Nobel laureate Michael Spence wrote in the Wall Street Journal in 2015 that, “Business investment in the real economy is weak. … We believe that QE has redirected capital from the real domestic economy to financial assets at home and abroad.” Jason Furman, then chair of President Obama’s Council of Economic Advisers, observed the same year that, “The investment slowdown is particularly notable because it comes at a time of high cash flows for businesses and substantial accumulated earnings both domestically and overseas.” Here’s The Economist: “While gross business investment has held steady, net business investment—that is, investment adjusted for depreciation—has been in long-run decline.”

Here’s an IMF paper from 2016: “We explore firm-level data on investment and document that investment fell relative to fundamentals at the turn of the millennium—well before the Great Recession.” Here’s a Brookings paper from 2017: “We analyze private fixed investment in the U.S. over the past 30 years. We show that investment is weak relative to measures of profitability and valuation.” Here’s the American Enterprise Institute’s Jim Pethokoukis in 2019: “And while the so-so recovery after the Great Recession has seen weak business investment, the same is true across rich-country economies.”

As with household budgets, the issue of investment trends is by no means straightforward. Debating it is important. But rejecting outright the weight of available evidence, in favor of a strained reading that conveniently forecloses consideration of real challenges, is unwise. The position demands a particularly awkward contortion in that it gets adopted only situationally, when faced with an argument about issues with businesses or the financial sector. When discussion turns to the corporate tax rate, the talking point returns that, “When President Trump took office, the U.S. had the highest corporate tax rate in the developed world and had experienced a decade of slow growth, low investment and stagnant employment and real wages.”

A final example, from this past week, concerns the question of inequality. American Compass published a primer on inequality trends in which we intentionally relied upon straightforward presentations of government data from the U.S. Census Bureau and the Federal Reserve. Predictably, this was declared “garbage” and “intellectual mush.” By contrast, Scott Winship’s 3,000-word effort at adjusting the data until the problem of rising inequality vanished was lauded as “going for the kill,” a “thorough dismantling,” and an “unfair fight.” (The martial rhetoric of statistical gnosticism is peculiar, if pervasive.) The subtitle placed on the piece at The Dispatch, “the latest from American Compass uses less than ideal data,” offered a more accurate, and amusingly tepid, description of what was going on.

Without going through every problem in Winship’s analysis, his first complaint about our use of the Census Bureau’s calculation of the “Gini coefficient” gives a good sense of his project and its shortcomings. The Gini coefficient is a measure of inequality calculated as a value between 0 (everyone earns the same income) and 1 (one earner has all the income). After three paragraphs of objection to the Census measure, Winship declares that “the best Gini coefficient estimates readily available are from the Congressional Budget Office.” So let’s take a look at the Census measure we published versus the CBO version:

The trends look almost exactly the same, with the CBO reporting higher levels of inequality. (Note: the line shown here is for “income before taxes and transfers,” which includes social insurance programs like Social Security and Medicare; a measure based purely on market income shows greater inequality.) Overall, for the years both datasets are available (1979–2017), the Census Bureau reports the value rising from .37 to .46 (9 points) while the CBO reports it rising from .41 to .52 (11 points). For the most recent two decades, 1997–2017, the Census estimate is +3 points and the CBO estimate is +4 points.

What, then, was the point of the endless complaints about methodological adjustments? And from where does Winship reach his conclusion that, “CBO shows inequality lower in 2017 than in 2000, 2005, 2006, 2007, or 2012. The CBO data also indicate that three-quarters of the increase in inequality from 1979 to 2017 was over by 1988. … Combine these estimates with the pre-1967 estimates from the Census Bureau and it looks like the only substantial increase in inequality since 1947 occurred during the 1980s.”

In fact, Winship’s complaints aren’t technical at all. He wants to use a different data series, provided by CBO, which shows income inequality after taxes and transfers. That is, he wants to count government benefits like Medicaid and SNAP in the “income” of lower-income households, which of course would reduce apparent inequality and slow its rise in recent years as those programs became larger. Here is his preferred line in red:

The question of when to use which line is an interesting conceptual one worth debating. But the handwaving about “smaller methodological changes that shift the post-2012 (see Appendix D) and post-2016 estimates upward” is a distraction. And Winship’s attempt to characterize any of these lines as “the rising income inequality that did not happen” is, to be polite, suspect.

Importantly, CBO emphasizes repeatedly that by far the largest and fastest-growing element of the means-tested transfers included in the “After Taxes and Transfers” calculation is Medicaid spending. How and when to value Medicaid is an important analytical question that I have written about at great length elsewhere, but for purposes of this discussion the vital point is that Medicaid is quite obviously not disposable income. So a post-transfer metric seems the wrong choice in making the claim, as Winship does, that “inequality in disposable income seems not to have risen noticeably in 30 years.”

Statistical gnosticism has done a great deal of damage to right-of-center policymaking in recent years, creating a small epistemological echo chamber that obscures troubling economic trends and seals political leaders off from the experiences of their constituents. Statistics can be misleading and it is always important to understand the assumptions underlying them and reconcile disconnects between the spreadsheets and the real world. But the task is to understand what is happening, not to construct an alternate reality that best supports a particular ideology. (Winship complains at least eight times that the Compass analysis mirrors the “center-left,” “progressives,” Hillary Clinton, or Elizabeth Warren, as if that represents an independent basis for critique.)

There will always be a market for the message that everything is fine, especially among those in society’s upper echelons for whom everything is fine. A successful conservatism will have to approach such questions with much more humility, a spirit of open inquiry, and a willingness to grapple with the nation’s challenges as they are.