NAM’s “Cost of Regulations” Estimate: An Exercise in How Not to Do Convincing Empirics
The latest effort to scaremonger about a rising regulatory burden on U.S. business was released yesterday by the National Association of Manufacturers (NAM). The report, by W. Mark Crain and Nicole V. Crain (C&C, henceforth) purports to (among other things) estimate the total cost of U.S. regulations. The claimed price tag is enormous—$2.1 trillion. The bulk of these costs (75 percent) are estimated using a cross-country regression analysis. This cross-country analysis, however, is completely unconvincing and should be ignored.
An earlier C&C study used a similar methodology as yesterday’s release—that study was shown to be deeply flawed by an EPI analysis. Yet, the methodology of the current study is largely the same. In fact, if anything, the current analysis is less robust and convincing than the previous one.
C&C undertake a cross-country regression analysis across 34 OECD countries for the years 2006–2013. This obviously leads to a first reason for being wary of results—one would expect the economic performance of rich countries during the 2006–2013 period to be utterly dominated by the Great Recession and its aftermath. C&C use dummy variables to control for the years 2008 and 2009, presumably to capture the official years of the Great Recession. But most of the OECD remains operating far below potential even in 2013 due simply to a shortfall of demand. Unless one is making the case that regulations impede recovery from recessions (a claim that they do not make) then it is extraordinarily hard to make large inferences about the effects of regulations on long-run economic performance in this short sample period.
The dependent variable used by C&C is the level of inflation-adjusted per capita gross domestic product. This is also cause for concern, as it does not reflect best practice in empirical examinations of cross-country economic performance. The vast majority of economic literature on comparative national performance aims to explain growth rates, not levels. Nearly every citation in the reference section that accompanies their analysis uses growth rates instead of levels. Additionally, if one wants to control levels of GDP across countries, it is standard to use purchasing power parity (PPP) measures of GDP. PPP adjusts measures of GDP across countries to reflect price differences. So if, for example, the per capita GDP of Norway is large relative to the United States when converted into U.S. dollars using market exchange rates, one would still want to know what this level of dollar-denominated GDP actually would buy Norwegian residents—meaning one would need information on comparative prices. The answer (in this particular case) is quite a bit less. Norwegian per capita GDP measured simply in U.S. dollars in 2011 was over $58,000, but dropped by about 10 percent if measured using PPPs that reflected the higher prices in Norway. For other countries this PPP adjustment would go the other way. A key question is whether or not the C&C results hold using this more appropriate measure of international living standards comparisons.
A further concern is the list of independent variables that C&C use to control for other influences on the level of GDP. Most glaring is the omission of any measure of the educational attainment of the workforce. Nearly all empirical growth studies have shown a powerful link between educational attainment in a country and GDP performance (both level and growth). In their previous study, they used a measure of educational attainment that actually had the wrong sign—showing that lower rates of education were associated with higher levels of GDP. The current study simply has no control for educational attainment.
An odd choice for inclusion into the regression is the size of the labor force. This will be mechanically and positively related to levels of GDP per capita. GDP for any nation is simply the total employment multiplied by average productivity. Total employment is simply the employment rate multiplied by the workforce. It is unclear what this variable is meant to capture.
Perhaps most importantly, the key independent variable—the measure of regulatory burden—is extremely problematic. It is a simple average of three surveys of subjective opinions of business executives about (1) the burden of government regulation, (2) the efficiency of the legal framework in challenging a regulation, and (3) the regulation of securities exchanges. The surveys simply ask executives to rate, say, the burden of government regulation from 1 to 7, with 1 being “extremely burdensome” to 7 being “not burdensome at all.” The benchmark against which C&C measures U.S. performance is the 5 highest-performing countries on each measure. These countries are generally: Norway, Finland, Sweden, the Netherlands, Luxembourg, Australia, and New Zealand. It is extremely unclear just what it could mean for regulatory policy to move the United States’s score on subjective regulatory burden from 3.4 to the 5-country benchmark average of 4.4. What possible policy tool or regulatory action could this map onto?
Further, it’s worth noting that the three indices averaged together by C&C could well be telling very different stories. The first index is a measure of regulatory burden as assessed by business executives. This is straight-forward enough (if incredibly difficult to interpret or operationalize). But the third index is “how effective is securities regulation in your country?”. For some people, effective securities regulation would imply very tight limits on what financial intermediaries can do. For others, effective regulation could imply a light regulatory touch. Again, operationalizing any claim that the U.S. should simply “increase” its score on this index is impossible.
It is also worth noting that these are just three of the sub-indices of a broader “competitiveness index” compiled by the World Economic Forum (WEF). On the broader index (which includes these regulatory measures) the United States is very highly-ranked. Other indicators besides subjective measures of regulatory burdens in this index may actually directly require regulatory interventions. For example, measures of the insurance of product market competition which sometimes demands more regulation—enforcement of antitrust law, for example. It also includes measures of broadband penetration, which often requires regulations that require telecom companies to provide service to remote areas.
Additionally, it’s worth noting that while the NAM report claims it is measuring the cost of “federal” regulations, there is so such reference in the measures of regulatory burdens used in the cross-country analysis that generates most of C&C’s results. The low score of the United States on these regulatory indices could in fact be driven by state and local regulations.
Of course, they also could be driven simply by a more partisan or ideological business class in the United States relative to other countries. The one undeniable thing the three indices used by C&C seem to be saying is that business executives in Scandinavia and Oceania seem less bothered by government regulation than U.S. executives do. This could be because regulations really are more economically expensive in the United States, but could also be because U.S. business executives are simply more partisan and ideological than their peers.
Is there any prima facie case to doubt the genuine economic burden of U.S. regulations and claims that growing regulatory burdens impose any growing drag on American growth? Well, yes. To the degree that regulations burden businesses specifically (a key claim of the C&C report), it should be by boosting the cost of doing business to the extent that profitability suffers. And yet profitability in the United States in recent years has certainly not suffered from anything outside the effect of the Great Recession. Profit margins in 2006 and 2007, right before the Great Recession, were high in historical perspective, and after falling sharply during the recession they reached their highest levels in nearly five decades during recent years (see figure H here).
All in all, the C&C report’s eye-catching number on the cost of regulations is, like their last report on this topic, far too flawed to form any basis for policymaking.