I recently wrote about a new study from economists John Dawson and John Seater that shows that federal regulations have slowed economic growth in the US by an average of 2% per year. The study was novel and important enough from my perspective that it deserved some detailed coverage. In this post, which is part two of a three part series (part one here), I go into some detail on the various ways that economists measure regulation. This will help put into context the measure that Dawson and Seater used, which is the main innovation of their study. The third part of the series will discuss the endogenous growth model in which they used their new measure of regulation to estimate its effect on economic growth.
From the macroeconomic perspective, the main policy interventions—that is, instruments wielded in a way to change individual or firm behavior—used by governments are taxes and regulations. Others might include spending/deficit spending and monetary policy in that list, but a large percentage of economics studies on interventions intended to change behavior have focused on taxes, for one simple reason: taxes are relatively easy to quantify. As a result, we know a lot more about taxes than we do about regulations, even if much of that knowledge is not well implemented. Economists can calculate changes to marginal tax rates caused by specific policies, and by simultaneously tracking outcomes such as changes in tax revenue and the behavior of taxed and untaxed groups, deduce specific numbers with which to characterize the consequences of those taxation policies. In short, with taxes, you have specific dollar values or percentages to work with. With regulations, not so much.
In fact, the actual burden of regulation is notoriously hidden, especially when directly compared to taxes that attempt to achieve the same policy objective. For example, since fuel economy regulations (called Corporate Average Fuel Economy, or CAFE, standards) were first implemented in the 1970s, it has been broadly recognized that the goal of reducing gasoline consumption could be more efficiently achieved through a gasoline tax rather than vehicle design or performance standards. However, it is much easier for a politician to tell her constituents that she will make auto manufacturers build more fuel-efficient cars than to tell constituents that they now face higher gasoline prices because of a fuel tax. In econospeak, taxes are salient to voters—remembered as important and costly—whereas regulations are not. Even when comparing taxes to taxes, some, such as property taxes, are apparently more salient than others, such as payroll taxes, as this recent study shows. If some taxes that workers pay on a regular basis are relatively unnoticed, how much easier is it to hide a tax in the form of a regulation? Indeed, it is arguably because regulations are uniquely opaque as policy instruments that all presidents since Jimmy Carter have required some form of benefit-cost analysis on new regulations prior to their enactment (note, however, that the average quality of those analyses is astonishingly low). Of course, it is for these same obfuscatory qualities that politicians seem to prefer regulations to taxes.
Despite the inherent difficulty, scholars have been analyzing the consequences of regulation for decades, leading to a fairly large literature. Studies typically examine the causal effect of a unique regulation or a small collection of related regulations, such as air quality standards stemming from the Clean Air Act. Compared to the thousands of actual regulations that are in effect, the regulation typically studied is relatively limited in scope, even if its effects can be far-reaching. Because most studies on regulation focus only on one or perhaps a few specific regulations, there is a lot of room for more research to be done. Specifically, improved metrics of regulation, especially metrics that can be used either in multi-industry microeconomic studies or in macroeconomic contexts, could help advance our understanding of the overall effect of all regulations.
With that goal in mind, some attempts have been made to more comprehensively measure regulation through the use of surveys and legal studies. The most famous example is probably the Doing Business index from the World Bank, while perhaps the most widely used in academic studies is the Indicators of Product Market Regulation from the OECD. Since 2003, the World Bank has produced the Doing Business Index, which combines survey data with observational data into a single number designed to tell how much it would cost to “do business,” e.g. set up a company, get construction permits, get electricity, register property, etc., in set of 185 countries. The Doing Business index is perhaps most useful for identifying good practices to follow in early to middle stages of economic development, when property rights and other beneficial institutions can be created and strengthened.
The OECD’s Indicators of Product Market Regulation database focuses more narrowly on types of regulation that are more relevant to developed economies. Specifically, the original OECD data considered only product market and employment protection regulations, both of which are measured at “economy-wide” level—meaning the OECD measured whether those types of regulations existed in a given country, regardless of whether they were applicable to only certain individuals or particular industries. The OECD later extended the data by adding barriers to entry, public ownership, vertical integration, market structure, and price controls for a small subset of broadly defined industries (gas, electricity, post, telecommunications, passenger air transport, railways, and road freight). The OECD develops its database by surveying government officials in several countries and aggregating their responses, with weightings, into several indexes.
By design, the OECD and Doing Business approaches do a good job of relating obscure macroeconomic data to actual people and businesses. Consider the chart below, taken from the OECD description of how the Product Market Regulation database is created. As I wrote last week and as the chart shows, the rather sanitized term “product market regulation” actually consists of several components that are directly relevant to a would-be entrepreneur (such as the opacity of a country’s licenses and permits system and administrative burdens for sole proprietorships) and to a consumer (such as price controls and barriers to foreign direct investment). You can click on the chart below to see some of the other components that are considered in OECD’s product market regulation indicator.
Still, there are two major shortcomings of the OECD data (shortcomings that are equally applicable to similar indexes produced by the World Bank and others). First, they cover relatively short time spans. Changes in regulatory policy often require several years, if not decades, to implement, so the results of these changes may not be reflected in short time frames (to a degree, this can be overcome by measuring regulation for several different countries or different industries, so that results of different policies can be compared across countries or industries).
Second, and in my mind, more importantly, the Doing Business Index is not comprehensive. Instead, it is focused on a few areas of regulation, and then only on whether regulations exist—not how complex or burdensome they are. As Dawson and Seater explain:
[M]easures of regulation [such as the Doing Business Index and the OECD Indicators] generally proceed by constructing indices based on binary indicators of whether or not various kinds of regulation exist, assigning a value of 1 to each type of regulation that exists and a 0 to those that do not exist. The index then is constructed as a weighted sum of all the binary indicators. Such measures capture the existence of given types of regulation but cannot capture their extent or complexity.
Dawson and Seater go out of their way to mention at least twice that the OECD dataset ignores environmental and occupational health and safety regulations. Theirs is a good point – in the US, at least, environmental regulations from the EPA alone accounted for about 15% of all restrictions published in federal regulations in 2010, and that percentage has consistently grown for the past decade, as can be seen in the graph below (created using data from RegData). Occupational health and safety regulations take up a significant portion of the regulatory code as well.
In contrast, one could measure all federal regulations, not just a few select types. But then the process requires some usage of the actual legal texts containing regulations. There have been a few attempts to create all-inclusive time series measures of regulation based on the voluminous legal documents detailing regulatory activity at the federal level. For the most part, studies have relied on the Federal Register, the government’s daily journal of newly proposed and final regulations. For example, many scholars have counted pages in the Federal Register to test for the existence of the midnight regulations phenomenon—the observation that the administrations of outgoing presidents seem to produce abnormally large numbers of regulations during the lame-duck period
There are problems with using the Federal Register to measure regulation (I say this despite having used it in some of my own papers). First and foremost, the Federal Register includes deregulatory activity. When a regulatory agency eliminates words, paragraphs, or even entire chapters from the CFR, the agency has to notify the public of the changes. The agency does this by printing a notice of proposed rulemaking in the Federal Register that explains the agencies intentions. Then, once the public has had adequate time to comment on the agencies proposed actions, the agency has to publish a final rule in the Federal Register—another set of pages that detail the final actions the agency is taking. Obviously, if one is counting pages published in the Federal Register and using that as a proxy for the growth of regulation, deregulatory activity that produces positive page counts would lead to incorrect measurements.
Furthermore, pages published in the Federal Register may be a biased measure because the number of pages associated with individual rulemakings has increased over time as acts of Congress or executive orders have required more analyses. In his Ten-Thousand Commandments series, Wayne Crews mitigates this drawback to some degree by focusing only on pages devoted to final rules. The Ten-Thousand Commandments series keeps track of both the annual number of final regulations published in the Federal Register and the annual number of Federal Register pages devoted to final regulations.
Dawson and Seater instead rely on the Code of Federal Regulations, another set of legal documents related to federal regulations. Actually, the CFR would be better described as the books that contain the actual text of regulations in effect each year. When a regulatory agency creates new regulations, or alters existing regulations, those changes are reflected in the next publication of the CFR. Dawson and Seater collected data on the total number of pages in the CFR in each year from 1949 to 2005. I’ve graphed their data below.
*Dawson and Seater exclude Titles 1 – 3 and 32 from their total page counts because they argue that those Titles do not contain regulation, so comparing this graph with page count graphs produced elsewhere will show some discrepancies.
Perhaps the most significant advantage of the CFR over counting pages in the Federal Register is that it allows for decreases in regulations. However, using the CFR arguably has several advantages over indexes like the OECD product market regulation index and the World Bank Doing Business index. First, using the CFR captures all federal regulation, not just a select few types. Dawson and Seater point out:
Incomplete coverage leads to two problems: (1) omitted variables bias, and, in any time series study, (2) divergence between the time series behavior of subsets of regulation on the one hand and of total regulation on the other.
In other words, ignoring potentially important variables (such as environmental regulations) can cause estimates of the effect of regulation to be wrong.
Second, the number of pages in the CFR may reflect the complexity of regulations to some degree. In contrast, the index metrics of regulation typically only consider whether a regulation exists—a binary variable equal to 1 or 0, with nothing in between. Third, the CFR offers a long time series – almost three times as long as the OECD index, although it is shorter than the Federal Register time series.
Of course, there are downsides to using the CFR. For one, it is possible that legal drafting standards and language norms have changed over the 57 years, which could introduce bias to their measure (Dawson and Seater brush this concern aside, but not convincingly in my opinion). Second, the CFR is limited to only one country—the United States—whereas the OECD and World Bank products cover many countries. Data on multiple countries (or multiple industries within a country, like RegData offers) allow comparisons of real-world outcomes and how they respond to different regulatory treatments. In contrast, Dawson and Seater are limited to constructing a “counterfactual” economy – one that their model predicts would exist had regulations stayed at the level they were in 1949. In my next post, I’ll go into more detail on the model they use to do this.