Legislatures in separation-of-powers systems regularly delegate policymaking authority to the executive branch. In this research note, we ask a central question about delegation in the US context: What percentage of major US laws delegate policymaking authority to the executive branch? We view this as a significant question for three reasons. First, there is a conventional wisdom regarding this issue, but little systematic data. Second, existing theories often presume that Congress chooses not to delegate. And third, the two previous attempts to answer this question produced widely varying estimates. We devise a new approach to assessing whether a law delegates policymaking responsibility to agencies, one that improves markedly on past assessments. Using this approach, we find that virtually all major laws in the US—more than 99 percent—contain delegation to government agencies. When Congress legislates, it delegates.
We reach this conclusion by drawing upon new data drawn from a source that has been little noticed or used by congressional scholars and by devising a new approach for identifying whether a law delegates. Our approach thus provides an innovative measure of the frequency of delegation, one that improves dramatically on existing estimates and provides the best systematic answer to our central question. It also allows us both to contribute a better sense of how frequently major laws delegate and also to generate a fresh set of questions about delegation that can build on our findings.Footnote 1
1. Perspectives on delegation
The stylized version of how delegation might occur is well known: when Congress sets out to make policy, it can spell out the details of policy itself or it can choose to either direct or allow agencies to develop the details of policy. What is not well known, however, is how often Congress chooses to delegate. Conventional wisdom holds that Congress regularly delegates to executive agencies. For example, as one prominent administrative law expert observed, “Congress regularly delegates broadly to agencies…Agency powers typically include the discretion to set far-ranging and expensive regulatory standards” (Mendelson, Reference Mendelson2010, 135). But how often is “regularly”? More specifically, when Congress passes major new laws, how many of these laws delegate policymaking authority to federal agencies? Surprisingly, despite the relevance of delegation to a thorough understanding of the US's separation-of-powers system, scholars know little about the answer to this question.
The well-developed theoretical literature on delegation often assumes that Congress makes use of the option to not delegate.Footnote 2 These studies typically begin with the assumption that when writing a law, a legislature decides between delegating to an agency and not doing so.Footnote 3 Some of these studies (e.g., Fiorina Reference Fiorina1986) investigate how Congress chooses between delegating to agencies and delegating to courts.Footnote 4 Their starting point is that Congress can do either—that is, sometimes it will choose to delegate to agencies, while other times it will choose to delegate not to agencies but instead to courts.Footnote 5 Other studies focus more generally on whether to delegate at all. In this view, Congress can either choose to delegate or it can “write detailed, exacting laws, in which case the executive branch will have little or no input into policy” (Epstein and O'Halloran, Reference Epstein and O'Halloran1999a, 961).Footnote 6
These theoretical approaches assume that Congress delegates only some of the time. But is this the case? More specifically, does Congress write some laws that delegate, but others that do not? We currently do not know the answer to this question. Most models assume that Congress can and does choose not to delegate, but as we will see, our analysis and results bring this assumption into question.
Finally, and perhaps most importantly, the only two studies that attempt to empirically assess which laws contain delegation reached widely varying conclusions. Farhang and Yaver (Reference Farhang and Yaver2016), as part of their insightful investigation of why Congress sometimes fragments responsibility for policymaking across agencies, began by first determining which laws delegate. Employing research assistants who read the texts of laws that Mayhew (Reference Mayhew2005) classified as major and having these students look for examples of language within each law that give agencies regulatory authority, their method revealed that 59.6 percent of these laws contained delegation.Footnote 7
Epstein and O'Halloran (Reference Epstein and O'Halloran1999a, Reference Epstein and O'Halloran1999b) took a different approach, using research assistants to code the summaries of laws that appear in each year's Congressional Quarterly Almanac. More specifically, they coded whether each summary indicated that a law delegated authority to any other political actor (and also whether procedural provisions constrained this delegation). Of the 257 laws they coded (again using Mayhew's list of major laws), they found that approximately 89 percent of major laws delegate policymaking responsibility—a considerably higher percentage than Farhang and Yaver.Footnote 8 Thus, there is broad disagreement across the only two empirical studies that systematically addressed this noteworthy topic.Footnote 9 What this means is that any study asking a question that first requires an assessment of whether a law delegates—for example, examining which delegating laws fragment authority, which ones allow for discretion, or anything including the number of delegating laws in a denominator—currently lacks an appropriate, accurate baseline.
2. Our approach
Using newly available data, we analyze more than 400 major laws over seven decades. We find three striking results. First, most of these laws—more than 99 percent—contain delegation to federal agencies. Thus, we find that delegation is far more common than earlier estimates suggest. Virtually every time Congress passes a major law, it delegates some policymaking authority to an agency. Second, the number of agencies receiving delegated authority from these laws increased over time, from the 1940s and 1950s through the 1980s and 1990s before declining slightly since 2000. Third, we find divided control of government has little effect on whether a major law delegates or on the number of delegated agencies.
These findings are surprising, not just because they run counter to empirical claims and theoretical assumptions, but also because it is seemingly so easy to identify laws that appear to lack delegation. Consider, for example, the Public Health Cigarette Smoking Act (PHCSA).Footnote 10 In this concise, three-page law, Congress required all cigarette packages manufactured or sold in the US to display warning labels. It provided the precise wording for the warnings and mandated their location and appearance. The law banned all advertising of cigarettes on television and radio and specified when this ban would go into effect. Yet while it appears to be a textbook case of a non-delegating law, a closer look reveals that it does indeed contain delegation. For example, it gives the Federal Trade Commission the power to issue rules covering the advertising of tobacco products, contingent on notifying Congress of its intent to do so, and confirms the Commission's authority to regulate any “unfair or deceptive acts or practices in the advertising of cigarettes.”
The PHCSA demonstrates that even a law that seems to completely specify policy can include delegation. Our goal, then, is to devise an approach that permits us to uncover these hard-to-find examples of delegation. For our analysis, we examined all of the 443 laws between 1947 and 2016 that Mayhew (Reference Mayhew2005) designated as significant laws.Footnote 11 To identify which of these law delegate, we first must determine what to use as an indicator of delegation. We proceed in two steps. To begin, we examine whether laws provide agencies with the authority to promulgate rules and regulations. We rely on a valuable, underutilized new source of information: ProQuest's Regulatory Insight database, which provides a regulatory history for all federal enactments since 1936, including all agencies that propose and issue final rules pursuant to each statute.Footnote 12 Hence, we can identify all agencies (if any) that drew statutory authority for their rules and regulations from each law in our dataset.Footnote 13 This approach, which follows Stiglitz (Reference Stiglitz2018), is conservative in that it is likely to miss ways other than rulemaking authority in which laws can delegate policymaking responsibility to agencies.Footnote 14
Because 38 Mayhew laws were not identified in this database, we turned to a second approach for those laws: conducting systematic searches of the full texts of laws.Footnote 15 We first searched for words or fragments that might indicate delegation, such as “department,” “secretar,” and “administ,” or those that pertain to judicial review of agency actions, like “judi,” “court,” and “appell” (since review provisions imply delegation).Footnote 16 When we found these terms, we then carefully read the surrounding and related sections to determine whether the law delegated regulatory policymaking authority to agencies.Footnote 17
By using these techniques, our approach improves on previous empirical studies. To begin with, we do not need to rely on Congressional Quarterly (CQ) summaries (Epstein and O'Halloran, Reference Epstein and O'Halloran1999b), which can vary in content across laws and over time and might not be consistent in mentioning delegation. Furthermore, while we use the laws themselves as one of our data sources, in contrast to Farhang and Yaver (Reference Farhang and Yaver2016) we do not rely only on reading laws, which can miss instances of delegation. The issue of false negatives—that is, identifying a law as non-delegating when it in fact delegates—is therefore especially a problem in these earlier approaches. A law might contain delegation, but the CQ summary does not mention it if it sees other aspects of the law as more noteworthy. And reading laws is notoriously difficult, especially if research assistants are not specialists in statutory language.Footnote 18 Indeed, we found a surprisingly high number of examples of false negatives in these two earlier analyses. We identify and discuss these examples in Appendix D.
3. Findings
Strikingly, we find that 99.1 percent of these laws—nearly every major law enacted in the US since World War II—contains delegation of policymaking authority to federal agencies.Footnote 19 Only four of the 443 laws did not delegate.Footnote 20 Furthermore, as Table 1 shows, the vast majority of these laws—almost 80 percent of them—delegated to multiple agencies. In the 439 laws we identified as containing delegation, there are 2560 instances of agencies receiving delegation.Footnote 21
Congressional delegation from 1947 to 2016
In Figure 1, we plot the count of the number of agencies that receive delegations of authority in major laws over time. The data show a growth over time in the average number of agencies to which each law delegates. In the 1940s and 1950s, the mean number of agencies per law was 2.8. This average increased slightly to 4.1 in the 1960s and 5.5 in the 1970s, peaked at 9.4 in the 1980s, then decreased to 8 in the 1990s and 6.1 in the 2000s. Compared to the years just after World War II, Congress now more regularly includes bigger bundles of delegation in major laws. Figure 1 also reveals an increase in variance over time, including a handful of identified laws that delegate to 40 or more agencies.Footnote 22 Since our analysis reveals that virtually all major laws contain at least some delegation, the proportion that delegate has stayed constant over time; however, we do find notable time-related patterns related to the number of agencies.
Number of agencies receiving delegated authority in major laws.
Finally, following previous studies of delegation, in Table 2, we consider differences in delegation choices under unified and divided control of government. Of course, since we found that virtually all major laws delegate, we also find that laws are equally likely to contain delegation regardless of whether control of government is unified or divided. Unified control has produced more laws that contain no delegation; still, 98 percent of unified government laws delegate.
Delegation and divided control
a Means differ at p < 0.05 with a two-tailed test.
Table 2's final column provides initial evidence that the mean number of agencies receiving delegated authority is significantly higher under divided than unified control, which is consistent with Farhang and Yaver's finding that divided government is associated with more fragmented authority.Footnote 23 However, since the prevalence of divided government has increased in recent years (i.e., it is not randomly distributed over time), to properly assess this relationship, we must control for the potential confounding effect of time. When we do, the association between divided government and the number of agencies is no longer significant.Footnote 24 Thus, we conclude that there is no significant difference between divided and unified government in terms of the number of agencies receiving delegation per law.
4. Conclusion
Utilizing new data sources and a novel approach, we provide a more precise and accurate measure of which laws delegate and which do not. As a result, we have found—contrary to previous studies—that nearly every major law passed by Congress over the past seven decades includes delegation to federal agencies. Even when Congress spells out policy details, it delegates.
We recognize that one potential reaction to our finding that delegation exists in virtually all laws is that “we already know this.” But in truth, prior to our research, this was not well known. Instead, we had only a general sense that statutes often delegate. When scholars attempted to measure the frequency of delegation, they came up with far lower estimates than we find (and their estimates disagree markedly with each other).
Our findings and our approach point to several avenues for future exploration. A common starting point for theoretical studies has been to ask whether Congress prefers to delegate to agencies, rather than delegating to courts or setting policy entirely on its own. From the starting point of “all laws delegate,” new questions emerge. For example, why does Congress delegate more to some agencies than others? What explains the changes over time in the number of agencies receiving delegated authority? Furthermore, we have looked only at major laws; what does the pattern of delegation look like for less significant laws?
Given that we have shown that almost all major laws delegate to agencies, future studies should now turn their attention to an assessment of the amount of delegation within each law. Our approach offers a count of the agencies with delegated authority in each law, but not how much authority those agencies were given.Footnote 25 Since our approach does not allow us to determine this, we have been cautious in discussing the implications of our findings. For example, we find that Congress is no more likely now to write a law that contains delegation to agencies than it was decades ago. It is possible, though, that grants of delegation are now both broader and more numerous within each law, that delegation is bundled to a variety of different agencies more often, or that states and courts are relied upon in concert with federal agencies.
Moreover, laws can delegate in different ways. Congress can tell an agency exactly what sorts of actions to take. It also can direct an agency to do something, but leave it up to the agency to decide exactly what to do. And it can create a new policy that provides an opportunity for an agency to take action, even if the law does not explicitly give power to the agency. This third scenario can yield ripple effects, where a law, by newly addressing a policy issue, opens the door for an agency to act. The first two types are thus explicit (and intentional) delegation, while the third is implicit (although not necessarily unintentional). Our approach cannot distinguish between these direct and indirect types of delegation; but examining such distinctions would be a welcome avenue for future studies.
Finally, although we show that virtually all major laws contain delegation, we do not examine the complementary issue of what agencies actually do with the grants of delegation they receive. Do these laws provide a base from which agencies pursue major rules? Or do they use them to generate minor rules? Building on previous points, does the extent of delegation, or the type of delegation, influence the types of rules that agencies then pursue?Footnote 26 Surprisingly, little research exists that ties the nature of legislative delegation to the nature of agency actions. By building on our foundation, which shows that nearly all laws delegate, future research can examine variation in the character of delegation over time.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2021.32.
Acknowledgements
We would like to thank Keith Dowding, John Jackson, Kenny Lowande, John Matsusaka, Nina Mendelson, Brendan Nyhan, Zoe Robinson, and Abby Wood for helpful comments and discussions. We would also like to thank Jordan Carr Peterson for excellent research support. Shipan would like to acknowledge the support of the RSSS fellowship at Australian National University.
