Policymaking by the Executive: Examining the Fate of Presidential Agenda Items

Abstract

Politicians routinely campaign on a large platform of new policies and policy changes they hope to pursue once in office. Presidents, however, are unique in their ability to accomplish this given the multiple avenues by which they can enact policy change. Using each presidential agenda between 1964 and 2010, we evaluate the circumstances under which a president’s policy goal is realized via an executive order or a law passed by Congress. Previous studies have only considered these avenues independently, but when we consider them in concert with one another, new patterns emerge. Our results indicate that a president’s proposal is most commonly addressed in tandem with Congress through the traditional legislative process. However, the level of polarization, presidential approval, the composition of government, and the type of issue being addressed increase the probability of a proposal being addressed unilaterally via executive order rather than by passing a law.

Notes

1 See Austin Ramzy, “McCain’s Vote Provides Dramatic Moment in 7-Year Battle Over Obamacare,” July 28, 2017, New York Times, https://www.nytimes.com/2017/07/28/us/politics/john-mccain-vote-trump-obamacare.html.

2 See Peter Sullivan, “Trump Eases ObamaCare Rules with Executive Order,” October 12, 2017, The Hill, https://thehill.com/policy/healthcare/355123-trump-eases-obamacare-rules-with-executive-order

3 For purposes of this article, we consider any piece of legislation that was passed by Congress and signed by the president to be a law. Any agenda item that was not designated a law or executive order is considered as having received no action.

4 Determining and debating what constitutes a presidential unilateral action has created a relatively robust literature for students of the American executive. Scholars have identified a variety of powers as potential avenues by which the president can create policy without the express consent of Congress. These include administrative strategies of dealing with the bureaucracy (Kennedy 2015; Rudalevige 2009, 2015), signing statements (Cooper 2014; Ostrander and Sievert 2013a, 2013b), executive proclamations (Cooper 2014; Dodds 2013; Rottinghaus and Maier 2007), executive memorandums (Lowande 2014), national security directives and recess appointments (Cooper 2014; Black et al. 2007), disaster declarations (Reeves 2011; Young 2013), and executive orders (Christenson and Kriner 2017a; Fine and Warber 2012; Howell 2003; Mayer 2001; Moe and Howell 1999; Rudalevige 2012, 2015).

5 We do not think that presidential executive orders are meant to supplant law. We understand and agree that if presented with the option of having their policies implemented either through Congress or an executive order, most presidents would choose Congress given the negative stigma sometimes attached to unilateral actions and the relative ease by which they can be undone by Congress or future presidents.

6 This is not meant to insinuate that there are only two ways for presidents to shift public policy toward their own preferences. The president has many means to potentially shift public policy, be it through Congress, executive orders, executive memorandums, or executive proclamations. We are also mindful that this short list of presidential tools is not exhaustive. For our purposes, however, we are focusing solely on executive orders to bring some cohesion to the literature that focuses on executive orders and public policy. In addition, prior research has shown that presidents utilize executive orders in order to shift policy toward the majority within Congress (See Law 2012).

7 The judiciary provides an important check on this authority by intervening from time to time when the president overreaches via executive order as we have seen most recently with Presidents Obama and Trump with respect to immigration.

8 We are not arguing that the legislative branch does not have important oversight responsibilities over the president and the bureaucracy, but both are insulated with respect to how they are able to operate.

9 The rationale for why presidential approval ratings affect legislative success is similar to the theories about why “going public” has the same effect. Because the national electorate votes for presidents and members of Congress are voted for by subsets of the same electorate, it follows that if the president achieves a certain level of mass approval, legislators will vote accordingly (Mayhew 1974).

10 We do not try to argue that every member of Congress will be beholden to the president after an election. We recognize that some will have the same pressures from their constituents, but research has shown some evidence of the honeymoon period across time (Beckmann and Godfrey 2007; Farnsworth and Lichter 2011).

11 We would argue this benefits presidents in terms of their long-term policy objectives because laws created by Congress can more easily withstand shifting political alignments and allegiances.

12 It is important to note that Christensen and Kriner (2017b) present a more nuanced interpretation by showing that Congress can diminish support for some of the president’s actions—including in the foreign policy realm—by raising constitutionally-based objections.

13 For an alternative perspective on legislative effectiveness, see Mayhew (2005).

14 We do not intend to argue that all coalitions in Congress present the same advantages throughout time. For instance, Democratic majorities in the 1960s were split with Southern Democrats aligning themselves with Republicans. With increasing polarization, coalition sizes in the 1990s and 2000s should produce coalitions that are much more aligned with the president. We utilize the size of the coalitions to present an alternative approach to the effects of divided government, which is included in the appendix.

15 We recognize that the failure of Congress to act occasionally has unintended consequences with regard to policy ending up closer to the president’s preferred position. On this point, see especially Law (2012) who argues that most executive orders move preferences closer to both the president and Congress.

16 We make no claims that the president is the only entity in which executive orders originate, but we must assume that the executive branch’s bureaucracy share similar goals on policy objectives.

17 To date, there has been little work that systematically examines why presidents might fail to advance a unilateral initiative given inaction on the part of Congress, most likely because studying non-events can pose a significant challenge empirically. Nevertheless, we believe it is important to control for this potential outcome in our empirical model since it does occur on a fairly regular basis. We would encourage other scholars to investigate this in future research.

18 Most recently, Belco and Rottinghaus (2014) explain when certain conditions are more likely to elicit unilateral action from the president. In addition, Belco and Rottinghaus (2017) explain when presidents use executive orders and proclamations related to topics within and outside the legislative arena. 

19 This time frame was chosen using the data available. However, we feel it provides enough variability in the composition of government and other relevant conditions to produce generalizable results.

20 We followed the direction of Eshbaugh-Soha (2005) when determining which proposals entered the president’s agenda. The proposal had to be mentioned twice publicly (in the State of the Union and somewhere else) to make the agenda. If a proposal had multiple parts, then it was disaggregated and each part was treated as an agenda item. This was done because certain aspects of a presidential proposal might be introduced in legislation while other parts would not.

21 A recent example of this would be President Trump’s decision in April 2017 to sign an executive order dismantling many of the environmental protections set up during the Obama administration. See Brady Dennis and Juliet Eilperin, “Trump Signs Order at the EPA to Dismantle Environmental Protections,” Washington Post, March 28, 2017, https://www.washingtonpost.com/national/health-science/trump-signs-order-at-the-epa-to-dismantle-environmental-protections/2017/03/28/3ec30240-13e2-11e7-ada0-1489b735b3a3_story.html?utm_term=.49d7871c7037

22 Please see the appendix for additional models calculated based on different subsets of the data. We utilize multiple models for a variety of reasons. First, estimating a multinomial model requires assuming the independence of irrelevant alternatives in the dependent variable. This means that each outcome must have the same probability of being selected regardless of the presence or absence of some other outcome. In this case, a multinomial model would only be appropriate if the president’s probability of choosing between signing a bill into law or not signing a bill into law (for whatever reason) was not conditioned on their ability to issue an executive order. As it is reasonable to assume that presidents would rather move policy on their own than see policy unchanged, the IIA assumption is not one we feel comfortable making. Furthermore, the multinomial model would only be appropriate if the president’s probability of issuing an executive order or letting policy remain the same was not affected by their ability to sign a bill into law. Given that executive orders cannot supersede law and require more political capital than signing a bill into law, it is fair to assume that presidents would prefer moving policy through legislative means rather than unilaterally. However, IIA is not an issue when estimating a multinomial probit model, and we therefore include it in our estimates.

In addition, we include all proposals in the first model and code all proposals that become law as 1 and all proposals that do not become law as 0 in a given year. Then, of the proposals that do not become law, we estimate a second binary logistic regression where proposals that became executive orders were coded as 1 and proposals that did not were coded as 0. This allows us to test our hypothesis in a methodologically sound and relatively intuitive way, as well as provides us with added robustness. We also estimate a single logit predicting whether a proposal became an executive order or a law. Although this excludes the possibility of examining a proposal that does not receive any action, it does allow us to evaluate the differences between pieces of legislation that ultimately become public law or an executive order.

Finally, some may argue a selection model is more appropriate to the specific question at hand. However, a selection model does not conform to the theoretical or methodological requirements necessary for us to justify its use here. A selection model accounts for the correlation between potential outcomes. For example, when modeling how a member of Congress feels about a bill, one must first estimate whether she voted or abstained before modeling whether she voted yea or nay. In this case, not offering support comes in the form of abstaining or voting nay, thus necessitating the need for a selection model. To create an analogous model in this article, we would need to know whether a proposal was acted on before and then modeling whether that proposal became a law or an executive order. Without having the entire population of proposals (data that do not exist and whose collection would fall well beyond the scope of this project), such a model cannot be easily estimated.

23 Any agenda item that was not designated a law or an executive order is considered as having received no action for the purposes of our work.

24 To code the executive orders, we follow Eshbaugh-Soha (2005, 2010) and searched for any executive order that contained information relating to the original policy proposal. If any part of the executive order related back to the president’s legislative proposal, that executive order satisfied the requirement of achieving the president’s policy goal.

25 We also estimate our models using the percent of the coalition size of the president’s party in each chamber. The results are substantively the same as when we use the simple dichotomy for divided government. Both variables (divided government and coalition size) have substantial correlation; therefore, having both variables in the same model seemed redundant. For more information on this, see the Appendix.

26 Eshbaugh-Soha (2005, 2010) originally coded presidential proposals into four different categories: minor, incremental, major, and meteoric. Madonna et al. (2016) collapsed the four categories into two, leaving only important and routine. For Madonna et al.’s (2016) purposes and our own, minor and incremental proposals are considered routine, whereas major and meteoric proposals are considered important. An example of an “important” proposal would be President Reagan’s 1985 legislative proposal for tax reform. A “routine” proposal would be similar to that of President Clinton’s proposal for an increase to NSF funding in 2000.

27 We borrow our categorization from Ragsdale (2014). We utilize Ragsdale’s coding scheme and condense them in order to have a more parsimonious model. In addition, we coded all the proposals based on the coding and categories provided by the Comparative Agendas Project. The Comparative Agendas Project provides more nuanced categories than those found in Ragsdale (2014), which allows us to better specify the issue areas of the legislative proposals. Similar to the way we condensed the Ragsdale categories, we applied the same process to the Comparative Agendas Project issue areas. The categories that we utilized produced the same results for the issue areas created by the Comparative Agendas Project as those from Ragsdale (2014). Therefore, we are confident in the condensed categories that we utilize from Ragsdale (2014).

28 The foreign affairs category includes issues such as foreign trade, diplomacy, and foreign aid. The defense, social welfare, and civil rights category include issues relating to defense and everything related to health care (i.e., Social Security and Medicaid). Government and economic management also contain all issues relating to education. The natural resources and environment category contains all agriculture related issues. Last, the other category includes issues relating to ceremonial/cultural, federalism, and personnel/agency requests.

29 Data on congressional polarization comes from www.voteview.com. We have used these data to create the average amount of polarization for a given Congress. Higher levels of polarization indicate that the parties are further apart from one another (Poole and Rosenthal 1997). This tends to lead to more legislative gridlock, which in our case, could signal greater opportunities for presidents to use their unilateral powers.

30 The presidential approval data was obtained from Gallup and indicates the percent of respondents approving of the job the president is doing during the month in which Congress was considering the proposal.

31 See Table A6 in the appendix for an examination of the model with an interaction term between divided government and polarization.

32 Table A1 in the Appendix contains estimates using coalition size instead of a divided government indicator. The results are substantively similar and do not alter our conclusions.