Does Bipartisanship Pay? Executive Manipulation of Legislative Coalitions During the George W. Bush Presidency

Abstract

Signing statements have become a common tool used by modern presidents to manipulate the implementation of public policy. Most previous work assumes that the inter-branch environment explains the president's usage of these tools. Little attention is paid to how dynamics within Congress may affect the president's decision-making. I advance a theory of legislative bargaining, coalition formation, and strategic calculus by the president that reveals when, how, and why presidents issue these statements. I hypothesize that we should be more likely to see signing statements on bills that pass through Congress with bipartisan support and with large enacting coalitions, regardless of the inter-branch context. Further, I argue that the magnitude of the challenges in these statements should also vary with the legislative cohesion on a bill. I evaluate these hypotheses with data on signing statements from the George W. Bush administration and find strong support in favor of the claims.

Acknowledgments

The author thanks Ken Moffett and Ian Ostrander for their helpful comments and suggestions and Chloe Dennison for her diligent research assistance.

Notes

1 HR 3763 passed through the 107^th House 423–3 and the 107^th Senate 99–0.

2 Quote from public statements of George W. Bush: https://www.govinfo.gov/content/pkg/WCPD-2002-08-05/pdf/WCPD-2002-08-05-Pg1283.pdf (page 1285).

3 Text available here: https://www.presidency.ucsb.edu/documents/statement-signing-the-sarbanes-oxley-act-2002)

4 Or she, one day

5 To this point, there has been little systematic analysis of the efficacy of signing statements, as implementation is a notoriously difficult thing to study. This is especially complicated because many signing statements are issued on policies with national security implications. However, the evidence we do have suggests signing statements can and do have policy effects.

6 The text of the legislation is available here: https://www.congress.gov/bill/109th-congress/

7 See GAO report here: https://www.gao.gov/products/A70992. It is important to note that the GAO study did not examine whether the signing statement itself was abided by the executive agencies, nor could it say definitively that the signing statement was the cause of the enforcement issue. Further, the GAO results should be carefully interpreted, as the study was not done systematically. The GAO examined only legislation that had been challenged by the president in statements, though they did look at provisions that both were and were not specifically challenged.

8 Remarks available here: https://fas.org/irp/congress/2006_hr/062706leahy.html

9 Remarks available here: https://fas.org/irp/congress/2007_hr/signing.html

10 Quote from DOJ memo to Kenneth Cribb from Ralph Tarr (page 13): https://www.archives.gov/files/news/samuel-alito/accession-060-89-269/Acc060-89-269-box3-SG-ChronologicalFile.pdf.

11 The Sarbanes-Oxley Act amended the U.S. Code to say that those who “corruptly” alter or destroy documents or obstruct justice can be fined or imprisoned under the law. In his signing statement Bush claimed that his administration will interpret “corruptly” to mean “requiring proof of a criminal state of mind on the part of the defendant” in order to protect constitutional rights of those who petition the government with grievances. This was not only at odds with the intent of the law but with part of the U.S. Code, which maintains that for such offenses “no state of mind need be proved with respect to the circumstances” (Birdsell 2007; Rountree 2012).

12 The PATRIOT Act Reauthorization passed with a relatively small enacting coalition (340 legislators) and was not passed along bipartisan lines (the bill was highly partisan in the House, though bipartisan in the Senate). Thus, the attempts to override the signing statement on this bill are consistent with expectations.

13 I believe this logic should hold under both unified and divided government. When there is unified government, the president’s fellow partisans can likely benefit from smaller shifts in the policy that were the result of concessions made in bargaining with the other party. In times of divided government, the concessions were likely even greater and the incentive to maintain a signing statement higher.

14 Fowler and Marshall argue that signing statements should be more likely than vetoes when the president faces a veto-proof enacting coalition. This logic is consistent with, though distinct from, the logic presented here. I am not analyzing the president’s choice to issue a signing statement in direct comparison with the veto calculus and hold that large coalitions, whether or not they are veto-proof, are more likely to result in bills that receive signing statements.

15 Please note that, as a robustness check, I use a categorical dependent variable that also considers rhetorical statements as an option to the president. These findings are discussed in the results section; I find that coalition dynamics affect the issuance of constitutional statements but not rhetorical ones, as expected.

16 The data is publicly available at http://congressionalbills.org/. For this analysis, I excluded all commemorative and private bills, as constitutional signing statements are unlikely to be attached to such legislation. These bills are likely to pass with large margins and bipartisan support, as they often do not have meaningful or widespread policy implications. As such, their inclusion would bias against finding the hypothesized effects. This choice reflects other work on presidential power. Fowler and Marshall (2017) advance a data set of nontrivial legislation during the same period of time for their analysis of vetoes and signing statements. Their data line up closely with that used here, and the results of the analyses advanced do not change if this alternate data set is used.

17 There are some constitutional signing statements in which the president does make a constitutional or policy objection but does not mention a specific part or section of the bill.

18 Of course, there may be a bias in which bills receive roll call votes within Congress, which could affect the results (Lynch and Madonna 2013). I hold that it is most likely that noncontroversial legislation is systematically passed by voice vote, whereas controversial, partisan, and close votes will be over-represented in roll call voting. If this is the case, I believe that any roll call bias will bias against finding results consistent with the theory advanced.

19 However, the results do hold when examining the effects on a chamber by chamber basis as well. These results are provided in a methodological appendix.

20 In addition to this measure of bipartisan support, I also coded for whether a coalition was cross-partisan in nature as an additional way of testing the hypothesis of interest. A cross-partisan coalition is one for which a majority of one party supports a bill but relies on a critical portion of the opposite party to pass the legislation. The results held.

21 In addition to using the raw count of “yea” votes to capture the size of the enacting coalition, I also use an indicator that captures the ratio of yea votes to all those cast on a particular piece of legislation to test the coalition size hypothesis. The results held.

22 In the methodological appendix to this article, I utilize cosponsorship data rather than roll call data to evaluate the hypothesis of interest, as an additional test on the theory. I find some evidence that as the cosponsoring coalition grows in size and becomes more bipartisan, signing statements are more likely.

23 I also controlled for importance using CQ’s Key Vote measure as an alternative. The results held when I swapped out the measures. Fifty-five of the included bills made it onto CQ’s list.

24 Data is is publicly available at http://www.policyagendas.org

25 Please note that Evans (2011) finds that sections of legislation within a particular bill are more likely to be challenged if they implicate the constitutional powers of the president. While his analysis is not at the bill-level, as mine is, his findings generally support my argument that bills that implicate the president's constitutional powers are also more likely to receive signing statements.

26 I re-evaluated the Models 1 and 2 by relaxing the hierarchical structure of the multilevel modeling technique, given that such a modeling procedure demands a lot from a relatively small dataset (only 130 observations). To do so, I treated unified government as a bill-level indicator and re-estimated the models using simple logistic regressions. The results held for all models.

27 The predicted probability for omnibus bills increases from 20% to 66.7%. For foreign affairs bills, the increase is from 20% to 57.6%.

28 The range of coalition size is from 260 to 525. All other predictors are held at zero.

29 This assumes average coalition size, unified government, and that all other controls are held at zero,

30 This assumes all other indicators are held at zero,

31 The probability of the president signing the bill without a statement decreases from 83.6% to 51.4%.

32 On top of those described here, I also estimated pooled models for each Congress separately. These analyses seriously restricted the number of bills analyzed. Coalition dynamics were significant in two of the four Congress-specific analyses, which is not surprising given the small data sets. However, the substantive results on coalition dynamics were large and positive for all four Congresses, as expected. This was true even in the 110^th Congress, during which Bush’s use of signing statements declined due to congressional reaction and his approval rating also dropped.

33 The dependent variable is over-dispersed. The mean is 7.2 and the standard deviation is 7.7. The assumption for Poisson models is that the mean is equal to the variance, which is clearly not the case here. Also, because of the extremely small data set (57 observations), this is not estimated as a multi-level model. Unified government is treated as a bill-level predictor in the negative binomial models.

34 HR 5005 (Homeland Security Act of 2002) passed with an enacting coalition of 385 across both chambers in the 107^th Congress, whereas HR 2673 (Consolidated Appropriations Act of 2004) passed with a coalition of 307 in the 108^th Congress. Neither coalition was bipartisan in nature.

35 This assumes all variables are held at zero except for unified government.