Until recently there has been little scholarly consideration of why presidential vetoes occur, perhaps because the answer seemed obvious. Vetoes occur, one might reasonably conclude, because Congress passes bills the president does not want to become law. Passage of offensive legislation is undeniably the root cause of vetoes, but that explanation is incomplete as it begs the question of why Congress passes bills that the president will veto. It is a puzzle that Congress passes bills the president will veto because the president is in frequent communication with Congress about what he likes and dislikes in proposed bills, and is not shy about threatening to veto a bill he does not like. Given the ample communication between branches, it is surprising that differences cannot be resolved without a veto. Further, a vetoed bill represents lost time and effort by Congress. Vetoes represent bargaining failures, and the cause of these failures requires explanation.
There are currently two well-developed, contradictory theories explaining why presidential vetoes occur. A "blame game" theory holds that vetoes occur because Congress deliberately provokes them by passing bills that the president dislikes, knowing they will be vetoed. A "sequential veto bargaining" (SVB) theory holds that vetoes occur because Congress lacks complete information about the president's true position on legislation and sends the president bills unsure of whether they will be vetoed. This article tests these alternative explanations and finds that blame game politics accounts for far more vetoes than incomplete information.
Both theories begin with the dominant model of legislative executive relations in contemporary scholarship, a model of policy competition known as the "legislative agenda control" model (LAC). It is based on work by Romer and Rosenthal (1978), and has at its heart a two-person game in which the legislature makes offers to the executive, which the executive can accept or refuse, but cannot amend (McCarty 1997). Presidential vetoes occupy an anomalous role in the LAC model, because the model, in its simpler forms, predicts that no vetoes will occur. When the players have complete information about each other's preferences, Congress knows what the president will sign and veto, and does not send bills that will be vetoed. That is, Congress will send the president only bills that improve the president's position relative to existing policy. Congress, knowing what policies improve on existing policy for the president, sends the president no bills that he will veto. The president, knowing that Congress knows what he will sign and veto, understands the uselessness of threatening to veto bills that actually represent an improvement, and so he signs all bills presented to him.
Vetoes do not occur in the LAC model because of two key assumptions: first that there is complete information and second that the game provides no payoffs except from policy proximity. Both of these assumptions are in varying degrees open to question, and both of them could account for the inaccuracy of the pure LAC model. The two alternative theories of vetoes account for vetoes by relaxing different assumptions. Groseclose and McCarty (2001) and Gilmour (1995) allow Congress and the president to compete over blame and credit as well as over policy, and Congress intentionally passes bills that will be vetoed. Cameron (2000) adopts a model in which Congress lacks complete information regarding the president's preferences and accidentally passes bills the president will veto. Both of these modifications of the LAC theory can account theoretically for vetoes. For an excellent summary of this debate see Cameron and McCarty (2004). The goal of this paper is to test empirically which theory best accounts for the vetoes that actually occur.
Alternative Explanations of Vetoes
Groseclose and McCarty (2001) and Gilmour (1995) contend that vetoes occur in part because Congress passes bills designed to be vetoed and draw a clear distinction between the majority party in Congress and a president of the other party. These vetoes, called blame game vetoes by Groseclose and McCarty, are not accidental, but are part of a deliberate political strategy by Congress to cause political trouble for the president. Members of Congress are well aware in advance that certain bills will be vetoed but pass them anyway. Gilmour (1995, 131) argues that "Congress passes a bill despite the expectation of a presidential veto, because it expects the veto to help their party and hurt the president." Groseclose and McCarty (2001, 113) write that "Instead of working toward a compromise, Congress chooses instead to create a campaign issue for the next election." The practice of provoking a veto is "an important and attractive tactic [that] accentuates the differences between the parties and provides useful information about the parties' positions" (Groseclose and McCarty 2001, 120). If Congress presses forward with a bill despite threats of a veto, and then the president signs it, Congress wins by getting exactly the bill they preferred without accepting a compromise.
The explanation of vetoes developed by Groseclose and McCarty and differs from the LAC model of vetoes by introducing an audience that observes political bargaining. Politicians are interested in obtaining policy goals, but in bargaining they can seek to appeal to constituency groups and can benefit or be hurt by the information conveyed to the audience by vetoes. By passing a popular bill that the president will veto, members of the majority in Congress can claim credit for themselves and generate blame for the president. Blame game politics is a variant on "position taking," which, as Mayhew (1974) explains, does not depend for its success on actually enacting a law.
By contrast, Cameron (2000, 2009) asserts that vetoes result from incomplete information in Congress about the president's preferences. Congress wants to adopt the bill closest to its ideal point that the president is willing to sign. The problem for Congress is that, because of uncertainty about the president's true preferences, it is hard to identify exactly the best possible bill. Sometimes Congress will adopt a bill that the president will not sign. But when this happens, it is accidental, not by design. Cameron explains:
Congress begins with a notion of the range of possible presidential preferences and so bases the content of its initial bill on its expectations about what the president might accept. Congress will not pass a bill that every likely type of president would veto--such a bill would be pointless. Nor is it likely (except under special circumstances, detailed shortly) to pass a bill the president would surely sign whatever his true preferences. To do so would yield too much. Thus, Congress is likely to pass a tough bill but one with a reasonable chance of enactment. (2000, 111) In this theory, vetoes occur when Congress miscalculates, passes a bill that is too "tough," and the president will not sign.
In Cameron's model, there is no benefit to Congress in passing a bill that cannot become law because there is no payoff from public approval. His innovation is to introduce incomplete information about the president's preferences, and by this means vetoes become explicable in the LAC model. Congress passes bills that are later vetoed because Congress cannot know for sure what the president will sign. They pass bills that are close to the congressional ideal point, hoping the president will sign. If the first bill is vetoed, Congress passes another, closer to the president's position. If that is vetoed, they can try again and again until they find a bill that is mutually acceptable. Cameron calls this process SVB. A key element of Cameron's treatment of the veto process is the occurrence of "veto chains," which are characteristic of Congress and the president groping for agreement, and occasionally failing, necessitating the passage of another bill.
At the level of the individual bill, blame game and sequential bargaining are incompatible explanations. At the aggregate level, they are not mutually exclusive explanations. That is, some vetoes can be blame game and others can be due to SVB, but no one bill can be both. Cameron acknowledges the possibility of blame game vetoes but dismisses them as unimportant. He finds only five clear instances of blame game vetoes in the period of his study, 1945-1992 (Cameron 2000, 195).
To date, neither theory has been adequately tested. Cameron adduces evidence that is consistent with his models, but he never subjects the models and their predictions to a critical test, and he never tests his models against alternative explanations. Some of the predictions of his model are consistent with other explanations or simple commonsense. He writes, for example, that "The model neatly explains why veto threats are rare during unified government but relatively frequent during divided government" (Cameron 2000, 192). Obtaining empirical evidence consistent with a pedestrian prediction of this kind does not prove the model is correct. In the empirical part of their article, Groseclose and McCarty show that major vetoes are followed by a decline in the president's popularity--a finding that is consistent with and supports their theory. For both blame game and SVB theories, more rigorous testing that compares alternative explanations is required.
Testing Competing Theories
This article proposes and implements a critical test to determine the extent to which individual vetoes are caused by blame game politics or by uncertainty about the president's position and SVB. The key issue is the existence of uncertainty about the fate of the bill at the time it achieves final passage by Congress. If a veto results from SVB, we should expect to observe, at the time of the bill's final congressional passage, uncertainty about whether it will be signed...