For years, political scientists have explained this seeming paradox by pointing out that members of Congress can win reelection by running against Congress. A representative can urge his voters to send him back time after time so that he can keep working to fix the broken system. As Fenno wrote in Home Style, “It is easy for each Congressman to explain to his own supporters why he cannot be blamed for the performance of the collectivity . . . because the internal diversity and decentralization of the institution provide such a wide variety of collegial villains to flay before one’s supporters at home” (1978, 167).²

In his textbook on Congressional elections, Gary Jacobson sums up the dominant view among political scientists: “Members are not held individually responsible for their collective performance in governing.” (2004, 227).

David Jones has a recent article in AJPS that challenges this long held view. Jones looks back 60 years to a report commissioned in 1950 by the American Political Science Association Toward a More Responsive Two-Party System. That report urged “greater party cohesion in Congress,” suggesting that the presence of liberal Republicans and conservative Democrats makes it more difficult for individual voters to hold their representative accountable for Congress’s collective activities.

Jones argues that the APSA report was correct: If the two parties become more distinct (i.e. polarized), then it should be easier for voters to blame members of the majority party for Congress’s collectively bad (or good) performance. And, as it happens, there’s been quite a bit of research in recent years showing that Congress has, in fact, become more polarized.

If Jones is right, then we’re in a new era. It may have been true 20, 30, or 40 years ago that members of Congress could evade accountability for Congress’s overall activities, but rising polarization has enabled voters to punish or reward Representatives for Congress’s collective performance.

Evidence that Overall Congressional Approval Matters

To test this possibility, Jones compiled each incumbent Representative’s electoral margin going back decades, producing thousands of data points. He then regressed those vote margins on a variety of independent variables. Among others, he regressed vote margins on Congress’s overall approval ratings. More importantly, he also interacted those approval ratings with measures of polarization (party unity).

Take a look at Figure 1 from Jones’s article (below). Along the X axis, Jones shows each year back through 1976. Along the Y axis, he shows that estimated effect of overall Congressional approval on individual incumbents’ vote margins for that year.³ By the end of the series, the estimated effect of overall approval rises above 0.50 (for members of the majority party). In other words, a one percentage point drop in Congressional approval (perhaps from 40 to 39) leads us to expect a 0.50+ drop in each incumbent’s vote margin. This is a powerful effect, subject to a powerful interaction.

Figure 1, Jones 2010

It’s worth noting that there is not similar interaction for members of the minority. Back in the days of low polarization, minority members could win reelection by running against Congress, just like members of the majority. Rising polarization has not prevented minority party members from continuing to run against Congress–and why should it? Minority party members can continue to win by running against Congress, citing all the majority’s misdeeds. Nothing has changed for the minority.

Comment and Criticism

This is an interesting and worthwhile article. It leaves me wondering, though, why 90+% of incumbents continue to win reelection. I began this review by pointing out a seeming paradox from 2008–not from 1976. Even in the most recent Congressional elections, 36% of voters said that most members did not deserve reelection, yet 94% of members won reelection. More generally, we continue to observe Congress (overall) receiving markedly low approval while individual members receive very high approval from their constituents.

If Jones is correct, then we ought not to observe this pattern so strongly anymore, yet we do. I’m not sure how to respond to Jones’s analysis given this continuing disconnect between overall and individual Congressional approval. Perhaps Jones has a serious problem in his statistical analysis that I’m not seeing resulting in an inflated estimate of the interaction. Or perhaps Congressional elections are sufficiently different from Congressional approval that this paradox can persist in approval data even as it evaporates in election results. I’m at a loss to resolve this puzzle.

The question: How do candidates decide what to emphasize in their campaign communications? When do they go negative? When do they stick to the issues? When do they emphasize their experience and community ties? And given how many thousands of campaigns are run around this country each election cycle, how can we possible study all this?

There has been some previous work on this question, but most of it has looked at television ads or media coverage. Both sources have flaws. Television ads and media coverage are more common in the most competitive races, since safe incumbents don’t spend money on ads. They are also more common in Senate races than House races. How, then, can we use television ads to see how rhetoric is different in competitive races or in House races than in other races?

In a recent article, Druckman et al. avoid these problems by looking instead at rhetoric in Congressional campaign websites in 2002, 2004, and 2006. Although not all candidates had websites in 2002, by 2004 and 2006 just about every major-party Congressional candidate had a website. And what do we learn?

Findings

• Challengers are far more likely than incumbents to use negative attacks (whether personal attacks or issue contrasts) than incumbents.
• Challengers are more likely than incumbents to use “Web 2.0″ technologies that enable visitors to create content (e.g. via commenting on the site).
• Challengers are more likely than incumbents to stress their party affiliation and to bring up issues that their party “owns,” particularly if the incumbent is not a member of the district’s majority party.²
• Incumbents are more likely than challengers to emphasize their experience, their long history in the district, and the specific benefits (pork) they have provided to the district. This is especially true in hotly contested districts; elsewhere, incumbents are unlikely to put much effort at all into their websites.

Theory

Druckman and his colleagues tell a compelling story to explain these findings. Briefly, and with considerable re-interpretation by me:

• Going negative and enabling interactive Web 2.0 technologies are risky. Since incumbents enjoy significant electoral advantages³, they feel no need to take on these risks. But challengers seeking to overcome these advantages may find these risks worth taking. Thus, challengers are more likely to pepper their sites with negative comments and also to provide forums, wikis, or commenting interfaces.
• Partisanship is often as valuable a cue to voters as incumbency. Thus, challengers make a partisan case against incumbents who belong to the district’s minority party.
• Incumbents want to do what they can to strengthen their incumbency advantage. Thus, an emphasis on talking about things that only incumbents have (experience, credit claiming opportunities, etc).

Importance

The authors’ biggest contribution is their thorough, thoughtful, and insightful use of campaign websites. This data source allows for a near-universal (especially after 2002), unfiltered look at what candidates want voters to hear. And this method does, indeed, yield different results than we would find if the authors had relied on more traditional data sources, such as TV ads or media coverage. When the authors restrict their analysis to those races that had significant ad buys or media coverage, many of their important results disappear into statistical oblivion. The authors have identified a cheap, easy way to capture a fuller sample of current campaign messages.

Quibbles and parting jabs

I like this article, but the authors need to be careful not to oversell their contribution.⁴ Reliance on websites as a true measure of what messages campaigns are actually pushing may not be as much a panacea as claimed.

First: Who reads campaign websites? The authors use a survey of campaign web developers to show who, in the developers’ minds, reads the websites. Even the developers concede that the main target audience–swing voters–is the least likely of all to actually visit the site. But remember that the web developers probably dramatically overestimate the importance of their product. How else would they sell their services? Consider the websites of two prominent members of Congress: Eric Cantor (Republican Whip) and Jim Clyburn (Democratic Whip).⁵ Alexa.com is a free service that tracks how many people view websites.⁶ Alexa reports that so few people visit Cantor’s (report) and Clyburn’s (report) sites that it can’t even provide an estimate of their reach.⁷ So, I ask: Do we have evidence that anybody actually reads these websites?

Druckman et al. would probably counter that it doesn’t matter whether anybody visits the site. What matters is that the site summarizes the campaign messages being used by the candidate generally, both online, in ads, and in appearances. Well…

Second: Do candidates actually push the same messages in the real world as in their websites? You can fit many, many campaign messages into a website. You can only fit a small handful into a single ad, appearance, or mailer. Perhaps candidates intentionally place their most provocative messages on their websites to avoid having to say them to their opponent’s face during a debate. We saw this in the 2008 presidential race. Both presidential campaigns released their worst ads online only, knowing that the media would see them and relay the attack’s message to conflict-hungry viewers. This was a hands-off way for candidates to get negative messages out into the blogosphere without having to push the messages personally. If that happens in Congressional races too, then this study is flawed–perhaps deeply.

But again. I like the article. It makes a fabulous contribution. The authors do as most of us do, however, by overselling their point.

Gordon and his colleagues take this widely accepted argument a step further. Their concern lies not with the candidates’ thinking, but with the voters’. If, as previous research indicates, high-quality challengers act strategically when deciding whether to run against a Congressional incumbent, then this decision ought to convey important information to voters.

As long as voters believe that the challenger actually bears the sort of costs discussed in the literature,¹ then the challenger’s willingness to incur these costs signals to voters that the incumbent may be weak.²

Democracy is supposed to provide voters with an opportunity to hold elected officials accountable for their performance in office. With so many elected officials to monitor, however, voters would have a difficult time fulfilling this task without assistance. We have long known that challenger behavior assists voters in this judgment by determining whether voters even have a real choice to make on election day. Gordon and his colleages add to this argument by explaining that challengers also provide information to voters about the incumbent’s performance merely by deciding whether to run.

This study builds on two fields. First, it seeks to develop our understanding of retrospective voting (Fiorina 1981; Kramer 1971). Second, it contributes to the literature on the incumbency advantage and challenger quality (Cox and Katz 1996; Jacobson and Kernell 1983).

The authors present no empirical evidence to support their claims, only a formal model. Given the large body of literature already available about strategic entry, this approach is not bothersome–yet. Personally, however, I hope to see empirical evidence for these arguments in future research. The claims make good sense, but sensible, intuitive claims often lead to unexpected, counterintuitive findings when tested on real-world data.

What is less clear is the source of this incumbency advantage. Previous research has suggested three answers. First, perhaps the perks of office, particularly the franking privilege and opportunities for casework, allow incumbents to endear themselves to their constitutions (Mayhew 1974; Fiorina 1977). Second, perhaps gerrymanding and weakened party loyalty are to blame (Erikson 1972). Or third, perhaps incumbents are so formidable that the best potential challengers simply refuse to invest their time running (Cox and Katz 1996; 2000).²

The trouble with these theories is that they have been developed and tested in the post-World War II context. Carson et al., by contrast, look to historical (1870-1900) House elections for further illumination. In so doing, they cast doubt on the first two theories, both of which claim that factors present today (but absent in the nineteenth century) cause the incumbency advantage. As it turns out, there was an incumbency advantage even then, even before members of Congress had large staffs (for casework), the frank, high salaries, and constant redistricting.

The authors then turn to the third theory. Cox and Katz supposed that their “scare-off” theory explained why the incumbency advantage had strengthened since 1964, the year of the Supreme Court’s “one man, one vote” ruling that led to perpetual redistricting. However, Carson et al. find 19th-century evidence of the same “scare-off” effect that Cox and Katz observed post-1964. Thus, although Cox and Katz may have identified a true effect, their “scare-off” effect cannot explain the recent rise in the incumbency advantage since the effect was present a century ago, long before “one man, one vote” took effect.

Candidate Quality in the Nineteenth Century

Much of the authors’ theory merely applies previous arguments from the literature. However, they do find themselves needing to defend one particular point. Late 19th century elections were much more party-centered than today’s candidate-centered affairs. With party leaders selecting Congressional candidates (rather than candidates openly competing in primaries) and with voters using party-printed tickets, should we even expect candidate quality to matter in this era?

Jacobson (1989, 787) argues that we should not. The reason lies in the competing incentives of parties and candidates. As for candidates, it is now accepted that Congressional election results depend crucially on challenger quality. When national partisan tides (and district-level concerns) favor an incumbent’s party, she is unlikely to attract a well-funded, experienced challenger (Jacobson and Kernell 1983). After all, potential challengers must bear costs to run–such as placing their political career at risk–so the best challengers will run only when success seems likely. The trouble is, however, that parties have the opposite incentives. When the national tide favors the Democrats, the Republican party wants to recruit the best possible challengers in an effort to forestall the pro-Democratic tide. In our candidate-centered era the party rarely gets its way, giving rise to the literature on challenger entry. But in the party-centered era that Carson et al. study, the criticism goes, we should not expect challengers to behave strategically; party leaders had pork, patronage, and other enticements that they could use to persuade high-quality challengers to run even against the partisan tide.

Carson et al. dismiss this concern, noting that self-interested politicians could refuse to run despite these incentives. Therefore, we should expect the same variations in challenger quality that we observe today–making the “scare-off” effect a reasonable thing to expect in the late nineteenth century. Moreover, candidate quality did matter, the authors claim; even though voters used party-printed ballots, the authors present anecdotal evidence to suggest that it was the Congressional candidate’s quality–not the presidential candidate’s–that influenced voters the most.

Empirical Evidence

The authors seek to predict the Democratic candidate’s vote share as a function of the lagged vote, the incumbent’s party (also lagged), a dummy for whether an incumbent was present (the primary coefficient of interest), and the Democratic quality advantage (a dummy indicating which party had the more experienced candidate; also lagged). They find that both incumbency and quality matter.

They then predict the Democratic quality advantage using essentially the same variables, and find that an incumbent’s presence lowers the likelihood of a quality challenger.

The Main Point

In case you missed it, here it is: Previous researchers have advanced three main arguments in an attempt to explain the origins of the incumbency advantage. But although their arguments may be true to an extent, they cannot be the full story. The first two theories depend on conditions that were not present in the nineteenth century, yet the incumbency advantage was–therefore, they cannot be the complete explanation. And the third theory claims to explain the increasing power of the incumbency advantage since the 1960s, yet it was present in the nineteenth century–so something else must account for the increasing advantage.

The main point: We still don’t know everything about what causes the incumbency advantage, so we need more research into it.