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.

Broad bills ensure that general state policies remain current and fair, but legislators might avoid them for two reasons. First, they are technically complicated; if you wish to revise the state highway code, for example, you will need expert advice and probably a few studies. Second, they are politically difficult to pass; since they influence the entire state, you’ve got to work to bring a coalition of legislators on board with your proposal. By contrast, “district” bills are technically less complicated; the process of campaigning generally gives legislators all the information they need about some pressing local problem. District bills are also politically easier to pass; since they don’t have any impact outside of a small geographical area, other legislators have no reason to oppose most district bills.

By examining every bill introduced in 13 states in 1881, 1901, 1921, 1941, 1961, 1981, and 1997–that’s over 165,000 bills–Gamm and Kousser try to explain why some states produce so many more district bills than others. In Alabama, only 53% of bills had statewide impact; in Nebraska, 77% did. In general, Gamm and Kousser find that states pass more district bills when legislators have incentives to build up their reelection constituency or to make themselves stand out as an individual, but they pass more statewide bills when legislators have incentives to develop their influence and power within the state legislature. There’s more to it than that, though.

Hypotheses and Findings

Gamm and Kousser identify specific conditions under which legislators might prefer to introduce district bills or statewide bills.

Hypothesis 1 (one-party dominance): When the major parties are evenly balanced in a state, you’ll see more statewide bills; when one party dominates, you’ll see more district bills. This prediction is based on Key’s seminal 1949 book, Southern Politics in State and Nation. Key wrote that two-party competition produces “conditions favorable to government according to rule or general principle” (e.g. statewide bills); by contrast,  “in a loose, catch-as-catch-can [one-party] politics highly unstable coalitions must be held together by whatever means is available…. A loose factional system lacks the power to carry out sustained programs of action [e.g. statewide bills]” (pages 305, 308). In other words, two-party competition creates strong incentives for each party to form a statewide brand name by competing over statewide policies; in a one-party system, unstable factional coalitions within the supermajority party need to be held together with pork and other district bills. This hypothesis is confirmed in the analysis.

Hypothesis 2 (legislative salary): Legislators will introduce more district bills if they are paid more. California’s legislators earn six-figure salaries each year; Montana’s earn less than $10,000. Higher salaries create higher incentives to hold onto your job. Thus, higher salaries create incentives to deliver pork and other particularistic bills to voters in your district, even at the expense of pursuing the state’s general interest. This hypothesis is also confirmed in the analysis.

Hypothesis 3 (session length): Legislators will introduce more statewide bills if they are in session longer. California’s legislators are in session almost year round; in other states, legislators meet only one or two months per year. Because statewide bills are more complicated to write and pass than district bills, you will see fewer statewide bills in states with shorter sessions. As it happens, however, the data do not bear out this expectation; session length appears to have no relationship with the types of bills that pass.

Hypothesis 4 (turnover): Legislators can increase their statute within the legislature by building coalitions and passing statewide bills. As turnover decreases and legislators serve longer, they gain incentives to develop their stature within the legislature. As such, decreased turnover should lead to increased statewide bills. That’s the expectation, anyway. In the analysis, Gamm and Kousser find the opposite: Higher turnover leads to more focus on statewide bills. Perhaps their logic was off. Perhaps what’s really going on is this: Legislators who know they won’t stay long have less need to build up their reelection constituency (with district bills), so they can instead focus on statewide needs. I’m sure Kousser would rather not say this, but…score one for term limits?

Hypothesis 6 (district size): If districts are larger, legislators will focus more on statewide bills. Larger districts include more cities and towns, meaning that an individual district bill will reach a smaller portion of the district. In this situation, a more effective way to show the entire district that their legislator is attentive to their needs might be to pass statewide bills rather than district bills. This hypothesis is confirmed in the analysis.

Yes, I skipped hypothesis 5. I found it unpersuasive, and it ended up being statistically insignificant in the final analysis.

Punchline: If you want your legislators to pass general policies that benefit the state as a whole, pay them less, make districts bigger, and strive for partisan balance. If you want your legislators to pass pork and other district-focused bills, pay them more, make districts smaller, and promote one-party government.

Comment and Criticism

Gamm and Kousser are right to look to state legislatures to answer this question. Too often, folks think that they have to look at Congress if they want to study legislative process. The trouble with that is that there are no cross-sectional variations in the variables that we might expect to matter. In any given year, the majority party margin is the same for every member of the U.S. House; so is the salary, session length, average turnover, and so on. Sure, these things vary over time–but ALL of them vary from year to year, making it hard to tell which over-time variation is influencing the results.

Sometimes, folks who give up on Congress try instead to compare countries to one another. This approach is useful, but it introduces a host of other complications (variations in culture, constitutions, voting rules, etc.). The state legislatures are the perfect venue for this research question. The authors are able to study all the variables they care about, but all their observations are similar by virtue of their status as members of a common political system.

Previous Work

Because most of the states that adopted the initiative in this period were agrarian, western, and mostly white, historians have traditionally answered this question by claiming that the initiative was most likely to be adopted in “younger states with a strong history of populism and antimonopoly sentiments, more homogeneous [white] populations…, more radical (but weaker) political parties, and stronger interest groups” (pg 334). Smith and Fridkin contend that these conclusions have been based on deeply flawed logic and methods.¹

In particular, scholars have erred in asking why some states adopted the initiative and others did not; such an approach focuses on voter demand for direct democracy. But since the initiative is usually adopted via legislative referendum, in which legislators invite voters to choose whether to adopt the initiative, scholars should instead be asking why legislators in some states chose to offer the initiative to voters, while legislators elsewhere did not.²

Theory and Findings

Smith and Fridkin argue that legislators were most likely to offer the initiative when states experienced close partisan competition. When a minority party gains tenuous control of the legislature, or when a majority feels that it might become the minority soon, then legislators gain an incentive to weaken the legislature by empowering the median voter. This is their main causal argument, which is supported by an event history analysis.

Other conditions also play a role. In particular, legislatures in newer states are more like to offer the initiative than legislatures in older states. The authors attribute this difference to the weaker party organizations in younger states–organizations that might have less ability to rein in independent-minded legislators. Likewise, legislators were more likely to offer the initiative if there were more third-party legislators or if there was a strong interest group pressure in a particular state for instituting the initiative.

Contrary to previous work, the authors do not find any evidence of a contagion/diffusion effect. Nor do they find that initiative adoption was more common in racially and ethnically homogeneous states. Significantly, they find that dummies for the western and southern region are not statistically significant in their models; controlling for other variables apparently eliminates the regional bias in initiative adoption.

Critique

This paper’s methodological rigor makes it a welcome addition. As the authors point out, most previous work on this topic has been somewhat impressionistic and ad hoc. Event history analysis is the right way to study adoption of the initiative. Their shift from studying adoption to studying legislative support for the initiative is very smart.

That being said, a few things seem off to me. First, why did initiative adoption largely cease after 1918? Imperfect as previous work may be, its emphasis on western culture and populism seems to have an answer to this question. Smith and Fridkin don’t so much as speculate; they simply restrict their analysis to 1898-1918. But if some omitted variable caused initiative adoption to cease in 1918, is it possible that the same omitted variable, if measured, would change the results in Smith and Fridkin’s analysis?

Also, it seems that the role of governors may be downplayed. The authors start well (p 339) by listing several Democratic and Republican governors who pressured their legislators to pass an initiative reform; they even note that Republican Hiram Johnson of California made this a major goal. Yet inexplicably, the authors “control” for these considerations by including a dummy for whether the governor belonged to a third party. They provide no justification of this measure which, unsurprisingly, is not statistically significant–leading them to conclude (without basis) that gubernatorial action plays little role in adoption of the initiative. But if we believe that executives can pressure legislators by “going public” or through other means, then we should be looking more at the role governors may have played.

Third, they take no account (empirically) of variations in initiative institutions. Initiatives vary widely in (1) how hard they are to use and (2) how easily the legislature can ignore them (Bowler and Donovan 2004). Why did some states (California, Oregon) pass powerful initiatives, but other states (Utah, Idaho) passed weak ones? The authors acknowledge this variation, but do not seek to explain it. My speculation: The “party competition” variable that the authors discuss might be two different things. If a minority (perhaps allied with third parties) seizes control for a short while, or if a majority genuinely fears becoming a minority, then we might see an effort to create a truly powerful initiative. But if a majority sees its power slipping and wants to do something popular to build up support, then we might see a weak initiative created purely for its symbolic value. Thoughts?

Every study has weaknesses, and the flaws in this one are by no means fatal, nor are they necessarily more numerous or severe than what is typical for an article in a top journal. But I think they point to some avenues for further research.

It is one of the happy incidents of the federal system that a single courageous state may, if its citizens choose, serve as a laboratory; and try novel social and economic experiments without risk to the rest of the country.

These claims have inspired a literature on “diffusion,” mostly focused on the American states, that has asked whether there is evidence that the states do, in fact, learn from one another. For the most part, the existing literature has concluded that diffusion occurs regularly among the American states, although there is some disagreement as to the mechanism.¹

By contrast, Volden, Ting, and Carpenter argue that none of the existing work (including their own previous work on diffusion, apparently) has provided any evidence of policy diffusion. The methods and assumptions used in previous research cannot differentiate between innovation (isolated experimentation by myopic states) and diffusion (learning from experiments in other states).

Overview of the Model

To correct this problem, the authors present two formal models of experimentation. Both begin from the same basic setup: States have policy makers (legislators, bureaucrats, whatever) that can be placed along a unidimensional ideological line (i.e. we can classify the state as conservative or liberal). Within a particular issue area, there is a status quo policy and a proposed experimental policy. Each policy has two characteristics. First, it can be placed along the ideological line, and this “spatial” characteristic is common knowledge–that is, everybody agrees as to which policy’s goals are more liberal or conservative. Second, each policy has a “valence”–that is, each policy might be more or less effective at reaching its stated goals.

Although each proposal’s “spatial” characteristics are assumed to be common knowledge, “valence” is known only for the status quo–the experimental proposal’s valence is unknown. Thus, policy makers have a choice: They can stick with the status quo (with known valence) or they can switch to the experimental policy (with unknown valence). If they choose to experiment, then in “period two” (e.g. the next legislative session), when the experimental policy’s valence is known, they can choose to stay with the new policy or revert to the old one.

After setting up this basic model, the authors derive two models from it. The first is a decision-theoretic model that assumes each state exists in isolation; states may innovate similar policies, but there cannot (by definition) be diffusion. The second is a game-theoretic model that assumes each state can learn from policy experiments in other states; either innovation or diffusion can occur.

The decision-theoretic model

In this model, states do not have the option of learning from other states’ experiments.

In an extremely liberal state, the policy makers will choose the most liberal policy proposal, regardless of expectations about valence. (Likewise for extremely conservative states). But in moderate states, policy makers will balance valence against ideology. A moderate conservative would prefer an efficient but liberal policy over an inefficient but conservative policy; a moderate liberal would prefer an efficient but conservative policy over an inefficient but liberal policy. These are the conditions under which innovation occurs.

Thus, we would expect to see only moderate states experimenting. If they learn that the policy is inefficient, then they would revert to the previous policy–the one that is ideological preferable.

The game-theoretic model

In this model, states do have the option of learning from other states’ experiments.

If states can learn about a proposed policy’s valence by observing policy experiments in other states, then the incentive to experiment drops. Experimentation is risky; if you can learn from others’ mistakes rather than having to make those mistakes yourself, then why experiment at all? As such, those policy makers willing to experiment will fit into a narrower ideological range than those willing to experiment in the decision-theoretic model.

Implications for the Literature

Previous work on “diffusion” has not appreciated the differences between these two models. All of the evidence for policy diffusion presented in previous work can be explained in terms of the decision-theoretic model. In order to conclude that diffusion actually occurs, we must find evidence of behaviors that are predicted by the game-theoretic model and NOT by the decision-theoretic model.

The literature has presented five different causal mechanisms to explain “diffusion,” but all five mechanisms can be explained with the decision-theoretic model:

• Walker says some states are inherently more disposed to innovate than others.
• Gray says diffusion happens when states face similar policy problems.
• Others say that diffusion happens when neighboring states are ideologically similar.
• Still others say that diffusion happens when any states (neighboring or not) are ideologically similar.
• A final argument is that diffusion happens when policy advocates take their arguments to multiple states.

As the authors put it: “Much of the empirical work to date has not adequately distinguished [game-theoretic] learning-based policy diffusion from [decision-theoretic] myopic individual adoptions.”

The authors conclude by listing specific empirical implications of their model that future research should evaluate in order to determine whether or not diffusion actually occurs.

Critique

Much of the work on diffusion was published in the 1960s and 1970s. I suspect that modern political scientists have hesitated to take up the question again because they thought it was settled. This article demonstrates persuasively that the question is far from settled. This is the article’s most important contribution–to point out that we do not yet have any evidence that states do (or do not) act as “laboratories of democracy” that learn from one another. This question is normatively important; if states do not learn from one another, then we lose an argument for devolution.

However, I was less than satisfied with the authors’ empirical suggestions. They conclude by pointing out several ways that we can use their theory to empirically determine whether states follow the decision- or game-theoretic model. Unfortunately, these empirical implications are extremely nuanced and may be difficult (or nearly impossible) to apply in practice.

In the absence of direct democracy, political parties might not have much reason to promote moderate policies. Republican legislators would generally prefer policies to the right of the median, while Democratic legislators would prefer policies to the left. Whichever party has the legislative majority has a variety of tools at its disposal to help it push policy away from the median and towards its own ideal point. These tools include control of the legislative agenda, control of committee chairmanships, and so on. In addition, the need to appease major campaign donors and partisan activists increases the incentive to use these tools. As a result, it is not the legislative majority that governs; it is the majority of the majority party that governs. When legislative control shifts from Republicans to Democrats, you might see a large shift in policy outcomes–even if only a couple of legislative seats changed hands.

But this is exactly the sort of thing that the Progressive movement sought to end. A purpose of direct democracy was to force legislators to produce policy outcomes closer to what the median voter would want–regardless of which party has the legislative majority. The purpose of Phillips’s article is to ask whether they succeeded.

There are two main mechanisms by which direct democracy can be a median-enhancing institution (from Gerber 1996). The first mechanism is direct; voters can impose specific legislation on the legislature, as happened with California’s famous Prop 13. The second is indirect; even if voters never use the initiative, its presence acts as a deterrent against extreme behaviors by the legislature.

Phillips finds evidence that direct democracy does matter. Party government is weaker in states with the citizen initiative than in states without. Here’s how he did it:

• Dependent variable: The state tax burden.¹
• Main independent variables: First, a few dummies to measure partisan control.² Second, a dummy indicating whether a state has direct democracy. Third, interactions between these two sets of variables.

As expected, he finds that the partisan dummies have strong relationships with the state tax burden. Democratic governments tax the most; Republicans tax the least; divided governments are in the middle. But here’s the rub: these relationships disappear in states with direct democracy.

The existing literature on the relationship between partisan control and size of government has had mixed results. Phillips contends that these mixed results can be explained, at least in part, by looking at direct democracy.

I have only one complaint with this article. I’m not convinced that a simple dummy variable can adequately measure direct democracy. From state to state, there are huge variations in how easy it is to use the initiative process, leading to huge differences in how frequently the process is used. This article could be more persuasive if it discussed this problem, which Phillips does not even mention. In fact, this problem is severe enough that, to me at least, it undermines the entire argument.