CHAPTER 1
The Electoral Bases
coauthored with Jonathan Menitove
Even in a system of just one elected institution, it can be tricky to poise a government on an electorate. Nineteenth-century Britain is the textbook case. Beyond questions of suffrage expansion, apportioning the House of Commons geographically proved to be a continuing, often tense project. Skew was the problem.1 According to one account, âBritain has never in modern times been closer to revolution than in the autumn of 1831ââa run-up to the Great Reform Act of 1832, which targeted the ârotten boroughsâ of the House of Commons and gave parliamentary voice to the new industrial cities of the North in answer to their processions, petitions, and riots.2 Skew arose in another major constituency overhaul in the 1880s testing whether âthe varying opinion of the country was to be fairly represented in the House of Commons.â3 A goal of that new reform was âa general correspondence between the support in [popular] votes and the representation of the two great parties.â4
In American history, again bypassing questions of suffrage, the national government during the decades after the Civil War may offer the best case of a representational question mark. Both skew and dissonance figured in a mix highlighting the Senate.5 Starting in the 1860s, Republicans in control of the government created a series of new states small in population and friendly to their party. Republicans in control of most of those statesâ legislatures (as well as others in the North) followed the party line in choosing senators.6 A Republican edge in Senate membership came to jar against a better Democratic showing in the membership of the House and the presidential popular vote.7 A near miracle seemed to be required for the Democrats to win the Senate.8 Tension built up. Keynoting the Progressive era, the muckraker David Graham Phillips in his celebrated essay of 1906, âThe Treason of the Senate,â assailed that body as âthe eager, resourceful, indefatigable agent of interests as hostile to the American people as any invading army could be, and vastly more dangerous; interests that manipulate the prosperity produced by all, so that it heaps up riches for the few.â9 This harsh language is worth reading twice. The Senate was running into legitimacy trouble. Reform came in the adoption of the Seventeenth Amendment in 1911â13 requiring direct election of senators.10 According to one analysis, this shift to direct Senate elections erased that chamberâs outlier Republican edge in the government once the electorates of the northern states proved to be less lockstep Republican in their choice of senators than their legislatures had been.11
That was a century ago. What is the record more recently? More specifically, how has the American mix of national elective institutionsâthe presidency, the Senate, and the Houseâconnected to the nationâs electorate during the decades since World War II?
To pursue this question here, I make heavy use of a particular statistic, the division of the major-party popular vote cast for presidentâthe Bush versus Gore vote in 2000, the Nixon versus Humphrey vote in 1968, the Eisenhower versus Stevenson vote in 1952, and so on. Without much doubt, the popular vote for president is the most important formal expression of political opinion in American public life. A presidential election seems to tell where the country stands. The stakes are high, turnout and interest peak, and the voting behavior of states, regions, and demographic groups is analyzed and accorded significance. Noted are the shape of the vote (Kansas voted one way, Rhode Island the other) as well as the thrust of the vote (the Republicans had a good year). All this has been true since the 1830s. When American elections are compared with those in parliamentary systems, the presidential vote is ordinarily used. In addition, the popular vote for president seems to possess a moral authority beyond whatever may happen in the Electoral College. We often hear that Al Gore âreally wonâ the election of 2000.
Granted, as a diagnostic of âwhere the country stands,â the major-party vote-share statistic does pose difficulties. There are at least six. First, the measure ignores an often sizable third-party vote for president as in 1968, 1980, and 1992. Second, tens of millions of people do not vote at all in American elections. Third, there is the Electoral College. We cannot know what kinds of appeals and mobilizing drives the parties would employ in any imaginable counterfactual systemâsay, to cite the most obvious alternative, one of direct national voting for president. The voting patterns, as well as the partiesâ win-loss outcomes, might differ in an alternative system. Fourth, again regarding the Electoral College, a particular distortion arises: owing to strategies induced by that system, voter turnout tends to slump a bit in nonbattleground states relative to battleground states.12 Fifth, a pinch of salt has to accompany summary election returns in a decentralized system like the American one where states and locales have their own customs and procedures. Exact correctness is an unwise expectation. In 2000 Florida had its hanging chads and butterfly ballot. In 1960 Illinois had a clouded count,13 and Alabama in that same year staged a vote for individual electors, not directly for presidential candidates, in a fashion that has spurred discussion then and since about how the stateâs votes should have counted toward a national popular vote total that year, and indeed, in light of the Alabama complexity, whether Kennedy really did win a national vote plurality over Nixon. To award the national vote edge to Kennedy âinvolves a moderate amount of license,â the political scientist V. O. Key, Jr., concluded in 1961.14 Sixth, to raise a different kind of point, the presidential popular vote can offer obvious idiosyncrasies associated with candidates. Reaganâs landslide 59â41 win over Mondale in 1984 and Nixonâs 62â38 win over McGovern in 1972 were, in this sense, distortions of where the country really stood. Also, the Kennedy vote in 1960 had a signature Roman Catholic flavor, the Carter vote in 1976 a southern flavor, the Reagan vote in 1980 a possibly southwestern flavor, and so on.
There is another possible kind of difficulty. Whether, in a normative sense, the presidential popular vote should be seen as indexing âwhere the country standsâ is a separate matter. For one thing, American âpolicy mandates,â for all the claims about them, do not seem to enjoy very sure empirical roots.15 Also, the style of contestation in American presidential elections has aspects of both plebiscitarianism and vacuousness that have left some observers uneasy.16 The less flashy link of legislative representation has its own kind of integrity.17 But this is a road we do not need to go down. I am engaging here in positive analysis, not normative assessment.
Notwithstanding all its impurities, distortions, and possible dubiousness, the presidential vote is out there as an exceptionally prominent artifact. It can be used as an analytic wedge.18 Conveniently, the presidential major-party vote-share statistic is available for the country at large but also in calculations for individual House districts (at least since 1952), individual Senate districts (that is, states), and the District of Columbia (starting in 1964 when voters there were enfranchised for presidential contests). For purposes here, I ask a âgeneral correspondenceâ question of a special sort: How does the presidential vote-share statistic play through the universes of House districts, Senate districts, and the Electoral College as compared with its performance in the country taken as a whole? More specifically, as will be discussed, I home in on the presidential vote-share statistic in the median district for the House, the Senate, and the Electoral College once the units of each of these universes are ranked for any election according to that value. This analytic course is a way around many of the impurities or distortions outlined above. The nonvoters and the counterfactual electoral systems lose relevance: in question is how American opinion formally registers itself in the system that actually exists. The problem of presidential landslides recedes: in question is the relation between the presidential vote share in the country as a whole and the presidential vote share strained through the various systems of subunits, regardless of whether the national results were close or one-sided. The third-party problem remains, but there is no way to manage it well and it does not seem to vitiate the analysis.
An intuition lurks behind this plan. It is that the political texture of constituencies is fundamentally important. Constituencies express their tastes by, for one thing, electing officials of one party or another, and those officials once in office tend to behave as partisans, but that is not the end of the story of representation. Party still has to struggle with constituency.19 In Congress, a Maine Republican is not likely to compile the same kind of voting record as a Texas Republican. We are not surprised when Republican Senator Susan Collins of Maine disagrees with Republican Senator James Inhofe of Oklahoma. In a pattern that transcends party, politicians tend to tailor themselves toâor at least to matchâtheir constituencies. This can be for reasons of electoral calculation, principled deference to home opinion, or, for that matter, accurate sampling of constituencies in the selection of candidates. No doubt all these factors have a role.
Enter the presidential vote-share statistic. As an indicator of the political texture of a constituencyâthe whole nation, a state, or a congressional districtâit seems to be very good. At the congressional level, political scientists have increasingly used the statistic as such an indicator. It seems to capture a blend of partisanship and ideology. For any election, states and districts can be scored and ranked according to their presidential performance. For any constituency, the statistic can supply information to an outside observer trying to predict how an office-holder will behave, and it can supply information to an office-holder trying to figure out how to behave. To be sure, presidential landslides can pitch the statistic up and down, which makes for a kind of distortion in one sense yet in another sense, it is not: Even in 1984, âReagan carried my stateâ could supply nontrivial political cues to, say, a senator from Massachusetts.20
For an exhibit of the independent illumination that can be supplied by the presidential vote-share statisticâthat is, independent of the party memberships of office-holdersâsee figure 1.1, a scatterplot designed by Paul Frymer.21 Plotted here are the roll-call records of House members elected in 1988 according to the share of the presidential vote won by George H. W. Bush in their districts that year. Americans for Democratic Action (ADA) scores are the roll-call measure. For the mass of incumbents who ran again in 1988 and won, the roll-call scores here are from the preceding Congress of 1987â88. For newcomers they are from the succeeding Congress of 1989â90âa second-best solution in Frymerâs design. In principle, Frymerâs question was: Relying on information at hand, do voters in the same district on the same day tend to select the same ideological flavor across offices?22 Democratic members appear in figure 1.1 as filled-in triangles, Republicans as hollowed-out rectangles. A summary of the data appears below the scatterplot. The first lesson from this display is that, yes, Democrats certainly do have higher ADA scores than Republicans. But a second lesson is that the parties vary within as well as between. In general, within each party, the higher the Bush vote the lower the ADA score. These figures are for 1988. The patterns of these within-party relationships would no doubt vary in a fuller study extending across the last half century, but the relationships would always be there to some degree. On offer, it seems likely,...