Party, policy, democracy and candidate choice in U.S. elections

Party, policy, democracy and candidate choice in U.S. elections

Bright Line Watch

John M. Carey (Dartmouth College)
Katherine P. Clayton (Dartmouth College)
Gretchen Helmke (University of Rochester)
Brendan Nyhan (University of Michigan)
Mitchell Sanders (Meliora Research)
Susan C. Stokes (University of Chicago) 


How committed is the American public to democracy? Are there any demo­c­ra­t­ic prin­ci­ples that, if violated by politi­cians, would generate resis­tance from the public? Are citizens of all political stripes equally willing to punish can­di­dates for such vio­la­tions? Building on cutting-edge research by Graham and Svolik (2018), Bright Line Watch conducted an exper­i­ment that asked people to choose between pairs of hypo­thet­i­cal can­di­dates. We use a conjoint design, which allows us to estimate the relative impor­tance of candidate traits, par­ti­san­ship, policy positions, and support for demo­c­ra­t­ic prin­ci­ples in voters’ decisions. Our exper­i­ment yields the following findings: 

  • Partisanship outweighs all else for both Democrats and Republicans. Both groups are approximately 19 percentage points more likely to select a candidate from their own party than one from the other party — an effect that exceeds those observed for candidate policy positions and support or opposition to democratic principles. The parties also continue to divide over the issues of tax policy and racial discrimination.

  • Democrats, Republicans, and independents all punish candidates who violate democratic principles related to political control over investigations, judicial independence, and cross-party compromise. These effects are consistently negative across partisan groups and range from four to 13 percentage points.

  • Americans diverge most dramatically by party on the democratic principle of equal voting rights and access. Democrats are less likely to back candidates who endorse legislation requiring voters to show ID at the polls, whereas support for these candidates increases by eight percentage points among independents and 17 percentage points among Republicans.

Encouragingly, these findings indicate that several key demo­c­ra­t­ic prin­ci­ples maintain consensus support. However, the pun­ish­ment for violating them is often modest (cf. Graham and Svolik 2018). Moreover, the, the high levels of par­ti­san­ship we document create a context in which such prin­ci­ples can be called into question and politi­cized. As we show, the deep partisan divide over voter ID laws is now reflected in the choices Americans make when eval­u­at­ing can­di­dates. The polarized response to these policies illus­trates how partisans can become deeply split over which demo­c­ra­t­ic pri­or­i­ties are worth protecting. 


Proclaiming support for democracy is easy, but will voters follow through on these com­mit­ments and punish can­di­dates who violate important demo­c­ra­t­ic prin­ci­ples? In recent years, prominent studies have argued that public com­mit­ment to demo­c­ra­t­ic values in the U.S. has declined (Foa and Mounk 2016, 2017, Pew Research Center 2018, Wike and Fetterolf 2018), though these findings are contested (Alexander and Welzel 2017, Norris 2017, Voeten 2017). American voters are also becoming more partisan (Pew Research Center 2017), which is strength­en­ing feelings of in-group loyalty and out-group rivalry (Fishkin and Pozen 2018, Mason 2018). Taken together, this increas­ing partisan antag­o­nism could poten­tial­ly threaten the stability of the U.S. political system by eroding master norms of mutual tol­er­a­tion and for­bear­ance (Levitsky and Ziblatt 2018).[1]

A new line of research seeks to measure the con­se­quences of these changes  by examining how voters make trade-offs between their com­mit­ments to their partisan and policy pref­er­ences, and to the values of democracy itself (Graham and Svolik 2018; also see Svolik 2017). In a survey exper­i­ment, these scholars pit hypo­thet­i­cal can­di­dates for state leg­is­la­ture with exper­i­men­tal­ly manip­u­lat­ed char­ac­ter­is­tics against each other and ask respon­dents to choose between them. They find that partisan and policy con­sid­er­a­tions have much stronger effects on vote choice than “unde­mo­c­ra­t­ic” platforms related to electoral fairness or checks and balances. 

Building on this research, Bright Line Watch conducted our own conjoint survey exper­i­ment assessing the relative impor­tance of par­ti­san­ship, policy pref­er­ences, and a subset of demo­c­ra­t­ic values taken from Bright Line Watch’s original 27 prin­ci­ples among voters just before the 2018 midterm elections. Our survey asked respon­dents to choose between a series of paired hypo­thet­i­cal can­di­dates “in an upcoming election.” Each “candidate” was defined by a profile con­sist­ing of a set of personal attrib­ut­es and stances. For instance, a pair of can­di­dates might include a male Democrat who favored redis­trib­u­tive economic measures and thought that officials should not be con­strained by court decisions they saw as partisan, and a female Republican with similar policy positions but who expressed more deference to courts. 

Respondents made a series of choices among pairs of can­di­dates whose attrib­ut­es were randomly assigned. This design allows us to estimate the effect of each indi­vid­ual component in the can­di­dates’ profiles — for example, their race, their party, or their stance on respect­ing court decisions — on the prob­a­bil­i­ty that the respon­dent would support him or her. (For more on conjoint exper­i­ments, see Hainmueller, Hopkins, and Yamamoto 2013.) The design also allows us to see how the effect of candidate attrib­ut­es on vote choice differs between different types of respon­dents. For example, we can compare how much Democratic and Republican respon­dents care about can­di­dates’ views on respect­ing court decisions.

The candidate choice experiment

The study was conducted from October 24–31, 2018 among 962 online par­tic­i­pants. [2] The sample was matched and weighted by the survey company YouGov to approx­i­mate a rep­re­sen­ta­tive sample of the U.S. pop­u­la­tion. 35% of our respon­dents iden­ti­fied as Republicans or inde­pen­dents who lean Republican, 43% as Democrats or inde­pen­dents who lean Democratic, and 17% as inde­pen­dents who lean toward neither party. [3]  Respondents were presented with ten pairwise choices between can­di­dates in a hypo­thet­i­cal upcoming election. Each candidate was described using eight char­ac­ter­is­tics: name, par­ti­san­ship, positions on policies toward taxation and racial dis­crim­i­na­tion, and four positions reflect­ing demo­c­ra­t­ic values. Figure 1 provides an illus­tra­tive example of what respon­dents saw:

Figure 1: Sample conjoint table

The char­ac­ter­is­tics of the can­di­dates were randomly generated from the following sets of alter­na­tives:[4]

Name (gender and race/ethnicity): Each candidate was assigned a name from a list of 123 names designed to signal both gender (man or woman) and race/ethnicity (either white, black, or Hispanic). [5] [6]

Partisanship: Democrat or Republican.

Policy positions: The surveys included two attrib­ut­es reflect­ing salient party dif­fer­ences over policy — one on racial dis­crim­i­na­tion, the other on taxation (see Bartels 2018).

Racial dis­crim­i­na­tion

  • Believes the gov­ern­ment should do more to prevent dis­crim­i­na­tion against racial minorities.
  • Believes dis­crim­i­na­tion against racial minori­ties is less of a problem now than in the past. 


  • Wants to raise taxes on the wealthy.
  • Wants to lower taxes on everyone, including the wealthy.

Democratic values: The surveys included four attrib­ut­es related to core demo­c­ra­t­ic values. 

Voting rights/access

  • Opposes new leg­is­la­tion to require voters to show state-issued ID at the polls.
  • Supports new leg­is­la­tion to require voters to show state-issued ID at the polls.

Investigations/rule of law

  • Said law enforce­ment inves­ti­ga­tions of politi­cians and their asso­ciates should be free of partisan influence.
  • Said elected officials should supervise law enforce­ment inves­ti­ga­tions of politi­cians and their associates.

Judicial deference

  • Said elected officials must obey the courts even when they think that the decisions are wrong.
  • Said elected officials should not be bound by court decisions they regard as politicized.


  • Promises to work for com­pro­mise across party lines.
  • Promises to stand up to the other party.

Democratic prin­ci­ples

The conjoint exper­i­ment allows us to see whether the com­mit­ments to demo­c­ra­t­ic prin­ci­ples that voters express in the abstract are reflected in the choices they make between can­di­dates. Bright Line Watch’s survey research to date has largely focused on a set of 27 prin­ci­ples, many of which are directly reflected in the candidate choice exper­i­ment. Specifically:

Voting rights/access.  This attribute reflects two prin­ci­ples related to elections:

  • All adult citizens have equal oppor­tu­ni­ty to vote
  • Voter par­tic­i­pa­tion in elections is generally high

Advocates of voter ID laws contend that they are necessary to combat voter fraud, but research shows that voter fraud is exceed­ing­ly rare (Levitt 2007, Minnite 2010, Cottrell, Herron, and Westwood 2018) and that the dis­en­fran­chis­ing effect of such laws dis­pro­por­tion­ate­ly affects racial and ethnic minori­ties (Ansolabehere and Hersh 2017, Hajnal, Lavejardi, and Nielson 2017). The disparate effects of voter ID laws undermine the principle that all citizens should have an equal oppor­tu­ni­ty to vote and that par­tic­i­pa­tion in elections should be high (Sargent 2018).

Investigations/rule of law.  This attribute reflects three prin­ci­ples related to accountability:

  • Government officials are legally sanc­tioned for misconduct
  • Law enforce­ment inves­ti­ga­tions of public officials or their asso­ciates are free from political influence or interference
  • Government agencies are not used to monitor, attack, or punish political opponents

Legal inves­ti­ga­tions that are impartial and free from partisan influence are an essential element of the rule of law. When politi­cians or parties are able to “capture the referees” by appoint­ing, intim­i­dat­ing, or con­trol­ling legal inves­ti­ga­tions, those who hold elected office will be free from effective oversight and can poten­tial­ly target their political opponents with harass­ment and polit­i­cal­ly motivated legal sanctions.

Judicial deference.  This attribute reflects two prin­ci­ples related to checks on political authority:

  • The judiciary is able to effec­tive­ly limit executive power
  • The elected branches respect judicial independence

Institutional checks on executive authority are widely regarded as an essential component of liberal democracy (V‑Dem 2017). Respect for court decisions, even polit­i­cal­ly unfa­vor­able ones, is essential to the effec­tive­ness of such checks and to the rule of law more generally. Efforts by politi­cians to intim­i­date judges — for instance, by por­tray­ing them as corrupt or incom­pe­tent and ques­tion­ing the legit­i­ma­cy of their decisions — fre­quent­ly signal the early stages of demo­c­ra­t­ic erosion (Mounk 2018; Levitsky and Ziblatt 2018). 

Compromise.  This attribute reflects a principle related to norms of behavior among public officials: 

  • Elected officials seek com­pro­mise with political opponents

Willingness to com­pro­mise is a demo­c­ra­t­ic value. The peaceful res­o­lu­tion of political conflict requires accep­tance of policy outcomes that are less than outright victories. Indeed, Levitsky and Ziblatt (2018) contend that mutual tol­er­a­tion among political adver­saries and for­bear­ance in exer­cis­ing power — which, together, describe com­pro­mise — are the “master norms” that sustain all other demo­c­ra­t­ic prin­ci­ples, including the rule of law, electoral integrity, and guar­an­tees of indi­vid­ual rights. 

In September 2017, we asked a rep­re­sen­ta­tive sample of Americans to rate each of the eight prin­ci­ples itemized above (along with 19 others) as either “Not relevant,” “Beneficial,” “Important,” or “Essential” to democracy. Majorities valued all eight, but based on the per­cent­age of par­tic­i­pants who rated each as either “important” or “essential,” we found sub­stan­tial variation in priority across items. 

The impor­tance of equal voting rights was ranked near the top of the list by the public (89% important or essential) though high levels of electoral par­tic­i­pa­tion were somewhat lower (74%). The prin­ci­ples asso­ci­at­ed with account­abil­i­ty were also rated high by most respon­dents, who indicated that public office-holders should be sanc­tioned for mis­con­duct (84%), that inves­ti­ga­tions should not be com­pro­mised by politics (84%), and that gov­ern­ment agencies should be polit­i­cal­ly neutral (80%). Institutional checks on authority came in just behind, with 82% and 77% rating judicial inde­pen­dence and the judiciary’s ability to limit executive authority as important or essential, respectively. 

By contrast, only 59% rated the norm of seeking com­pro­mise with political opponents to be important or essential. The rel­a­tive­ly low priority the public attaches attached to com­pro­mise is poten­tial­ly at odds with the cen­tral­i­ty of mutual tol­er­a­tion and for­bear­ance to Levitsky and Ziblatt (2018). If they are correct, then the value that par­tic­i­pants place on com­pro­mise in this exper­i­ment is espe­cial­ly important to measuring the public’s com­mit­ment to American democracy. 

More generally, as we examine the results of our candidate choice exper­i­ment, we focus attention on how par­tic­i­pants react to each candidate attribute and to the relative impact of those related to par­ti­san­ship, to policy positions, and to demo­c­ra­t­ic principles.

Candidate choices: Republicans, Democrats, and independents

Our exper­i­ment confirms the over­whelm­ing impor­tance of par­ti­san­ship in Americans’ political behavior. Figure 2 illus­trates the relative effect of each attribute on the like­li­hood of voters sup­port­ing a candidate for Republican respon­dents (left panel), Democrats (middle panel) and inde­pen­dents (right panel). [7] [8]  Each attribute is assigned a “baseline” level, which serves as our point of com­par­i­son. [9] Each dot in Figure 2 rep­re­sents the estimated change, relative to the baseline, in the like­li­hood of a respon­dent sup­port­ing a candidate with that char­ac­ter­is­tic (for example, being a Democrat) compared to the baseline value (for par­ti­san­ship, Republican), if all other candidate char­ac­ter­is­tics were held equal. Dots to the right of the vertical line in each panel indicate positive effects, and dots to the left of the line negative ones. [10] The hor­i­zon­tal lines extending through these dots are “con­fi­dence intervals” indi­cat­ing sta­tis­ti­cal uncer­tain­ty around those estimated effects. [11]

Figure 2: Candidate pref­er­ences among Republicans, Democrats, and independents

The effects of candidate par­ti­san­ship are for­mi­da­ble. All else equal, Republicans are 19 per­cent­age points more likely to select a Republican candidate than a Democratic one. Partisanship is sym­met­ri­cal on the other side of the aisle. Democrats are similarly 19 per­cent­age points more likely to select a fellow Democrat than a Republican. It is worth under­scor­ing that these effects are inde­pen­dent of candidate char­ac­ter­is­tics, policy positions, or demo­c­ra­t­ic values, all of which varied randomly across the composite can­di­dates from which par­tic­i­pants chose. Independents, true to their name, appear indif­fer­ent between the two parties — they are no more likely to select a Republican than a Democrat.

Partisanship far outweighs demo­graph­ic traits in shaping voters’ choices. Race/ethnicity does not sig­nif­i­cant­ly shift the choices of either partisan group or of inde­pen­dents. That is, the effects of a candidate having a typically black or Hispanic name rather than a white one are not dis­tin­guish­able from zero for any partisan group. Among Democrats or inde­pen­dents, there is similarly no mea­sur­able effect of candidate gender, whereas Republicans are 2.5 per­cent­age points less likely to select a woman than a man, other things equal (and the effect just reaches sta­tis­ti­cal sig­nif­i­cance). On the whole, the effects of race/ethnicity and gender on candidate choice are slight. 

Policy positions tend to have a larger effect on respon­dents’ choices than candidate demo­graph­ics. Republicans are nearly nine per­cent­age points less likely to select a candidate who wants to increase taxes on the wealthy than one who wants to reduce taxes across the board, whereas Democrats are 11.5 per­cent­age points more likely to select a candidate who wants to increase taxes on the wealthy. Independents are, once again, in the middle; a candidate’s policy position on taxes has no mea­sur­able effect on their choices. 

Similarly, though the race and gender of can­di­dates them­selves has little impact on par­tic­i­pants’ choices, can­di­dates’ stances on dis­crim­i­na­tion and affir­ma­tive action do matter. Republicans are 3.8 per­cent­age points less likely to select a candidate who wants the gov­ern­ment to do more to protect minori­ties than one who thinks dis­crim­i­na­tion is less of a problem now than in the past. Democrats, in sharp contrast, are 13.1 per­cent­age points more likely to select candidate who wants the gov­ern­ment to do more to protect minori­ties. Here, inde­pen­dents appear to lean slightly toward the Democratic position, although the estimate falls short of sta­tis­ti­cal significance.

In short, these con­tro­ver­sial policy issues still divide partisans. The effects of our taxation and racial dis­crim­i­na­tion policy positions on candidate choices are not as large as the effects of party labels, but they are powerful.

Democratic prin­ci­ples and candidate choice

Notably, can­di­dates’ stances on demo­c­ra­t­ic prin­ci­ples make a dif­fer­ence in voters’ will­ing­ness to support them even after account­ing for the candidate’s traits, party, and policy positions. In most instances, voters punish can­di­dates who hold positions anti­thet­i­cal to the demo­c­ra­t­ic prin­ci­ples that map onto those from our Bright Line Watch surveys, as described above. We specif­i­cal­ly estimate the effect of taking four positions that violate these values: restrict­ing voter access, dis­re­gard­ing unfa­vor­able judicial decisions, politi­ciz­ing inves­ti­ga­tions, and seeking con­fronta­tion over com­pro­mise. On three of these issues, demo­c­ra­t­ic values are supported by all three groups: Republicans, Democrats, and inde­pen­dents. On ballot access, however, pref­er­ences diverge by party.

Specifically, we find that a candidate who questions judicial authority rather than deferring to it is 12.7 per­cent­age points less likely to be selected by Republicans, 9.7 per­cent­age points less likely to be selected by Democrats, and 9.7 per­cent­age points less likely to be selected by inde­pen­dents. These estimates are sta­tis­ti­cal­ly indis­tin­guish­able from each other. Likewise, a candidate who favors political control over inves­ti­ga­tions to neu­tral­i­ty is nearly four per­cent­age points less likely to win the support of Republican voters, 5.3 per­cent­age points less likely among Democratic voters, and 8.1 per­cent­age points less likely among inde­pen­dents. Again, the dif­fer­ences by party are not sig­nif­i­cant — from a sta­tis­ti­cal stand­point, Republicans, Democrats, and inde­pen­dents are equally inclined to punish partisan investigations. 

In turn, a candidate who promises to stand up to the other party rather than to seek com­pro­mise is 6.0 per­cent­age points, 5.6 per­cent­age points, and 11.5 per­cent­age points less likely to be selected by Democrats, Republicans, and inde­pen­dents, respec­tive­ly. In this case, the dif­fer­ences between Democrats and inde­pen­dents, and between Republicans and inde­pen­dents, are sta­tis­ti­cal­ly sig­nif­i­cant. Both Republicans and Democrats value cross-party com­pro­mise and reward can­di­dates who promise to seek it, but inde­pen­dents value com­pro­mise more.

In sum, there is consensus across parties on the three demo­c­ra­t­ic prin­ci­ples of judicial deference, inde­pen­dent inves­ti­ga­tions, and com­pro­mise. Participants from all groups reward can­di­dates who espouse these prin­ci­ples and punish those who do not.[12]

By contrast, we see sharp partisan polar­iza­tion on voting rights. Republican par­tic­i­pants are 16.6 per­cent­age points more likely to select a candidate who supports voter ID laws than one who opposes them. This effect is larger than for any other attribute except the candidate’s party label itself. (Independents also prefer can­di­dates who support voter ID laws and are more likely to support them by 7.9 per­cent­age points.) Democrats have the opposite pref­er­ence. If a candidate supports voter ID laws, Democratic par­tic­i­pants are 8.9 per­cent­age points less likely to select the candidate, all else equal. Overall, a candidate who supports voter ID laws is more likely to be selected than one who opposes them across our full sample of participants. 

What does this tell us about party, policy, and democracy?

The question of how voters make tradeoffs between party, policy, and demo­c­ra­t­ic principle could not be more relevant. In the wake of the 2018 midterm elections, we have seen the integrity of the Mueller inves­ti­ga­tion come under threat with the forced res­ig­na­tion of Attorney General Jeff Sessions and President Trump’s con­tro­ver­sial appoint­ment of an acting replace­ment who may quash the inves­ti­ga­tion. In Georgia, where the election is still being contested, Republicans must choose whether to support a guber­na­to­r­i­al candidate who shares their party and policy pref­er­ences but faces alle­ga­tions of voter sup­pres­sion. Similarly, Florida governor Rick Scott has threat­ened to use state law enforce­ment to intervene in the vote count in his race for U.S. Senate, prompting a sharp tradeoff between partisan and demo­c­ra­t­ic loyalties for Republicans.

Our results provide both encour­ag­ing and sobering con­clu­sions about the American public’s demo­c­ra­t­ic pri­or­i­ties in this difficult context. Not sur­pris­ing­ly, respon­dents divide sharply on par­ti­san­ship and on policy pref­er­ences, which reflects the state of con­tem­po­rary American politics and reassures us that our candidate choice exper­i­ment worked as intended. 

Consistent with previous BLW studies, we also find that partisans are most likely to find common ground on matters of rule of law and account­abil­i­ty. Participants from both parties (as well as inde­pen­dents) support can­di­dates who promise to respect court decisions and those who pri­or­i­tize the impar­tial­i­ty of inves­ti­ga­tions into wrong­do­ing. They also support can­di­dates who advocate for com­pro­mise rather than con­fronta­tion across the partisan divide. By contrast, can­di­dates who would politi­cize inves­ti­ga­tions, flout court decisions, or who broadcast intran­si­gence toward the other party may be crossing “bright lines;” they are less likely to be selected by voters of both parties as well as by inde­pen­dents. Not all political values, then, are polarized in our current context. 

We urge caution, however, in inter­pret­ing these seemingly encour­ag­ing findings. Democrats and Republicans both prefer can­di­dates who favor polit­i­cal­ly impartial inves­ti­ga­tions, but con­cep­tions of impar­tial­i­ty might well differ across parties. Democrats tend to regard Robert Mueller’s inves­ti­ga­tion as polit­i­cal­ly unbiased but Republicans view it quite dif­fer­ent­ly. Previous Bright Line Watch research has found consensus, for example, between sup­port­ers and opponents of President Trump that public officials should be punished for malfea­sance, but those groups might well have different officials or types of wrong­do­ing in mind.

We also find troubling levels of partisan polar­iza­tion on the key demo­c­ra­t­ic value of equal voting rights, reflect­ing a fun­da­men­tal division over who should be included in, and excluded from, the political community. Democrats and Republicans respond diver­gent­ly to can­di­dates who support voter ID laws that would restrict par­tic­i­pa­tion in elections and whose burdens could dis­en­fran­chise racial minori­ties at far higher rates than whites. 

It is likely that both sides would claim that they are sup­port­ive of free and fair elections, Republicans by combating voter fraud, Democrats by offering equal access to the polls. These positions echo those of party leaders. The evidence demon­strates, however, that voter fraud is exceed­ing­ly rare (Levitt 2007, Minnite 2010, Cottrell, Herron, and Westwood 2018). Moreover, strict voter ID laws are most fre­quent­ly — though not exclu­sive­ly — found in Southern states with a history of racially dis­crim­i­na­to­ry forms of election administration. 

What strate­gies do our results suggest for those concerned about demo­c­ra­t­ic erosion or back­slid­ing in the U.S.? Cross-party agreement that par­ti­san­ship should not play a part in inves­ti­ga­tions may partly reflect weariness with polit­i­cal­ly motivated inquiries (as in the Clinton impeach­ment). With Democrats poised to take over the House of Representatives, this dynamic is likely continue. In some settings, leg­isla­tive arms races are brought to a halt through a kind of agreed-upon mutual restraint, as both sides tire of conflict that does not offer a lasting advantage to either side. We are nowhere near such a “cease-fire” in the U.S. today. But our survey provides evidence that the public (and par­tic­u­lar­ly inde­pen­dents) might be open to one.

By contrast, such an accord seems still further away on ballot access, an issue where one major party benefits at the expense of the other. Is there no room for a shift away from this anti-demo­c­ra­t­ic practice? Looking beyond the polar­iza­tion evident in our exper­i­ment, we see some poten­tial­ly encour­ag­ing signs. First, strict voter ID laws have not con­sis­tent­ly dimin­ished turnout among minori­ties in practice (Citrin, Green, and Levy 2014, Valentino and Neuner 2016, Grimmer et al. 2018, Aytaç and Stokes 2018; but see Hajnal, Lavejardi, and Nielson 2017). Just as these efforts make it harder for voters to cast ballots, they also can motivate mobi­liza­tion and voter turnout. Such backlash effects could make tactics that restrict access to the franchise somewhat less appealing, espe­cial­ly if they produce lasting change in par­ti­san­ship or turnout in the com­mu­ni­ties that are dis­pro­por­tion­ate­ly affected (most notably, Hispanics). 

Finally, we note the outcome last week of the ref­er­en­dum on Florida’s Amendment 4, which rein­states voting rights for ex-felons who have completed their sentences. The measure won 61% of votes among Florida’s evenly divided elec­torate, which indicates that sub­stan­tial numbers of Republicans supported it. For instance, prominent evan­gel­i­cals argued in its favor as con­sis­tent with the Christian ideal of redemp­tion. So although voter ID pro­vi­sions divide the parties, it may yet be possible for the principle of inclusive suffrage and broad voting par­tic­i­pa­tion to gain purchase across partisan lines.



Polarization, and consensus, across other groups

Our data allow us to compare survey par­tic­i­pants by not only their par­ti­san­ship, but across other salient demo­graph­ic and atti­tu­di­nal divides. We can, for example, break out our par­tic­i­pants by their age, education level, level of political knowledge, or level of interest in politics to evaluate other dif­fer­ences in demo­c­ra­t­ic preferences.

We begin by con­sid­er­ing dif­fer­ences by age. Some scholars have argued that younger gen­er­a­tions of Americans are increas­ing­ly cynical about democracy (Foa and Mounk 2016, Drutman, Diamond, and Goldman 2018a, 2018b), while others hold that older adults have less faith in demo­c­ra­t­ic insti­tu­tions (Voeten 2017). However, we find little evidence for either hypoth­e­sis when we compare par­tic­i­pants who are 49 years old and above (the median age in our sample) to those who are younger — the effects of sup­port­ing demo­c­ra­t­ic values on judicial deference, inves­ti­ga­tions, and com­pro­mise do not mea­sur­ably differ between older and younger par­tic­i­pants. We do, however, find that positions on voter ID laws have no mea­sur­able effect on candidate support among younger voters, while older par­tic­i­pants are 6.5 per­cent­age points more likely to select a candidate who supports them. (The dif­fer­ence between the groups is sta­tis­ti­cal­ly significant.)


Turning to education level, political knowledge, and political interest, we asked par­tic­i­pants in our survey to report their education level, to answer a series of five questions on politics (Carpini and Keeter 1993), and to report their level of interest in politics. For education level, we divided respon­dents into groups based on whether they hold a bachelor’s degree, while for political knowledge, we used the number of correct responses they provided to classify them as having high or low knowledge of politics (using a median split). Finally, we clas­si­fied par­tic­i­pants who are “extremely inter­est­ed” or “very inter­est­ed” in politics as having high political interest and those who are “somewhat,” “not very,” or “not at all” inter­est­ed in politics as low in political interest.

Research suggests that support for democracy is weakest among the least educated and the least inter­est­ed in politics (Drutman, Diamond, and Goldman 2018a, 2018b). To an extent, our results cor­rob­o­rate this account. Participants with a bachelor’s degree or higher are 8.8 per­cent­age points less likely to select a candidate who says elected officials should supervise inves­ti­ga­tions than one who says inves­ti­ga­tions should be free from partisan influence, while those with less than a bachelor’s degree are 3.7 per­cent­age points less likely to select the candidate. More educated par­tic­i­pants are also 14.5 per­cent­age points less likely to support a candidate who says elected officials should not be bound by politi­cized decisions than one who says politi­cians should obey the courts; the estimate for less educated par­tic­i­pants is 9.0 per­cent­age points. For both attrib­ut­es, the dif­fer­ences between more and less educated par­tic­i­pants are sig­nif­i­cant at the 95% con­fi­dence level. These results replicate when we compare par­tic­i­pants by their level of political knowledge (7.8 per­cent­age points versus 1.3 per­cent­age points for inves­ti­ga­tions, and 13.5 per­cent­age points versus 6.2 per­cent­age points for courts) or political interest (7.7 per­cent­age points vs 2.3 per­cent­age points for inves­ti­ga­tions, and 12.4 per­cent­age points versus 8.5 per­cent­age points for courts). For the attrib­ut­es on voter ID laws and com­pro­mise, we find no sig­nif­i­cant dif­fer­ences by education level, political knowledge, or political interest, with all groups slightly favoring can­di­dates who support voter ID laws and who promise to work for com­pro­mise across party lines. [13]

Average marginal component effects among Republicans, Democrats, and independents

The chart below shows the point estimates for each attribute-level, relative to a baseline, as depicted in Figure 2. 


[1]  In their candidate choice exper­i­ment, Graham and Svolik (2018) also find that partisan and policy con­sid­er­a­tions have much stronger effects on vote choice than “unde­mo­c­ra­t­ic” platforms related to electoral fairness or checks and balances.

[2]  Concurrent with our survey exper­i­ment, we also conducted our seventh expert survey, and fifth public survey, to inves­ti­gate how Americans rate the per­for­mance of their democracy on 27 distinct demo­c­ra­t­ic prin­ci­ples. Our expert survey also included assess­ments of the impor­tance of 27 notable recent political events and eval­u­a­tions of how (ab)normal they are. See our Bright Line Watch report on these findings and the raw data for more information.

[3] This dis­tri­b­u­tion approx­i­mates that from a March 2018 Pew poll that found 43%, 46%, and 11%, for Democrats, Republicans, and inde­pen­dents, respec­tive­ly. The slight dis­crep­an­cies we find may be partially attrib­ut­able to wording dif­fer­ences between the par­ti­san­ship question on the surveys. 

[4] Each par­tic­i­pant in the exper­i­ment saw ten pairs of can­di­dates. Candidate names always appeared at the top followed by par­ti­san­ship. This format approx­i­mates how infor­ma­tion about can­di­dates is often supplied to voters, for example in voter guides or in news­pa­pers. Beyond name and par­ti­san­ship, we ran­dom­ized the order of the policy and democracy attrib­ut­es across par­tic­i­pants so that, on the whole, their placement (lower or higher in the tables) would not affect their relative salience.

[5] Our list of candidate names come from Butler’s (2014) audit study on racial dis­crim­i­na­tion among leg­is­la­tors, which used infor­ma­tion from previous studies of racial dis­crim­i­na­tion (Fryer and Levitt 2004, Word et al. n.d.) and from the U.S. Census to compile a list of 123 common puta­tive­ly white, black, and Hispanic first and last names. The full list is included in Butler and Homola (2017). We used candidate names to signal gender and race/ethnicity for a number of reasons. First, using names increases realism. Second, using a name to signal both gender and race/ethnicity is eco­nom­i­cal in terms of space in the survey instru­ment. Finally, using names reduces concerns about social desir­abil­i­ty bias in survey responses while still allowing us to estimate how gender and race/ethnicity affect pref­er­ences between can­di­dates (for example, Bertrand and Mullainathan 2004, Broockman 2013, DeSante 2013, White, Faller and Nathan 2015, Doherty, Dowling, and Miller 2018).

[6] Candidate gender was ran­dom­ized with prob­a­bil­i­ty 0.5. Candidate race/ethnicity was ran­dom­ized to be white, black, or Hispanic with prob­a­bil­i­ties 0.6, 0.2, and 0.2, respec­tive­ly, to better approx­i­mate race/ethnicity in the general pop­u­la­tion and among can­di­dates for public office in con­tem­po­rary campaigns (Bialik and Krogstad 2017, U.S. Census Bureau 2017, Reflective Democracy Campaign 2018).

[7] We follow the con­ven­tion of grouping inde­pen­dents who “lean” toward one party or the other with self-iden­ti­fied partisans (Petrocik 2009, Hawkins and Nosek 2012). Our survey included a question designed to screen out respon­dents who do not provide sincere responses to surveys. 3.8% of our respon­dents were removed from the sample on the basis of their response to that item. 

[8] All the estimates, and standard errors, presented in the figure are also presented in a table format in the appendix.

[9] Any level, or value, may serve as the baseline. Most of our attrib­ut­es take only two levels, though race/ethnicity can take three (white, black, or Hispanic). In that case, we use white as the baseline because it is the most commonly observed among elected officials (Reflective Democracy Campaign 2018).

[10] The dots for the baseline level of each attribute are, by def­i­n­i­tion, located on the vertical line at zero.

[11] If the con­fi­dence interval touches or crosses the vertical line, then we cannot say with more than 95% con­fi­dence whether the estimated effect is different from zero. By contrast, if the interval does not overlap the vertical line, the result is sta­tis­ti­cal­ly dis­tin­guish­able from zero at the 95% con­fi­dence level.

[12]Graham and Svolik (2018) also find that com­mit­ments to demo­c­ra­t­ic prin­ci­ples related to checks and balances and to electoral fairness are roughly symmetric across the partisan divide.

[13] For par­tic­i­pants with a bachelor’s degree or higher, there is no sig­nif­i­cant dif­fer­ence in like­li­hood of support for can­di­dates who support versus oppose voter ID laws. 



Alexander, Amy C. and Christian Welzel. 2017. “The Myth of Deconsolidation: Rising Liberalism and the Populist Reaction.” Journal of Democracy. Downloaded November 9, 2018 from“democratic-deconsolidation”.

Ansolabehere, Stephen, and Eitan D. Hersh. 2017. “ADGN: An Algorithm for Record Linkage Using Address, Date of Birth, Gender, and Name.” Statistics and Public Policy 4 (1). Taylor & Francis: 1–10.

Arnesen, Sveinung, Dominik Duell, and Mikael Poul Johannesson. 2019. “Do Citizens Make Inferences from Political Candidate Characteristics When Aiming for Substantive Representation?” Forthcoming, Electoral Studies.

Aytaç, S. Erdem, and Susan C. Stokes. 2018. Why Bother? Rethinking Participation Elections and Protests. Cambridge, U.K.: Cambridge University Press

Azari, Julia. 2016. “Weak Parties and Strong Partisanship Are a Bad Combination.” Vox.

Bartels, Larry M. 2018. “Partisanship in the Trump Era.” Working paper, available at:

Bialik, Kristen, and Jens Manuel Krogstad. 2017. “115th Congress Sets New High for Racial, Ethnic Diversity.” Pew Research Center.

Broockman, David E. 2013. “Black Politicians Are More Intrinsically Motivated to Advance Blacks’ Interests: A Field Experiment Manipulating Political Incentives.” American Journal of Political Science 57 (3): 521–36.

Butler, Daniel M. 2014. Representing the Advantaged: How Politicians Reinforce Inequality. Cambridge, U.K.: Cambridge University Press.

Butler, Daniel M., and Jonathan Homola. 2017. “An Empirical Justification for the Use of Racially Distinctive Names to Signal Race in Experiments.” Political Analysis 25 (1): 122–30.

Carpini, Michael X. Delli, and Scott Keeter. 1993. “Measuring Political Knowledge: Putting First Things First.” American Journal of Political Science 37 (4): 1179–1206.

Citrin, Jack, Donald P. Green, and Morris Levy. 2014. “The Effects of Voter ID Notification on Voter Turnout: Results from a Large-Scale Field Experiment.” Election Law Journal: Rules, Politics, and Policy 13 (2): 228–42.

Cottrell, David, Michael C. Herron, and Sean J. Westwood. 2018. “An Exploration of Donald Trump’s Allegations of Massive Voter Fraud in the 2016 General Election.” Electoral Studies 51: 123–142.

DeSante, Christopher D. 2013. “Working Twice as Hard to Get Half as Far: Race, Work Ethic, and America’s Deserving Poor.” American Journal of Political Science 57 (2): 342–56.

Doherty, David, Conor M Dowling, and Michael G Miller. 2018. “Do Party Chairs Think Women and Minority Candidates Can Win? Evidence from a Conjoint Experiment.” Forthcoming, The Journal of Politics.

Drutman, Lee, Larry Diamond, and Joe Goldman. 2018a. “Follow the Leader: Explaining American Support for Democracy and Authoritarianism.” Report prepared for the Democracy Fund Voter Study Group, available at:

Drutman, Lee, Larry Diamond, and Joe Goldman. 2018b. “Testing the Limits: Examining Public Support for Checks on Presidential Power.” Report prepared for the Democracy Fund Voter Study Group, available at:

Fishkin, Joseph, and David E. Pozen. 2018. “Asymmetric Constitutional Hardball.” Columbia Law Review 118.

Foa, Roberto Stefan, and Yascha Mounk. 2016. “The Danger of Deconsolidation: The Democratic Disconnect.” Journal of Democracy 27 (3): 5–17.

Foa, Roberto Stefan, and Yascha Mounk. 2017. “The Signs of Deconsolidation.” Journal of Democracy 28 (1).

Fryer, Roland, and Steven Levitt. 2004. “The Causes and Consequences of Distinctively Black Names.” Quarterly Journal of Economics 119 (3): 767–805.

Graham, Matthew, and Milan W. Svolik. 2018. “Democracy in America? Partisanship, Polarization, and the Robustness of Support for Democracy in the United States.” Working paper, available at:

Grimmer, Justin, Eitan Hersh, Marc Meredith, Jonathan Mummolo, and Clayton Nall. 2018. “Obstacles to Estimating Voter ID Laws’ Effect on Turnout.” The Journal of Politics 80 (3): 1045–51.

Hajnal, Zoltan, Nazita Lajevardi, and Lindsay Nielson. 2017. “voter ID Laws and the Suppression of Minority Votes.” The Journal of Politics 79 (2): 363–79.

Hawkins, Carlee Beth, and Brian A. Nosek. 2012. “Motivated Independence? Implicit Party Identity Predicts Political Judgments Among Self-Proclaimed Independents.” Personality and Social Psychology Bulletin 38 (11): 1437–1452.

Hainmueller, Jens, Daniel J Hopkins, and Teppei Yamamoto. 2013. “Causal Inference in Conjoint Analysis: Understanding Multidimensional Choices via Stated Preference Experiments.” Political Analysis 22 (1): 1–30.

Huddy, Leonie, Lilliana Mason, and Lene Aarøe. 2015. “Expressive Partisanship: Campaign Involvement, Political Emotion, and Partisan Identity.” American Political Science Review 109 (1): 1–17.

Lee, Frances E. 2009. Beyond Ideology: Politics, Principles, and Partisanship in the U.S. Senate. Chicago, IL: University of Chicago Press.

Levitsky, Steven, and Daniel Ziblatt. 2018. How Democracies Die. New York, NY: Crown.

Levitt, Justin. 2007. “The Truth About Voter Fraud.” New York, NY: Brennan Center for Justice. Available at Truth About Voter Fraud.pdf.

Mason, Liliana. 2018. Uncivil Agreement: How Politics Became Our Identity. Chicago: University of Chicago Press.

McConnell, Christopher, Yotam Margalit, Neil Malhotra, and Matthew Levendusky. 2018. “The Economic Consequences of Partisanship in a Polarized Era.” American Journal of Political Science 62 (1): 5–18.

Minnite, Lorraine C. 2010. The Myth of Voter Fraud. Ithaca, NY: Cornell University Press.

Norris, Pippa. 2017. “Is Western Democracy Backsliding? Diagnosing the Risks.” Journal of Democracy. Downloaded November 9, 2018 from“democratic-deconsolidation”.

Petrocik, John Richard. 2009. “Measuring Party Support: Leaners Are Not Independents.” Electoral Studies 28 (4): 562–572.

Pew Research Center. 2017. “The Partisan Divide on Political Values Grows Even Wider.” Trust, Facts, and Democracy. Pew Research Center.

Pew Research Center. 2018. “The Public, the Political System and American Democracy.” Trust, Facts, and Democracy.

Reflective Democracy Campaign. 2018. “A Rising Tide? The Changing Demographics on Our Ballots.” Report prepared by the Reflective Democracy Campaign, available at:

Sargent, Greg.  2018. An Uncivil War: Taking Back Our Democracy in an Age of Trumpian Disinformation and Thunderdome Politics. New York, NY: HarperCollins Publishers.

Svolik, Milan W. 2018. “When Polarization Trumps Civic Virtue: Partisan Conflict and the Subversion of Democracy by Incumbents.” Working paper, available at:

United States Census Bureau. 2017. “U.S. Census Bureau QuickFacts: United States.” U.S. Census Bureau QuickFacts.

Valentino, Nicholas A., and Fabian G. Neuner. 2017. “Why the Sky Didn’t Fall: Mobilizing Anger in Reaction to Voter ID Laws.” Political Psychology 38 (2): 331–50.

Voeten, Erik. 2017. “Are People Really Turning Away from Democracy?” Journal of Democracy. Downloaded November 9, 2018 from“democratic-deconsolidation”.

White, Ariel R., Noah L. Nathan, and Julie K. Faller. 2015. “What Do I Need to Vote? Bureaucratic Discretion and Discrimination by Local Election Officials.” American Political Science Review 109 (1): 129–142.

Wike, Richard, and Janell Fetterolf. 2018. “Liberal Democracy’s Crisis of Confidence.” Journal of Democracy 29 (4): 136–150.

Word, David L, Charles D. Coleman, Robert Nunziata, and Robert Kominski. 2018. “Demographic Aspects of Surnames from Census 2000.” Technical Report for the U.S. Census Bureau. Downloaded November 9, 2018 from