7. Peeking Under the Blanket
A Direct Look at Crossover Voting in the 1998 Primary
Anthony M. Salvanto and Martin P. Wattenberg
Before California's blanket primary went into effect in 1998, its proponents argued that the new system would produce more moderate candidates: with all voters now able to vote in any contest, they reasoned, candidates with broader appeals would win. Opponents of the new system saw it as an infringement on a party's right to choose its own nominees. They feared that Democrats' nomination fights could now be unfairly influenced—perhaps even determined—by registered Republicans, and vice-versa. Minor-party contests might be especially vulnerable; with their low vote totals, they could easily be flooded by voters from the major parties. Moreover, many opponents argued that the new rules could advantage incumbents, whose name recognition would draw voters from outside the party in numbers that their challengers could never match.
All of these predictions—whether hopeful or fearful—rested on one key idea: that voters would move back and forth across party lines in significant numbers. In this chapter we take a direct look at voting behavior in the blanket primary and examine whether or not this was the case. We show how much crossover voting actually occurred, and we address why it occurred, including the electoral conditions that encouraged or discouraged it. We also consider the critical issue of whether crossover voting could potentially change electoral outcomes, either because it favors a certain type of candidate or because the vote choices of party members differ from those of crossover voters.
Our data consist of more than one-quarter million actual absentee ballots cast in the 1998 primary election in Los Angeles County. With the actual ballots, we can provide a complete and accurate picture of voters' selections across the entire range of contests. The ballots are anonymous,
CROSSOVER VOTING: WHAT WE EXPECT
We consider three possible reasons behind crossover voting. These reasons correspond to the three motivations for crossover voting discussed in the introduction to this volume. First, as an example of sincere crossover voting, voters may cross party lines to support an incumbent when one is on the ballot. This is a form of sincere crossover voting in the sense that the voter prefers the incumbent to the other candidates by virtue of the former's incumbency status. We may attribute this preference to available information: incumbents enjoy a strong advantage in name recognition (Jacobson 1992), particularly in local and district-level contests. This factor might play an especially large role in a blanket primary, as voters who are faced with a long list of candidates might simply gravitate to the most recognizable name. By and large, incumbents also have more money and more avenues through which to communicate with the voters, and (if they are popular) they are likely to have supporters among nonpartisans and members of the other party. For these reasons, we expect to see large numbers of voters crossing party lines to vote for incumbents.
Electoral competition may suffer as a result of such behavior. Opponents of the blanket primary argued that the new system would greatly benefit incumbents. To the extent that voters engage in this sort of crossover voting, incumbents draw votes not only from supporters in their own party but also from any other voters who might approve of their performance, or who simply recognize their names. Meanwhile, candidates challenging incumbents for a nomination could rarely hope to match that influx of votes from outside their party. We expect that incumbents, when challenged, will receive the vast majority of all the crossover votes cast in the race.
A second hypothesized reason behind crossover voting is that voters might be drawn to where the action is: they might cross party lines to vote in competitive nomination fights whenever their own party offers an uncontested or noncompetitive race. This voting would not necessarily be sincere, as their most preferred candidate might still be the one running unopposed in their own party, but the chance to cast a meaningful or even decisive vote in the competitive race could provide enough incentive to make them cross (Downs 1957). In the language of the introductory chapter, these voters are engaging in strategic hedging. Such voting could also be inspired by a heavier flow of campaign information emanating from the close races, which usually produce more advertising and more news stories—thus making more information available. Information leads to participation:
In a more general way, a lopsided flow of information could spur crossover voting among registered partisans who live in districts dominated by the other party. These voters would likely be exposed to a steady stream of political information from the other party—and the blanket primary would now give them a chance to act on it. We expect to see them do so.
Under a blanket primary system, voters also have the chance to use their votes in a malicious way: they could try raiding into another party in an attempt to sabotage outcomes there. This is the third possible motive behind crossover voting that we consider. Registered members of one party can, conceivably, study the other party's candidates and try to pick out the weakest one—that is, the one with the least chance of winning the general election. These partisans could then cross party lines in the primary and vote for that weak candidate in an attempt to throw the nomination to her. If they prove successful, then their own party's candidate would face that weaker nominee in November.
Although it represents an insincere vote choice, such strategic behavior might sound like an irresistible, even ingenious, idea for party loyalists. Indeed, many opponents of the blanket primary feared that this type of sabotage would unfairly swing some election results. Yet we expect that raiding in a blanket primary system will be very rare. First, the amount of political knowledge required to engage in such behavior is very high, and probably beyond the level of most voters. It would involve not only learning about candidates in the other party, but also making savvy predictions about their long-term fortunes. Second, without a massive and coordinated effort from thousands of voters, any single person considering a raiding strategy would probably come to see it as a wasted vote. If she did not fully anticipate like-minded fellow partisans acting in concert, then her vote would be squandered on a weak candidate from another party who was probably certain to lose.
DATA SOURCES
The absentee ballots examined in this chapter offer a unique insight into crossover voting. There were more than two hundred and fifty thousand absentee ballots cast in L.A. County in the June 1998 primary election, which constituted more than 20 percent of all the ballots cast in the election. The absentee ballots were reasonably representative of the full election-day canvass. (See the appendix and table 7.10, at the end of this chapter,
One possible limitation with the absentee ballots, however, is that absentee voters fill out ballots early in the campaign. They would therefore have been at a disadvantage in planning strategic votes, because information that could have aided in that planning—such as media coverage of campaigns and the reporting of poll results—intensifies closer to election day. Thus, whatever amount of strategic voting we see here may represent the minimum found in the election. This factor could be offset, however, by the fact that absentee voters had plenty of time with the ballot in hand and perhaps used it to consider all their options. That extra time could also mean that absentee voters were the least likely (all else being equal) to pick the first candidate on a list whose name they recognized. Hence, the amount of crossover voting to support incumbents found among absentee voters was probably at least matched by the overall electorate. The final vote tallies support this conjecture as well.
A second possible limitation with the absentee ballots is that we cannot measure any possible campaign-specific effects that could have affected voters' decisions, such as a vote-by-mail drive in a particular district. However, we compensate for this possible limitation by examining more than fifty state and local contests and by observing voters across a range of electoral scenarios. Our conclusions about how voters behaved in any given electoral scenario (e.g., with the presence of an incumbent or of a competitive race) are drawn from crossover rates in a number of such cases.
OVERALL PATTERNS OF CROSSOVER VOTING
We begin the analysis by showing the overall patterns of voting in the sample. Table 7.1 shows how voters distributed their votes among the parties across eleven major offices. It shows the breakdown of votes cast by each of four types of voters: registered Democrats, registered Republicans, registered minor-party members, and nonpartisans (decline-to-states).
In the case of Democrats and Republicans, a plurality of voters cast votes only for candidates of their party of registration. However, the majority of registered partisans split their tickets in some way. For instance, 28.6 percent of Democrats mixed in some Republican votes, and another 10 percent selected at least one minor-party candidate. Registered Republicans were even more prone to cross party lines, with more than 37.8 percent selecting some Democrats and another 10 percent selecting minor-party candidates as well as Democrats.
Among nonpartisan voters, about one-quarter stayed loyal to a single party throughout. Some of these voters were probably people with strong
| Party Registration | ||||
|---|---|---|---|---|
| Pattern | Democrat | Republican | Nonpartisans | Minor Parties |
| Democrats only | 45.6% | 2.7% | 9.6% | 14.1% |
| Republicans only | 2.0 | 40.5 | 16.8 | 5.9 |
| Third parties only | 0.2 | 0.2 | 0.8 | 5.1 |
| Democrats and third parties only | 11.5 | 0.9 | 8.2 | 18.0 |
| Republicans and third parties only | 0.6 | 6.5 | 3.3 | 8.4 |
| Democrats and Republicans only | 28.6 | 37.8 | 37.4 | 19.6 |
| Democrats, Republicans, and third parties | 10.6 | 10.8 | 22.0 | 27.7 |
| Skipped all eleven offices | 0.9 | 0.7 | 1.8 | 1.2 |
| TOTAL | 100.0% | 100.1% | 99.9% | 100.0% |
party affiliations who preferred not to officially declare themselves. Thus, the new rules gave them a chance to participate in a primary for the first time. However, the vast majority of nonpartisans split their tickets, and did so predominantly between the two major parties.
Just 5.1 percent of all voters registered with a minor party selected only minor-party candidates. It was not as if they had no one to vote for: most minor parties fielded candidates in all the major statewide races and in many of the district-level races. Most of those candidates ran uncontested, however, which may suggest that minor-party members were more concerned with the contested races or incumbents elsewhere than they were with giving their party a larger share of the overall vote.
All of this ticket-splitting made the ballot patterns of this primary election look remarkably like a general election. To illustrate this, table 7.2 shows the distribution of votes in this election compared to the same county in the 1994 general election. (The 1994 data also come from the Los Angeles County Registrar and reflect a random sample of election-day ballots.) The percentages show a very similar pattern: voters' propensity to split their tickets, and the way in which they did so, was nearly the same under this new primary format as it was in the 1994 general election. Crossover voting may well be routine behavior for voters any time they are given choices among many parties.
| Pattern | 1998 Primary | 1994 General |
|---|---|---|
| Democrats only | 26.7% | 20.5% |
| Republicans only | 17.1 | 16.5 |
| Third parties only | 0.3 | 0.5 |
| Democrats and third parties only | 7.4 | 8.8 |
| Republicans and third parties only | 3.2 | 4.9 |
| Democrats and Republicans only | 32.6 | 30.6 |
| Democrats, Republicans, and third parties | 11.9 | 18.0 |
| Skipped all eleven offices | 0.9 | 0.3 |
| TOTAL | 100.1% | 100.1% |
| NOTE: Figures are percentages of ballots showing a given pattern. (Figures may not total 100 percent due to rounding.) |
This finding helps to shed light on the motivations behind crossover voting. The patterns in table 7.2 suggest that on the whole, voters were not acting maliciously toward the other party in their crossover voting, because the voting trends in this primary looked so much like they did in a general election. In a general election voters have no incentive to try to sabotage the other party because there is no second, subsequent election for them to influence. More direct evidence in support of this conclusion comes later, when we examine specific vote choices.
CROSSOVER VOTING AND STATEWIDE RACES
In order to show the conditions under which voters crossed party lines, we can examine voting behavior by party affiliation for the eight statewide offices that were on the 1998 ballot. Table 7.3 shows the amount of crossover voting by registered Democrats and Republicans, as well as how nonpartisan voters cast their votes. The figures show that the highest rates of partisan crossover for any one party occurred among registered Republicans voting for Democrats in the races for Governor and Controller. In each case, about 27 percent of registered GOP voters crossed party lines. These two races illustrate the two most viable explanations for partisan crossover: the appeal of a hotly contested race in the Governor's contest, and the appeal of a popular incumbent in the Controller's race. In both cases, crossover voting was no doubt stimulated because of the lack of any real contest in the other party.
In the Governor's race, the Democratic nomination fight was a highly visible contest between Gray Davis, Al Checchi, and Jane Harman. Although Davis pulled away at the end, the race had been considered close throughout
| Percentage of Other Party Crossing into: | Percentage of Nonpartisans Voting in: | |||
|---|---|---|---|---|
| Democratic Race for: | Republican Race for: | Democratic Race for: | Republican Race for: | |
| Governor | 26.9% | 8.7% | 62.6% | 24.0% |
| Lieutenant Governor | 10.2 | 12.9 | 37.3 | 33.1 |
| Secretary of State | 6.0 | 17.9 | 25.8 | 37.9 |
| Controller | 27.0 | 5.2 | 54.1 | 16.8 |
| Treasurer | 11.1 | 9.5 | 40.0 | 28.5 |
| Attorney General | 15.3 | 12.3 | 40.5 | 29.8 |
| Insurance Commissioner | 10.4 | 17.6 | 34.9 | 35.5 |
| U.S. Senator | 12.1 | 15.4 | 40.0 | 44.4 |
In the Controller's race, Kathleen Connell clearly benefited from incumbency, facing no heated competition from within her own party. On the other side, there was no contest among Republicans to draw the attention of voters. Thus, Republican voters who felt Connell was doing a fine job were free to cross party lines and vote for her. There is no way to tell conclusively from these data to what degree name recognition alone was a factor, but given the lack of a campaign on the Republican side, one can reasonably suspect that it was strong. To a lesser extent, crossover voting in the Secretary of State's race was also skewed, this time to the Republican side. This is likely a result of the same phenomenon, as Bill Jones was a well-known Republican incumbent, while Democrats had no nomination contest.
The figures for nonpartisan voting support these patterns and conclusions. Almost two-thirds of nonpartisans voted for a Democrat in the gubernatorial race. These voters decided to weigh in on the Democratic side
| Percentage of Other Party Crossing into: | Percentage of Nonpartisans Voting in: | |||
|---|---|---|---|---|
| Democratic Race for: | Republican Race for: | Democratic Race for: | Republican Race for: | |
| NOTE: Figures are percentages in Democratic area minus percentages in Republican area. | ||||
| Governor | +6.6% | −5.2% | +5.6% | −2.1% |
| Lieutenant Governor | +4.0 | −12.1 | +7.9 | −11.0 |
| Secretary of State | +1.3 | −6.2 | +9.7 | −7.7 |
| Controller | −3.1 | −2.6 | +3.9 | −0.4 |
| Treasurer | +2.5 | −4.7 | +4.7 | −3.6 |
| Attorney General | +3.5 | −6.7 | +6.3 | −4.8 |
| Insurance Commissioner | +2.3 | −8.4 | +8.8 | −5.7 |
| U.S. Senator | +4.8 | −9.7 | +17.4 | −20.6 |
Information effects may not be confined to individual races; they may also affect the overall rate of crossover voting. We hypothesized that a registered partisan, living in a district dominated by a different party, might receive more information about that party and be more inclined to cross into it. As an example of this, table 7.4 compares crossover voting rates in the statewide races in two congressional districts, one heavily Democratic (but nonminority) and the other heavily Republican. The figures reflect the percentage difference between crossover voting in the Democratic and Republican districts. They show that Republicans were more likely to cross over if they lived in the Democratic district; Democrats were more likely to cross over when they lived in the Republican district. The right side of the table, meanwhile, gives the difference in crossover among nonpartisans. It is clear that these voters were more likely to vote in accordance with the partisanship of the district in which they lived, probably because they were more exposed to information about these campaigns.
ELECTORAL SCENARIOS: CROSSOVER IN DISTRICT ELECTIONS
By examining specific district-level races, we can gain further insight into when and why voters cross party lines. There were a variety of electoral scenarios spread among the dozens of district races in 1998. For example, some had contested races in only one party, some had incumbents facing challengers, while others had incumbents running unopposed, and so forth. In this section, we set up some of those various scenarios as comparative cases, and use them to examine the electoral conditions under which voters cross party lines. We use three types of district-level offices in the analysis: U.S. House, State Assembly, and State Senate.
Races with Incumbents
We begin with table 7.5, which explores the drawing power of an incumbent and shows when partisans are willing to cross party lines in order to vote for one. Included here are all districts in which an incumbent was running for renomination in one of the major parties. The behavior of registered Democrats and Republicans is considered in the table. Significantly, when one party had an incumbent on the ballot, and the other party offered a competitive race for its nomination (defined here as a final margin between candidates of 10 percent or less), a full 30.8 percent of that party's registered members still crossed over to support the incumbent. This stands as a testament to the power of incumbency in district elections, at least in comparison to the drawing power of a competitive race. Nearly one-third of registered partisans left their own party even though a battle for the nomination raged within it—a battle that their vote could have helped decide.
Name recognition, especially given the long, pooled list of candidates on the 1998 ballot, may have played a role in this result. It is also possible that voters saw the contested race in their own party as meaningless anyway, because the eventual winner would probably lose to the incumbent in the general election. However, such reasoning would still not directly compel a partisan voter to cross the party line: why not put up your party's best candidate, anyway? The preference for the incumbent over all others was therefore most likely a sincere choice. The number of voters crossing over when their own party offered no real contest for the nomination was roughly identical at 27.1 percent, and more than half of all registered partisans defected to vote for the incumbent when their own party offered no candidate at all. This result probably reveals as much a cause as an effect: incumbents in these districts were probably seen as so widely popular that no one from either party bothered to mount a challenge.
This strong drawing power for incumbents will certainly be seen by some
| Other Major Party Offered: | |||
|---|---|---|---|
| Competitive Race | No Viable Contest or One Candidate | No Candidate | |
| NOTE: Figures are percentages of voters casting ballots in the race. Districts used include U.S. House, State Senate, and State Assembly. | |||
| Mean percentage crossing to vote for incumbent | 30.8% | 27.1% | 56.7% |
| N of cases | (2) | (22) | (7) |
There is potential for an indirect incumbency advantage to emerge from this scenario as well. Incumbents, helped by crossover voting, could increase their share of the total primary electorate and perhaps even collect more than 50 percent of it. That could translate into easier fund-raising for the general election: the incumbent could trumpet these results to potential contributors as a sign of a coming victory in the general election, and by so doing gather even more money for the November contest. If this were to occur, the other party's nominee would need to gear up for a general election in which he was not only pitted against an incumbent, but against one with an even greater monetary edge.
Competitive Race
When one party offers a competitive race for its nomination while the other does not, we expect voters to gravitate toward it: their votes are more meaningful
| Other Major Party Offered: | |||
|---|---|---|---|
| No Incumbent | Incumbent | Competitive Race | |
| NOTE: Figures are percentages of voters casting ballots in the race. Districts used include U.S. House, State Senate, and State Assembly. | |||
| Mean percentage crossing to follow the action | 41.7% | 6.7% | 16.0% |
| N of cases | (8) | (2) | (1) |
In table 7.6, we consider districts where voters had the chance to cross over into such a race. It clearly shows that voters did seize the opportunity to vote in competitive races—except when an incumbent was present. In openseat districts where both parties offered candidates but only one of those parties had a competitive race (again, defined as one in which the final vote margin was less than 10 percent), 41.7 percent of registered partisans left their own party to join the action in the other. Yet, when voters were faced with a choice between an incumbent in their own party and a competitive race in the other, they stayed put: only an average of 6.7 percent crossed party lines. This is the flip side of the effect shown in table 7.5. Incumbents were clearly bigger draws than competitive races when put head-to-head. In only one case did voters see a heated race in each party in an open-seat district, so it is hard to draw generalities from this. However the number of voters crossing in that district was but a sliver of those seen crossing in other districts; a contest in their own party kept partisans largely at home in this circumstance.
Strategic Sabotage?
Next we assess whether voters undertook strategic attempts to raid, or sabotage, other parties. The absentee ballots do not allow us to study candidate preference rankings such as those used by Sides, Cohen, and Citrin (chapter 5) and Alvarez and Nagler (chapter 6; both in this volume) to assess raiding, so we infer motivation from actual vote choices and the strategic context in which they took place. To do this, we compare the actual candidate choices made by registered partisans when they crossed into the other party with the vote choices of the party's own faithful. We make these comparisons in races where conditions were ripe for strategic sabotage to
| District | Party of Incumbent | Percentage of Other Party's Registered Voters Crossing Over ("Raiding") | Percentage of Partisan "Raiders" Voting for Incumbent | Incumbent's Support in Own Party | Incumbent's Margin |
|---|---|---|---|---|---|
| NOTE: Figures are percentages of all voters casting ballots in the race. | |||||
| HR 26 | D | 84.0% | 80.4% | 74.9% | 33.3% |
| HR 28 | R | 32.7 | 95.5 | 87.6 | 88.8 |
| HR 37 | D | 32.8 | 83.9 | 76.5 | 41.2 |
| HR 38 | R | 39.1 | 91.9 | 82.8 | 72.2 |
| HR 41 | R | 31.7 | 36.5 | 23.3 | −23.9 |
| AD 55 | D | 30.2 | 41.3 | 56.1 | 19.7 |
| AD 57 | D | 35.2 | 46.7 | 63.7 | 39.6 |
The data in table 7.7 indicate, however, that the behavior of crossover voters depends more on the incumbent's overall standing in the district than it does on any sinister strategies. Note that in the first four districts listed, overwhelming numbers of the crossover partisans voted for the incumbent in the opposite party. In these districts, the incumbents were strong in their own party as well, drawing huge numbers of their own partisans and winning by large margins. In districts where incumbents were on the ropes in their own party, however, they did not get the support of crossover partisans. The three districts listed at the bottom of table 7.7 (HR 41, AD 55, and AD 57) had incumbents with much lower levels of support among voters in their own party, and the levels of support from crossover partisans largely mirrored this trend. There were no districts in which an incumbent was the overwhelming choice of his own party while crossover partisans rushed in en masse to back a different challenger.
All this amounts to more evidence that widespread raiding simply did not occur. When an incumbent was popular in a district, that popularity was sweeping and crossed party lines; when an incumbent was in trouble, he or she was in trouble everywhere.
| Countywide | Minority Districts | |||
|---|---|---|---|---|
| Fong | Issa | Fong | Issa | |
| Democrats (19.3%) | 51.9% | 33.4% | 56.2% | 27.9% |
| Nonpartisans (8.0%) | 59.2 | 26.2 | 67.4 | 18.7 |
| Minor party (1.4%) | 42.3 | 37.7 | 48.8 | 26.4 |
| Republicans (71.3%) | 44.6 | 40.8 | 46.7 | 37.5 |
| Overall sample (100.0%) | 47.1 | 38.2 | 51.9 | 32.3 |
THE CONSEQUENCES OF CROSSOVER VOTING
The electoral consequences of crossover voting depend very much on whether the preferences of partisans and crossover voters match. If they do not, it is possible that a blanket primary will produce different nominees than a closed system would have, even with all voters voting their sincere preferences.
One instance where this may have occurred was in the Republican U.S. Senate contest between Matt Fong and Darryl Issa. Table 7.8 shows how the votes were split among the different sets of voters who selected one of those candidates. Note that while Fong held a slight plurality among registered Republicans, 44.6 percent to 40.8 percent for Issa, registered Democrats who made a choice between the two were solidly in favor of Fong, 51.9 to 33.4 percent. Nonpartisans showed an even larger discrepancy. In the last two columns of the table, these figures are shown only for congressional districts with nonwhite majorities (districts 30–35 and 37). In these districts, the discrepancy between Republican and other voters was even larger.
Because this analysis is relegated to one county, we cannot prove that crossover voting caused Matt Fong to win his party's nomination. Yet these data do illustrate that there can be large differences between the preferences of a party's own voters and the preferences of voters who cross into that party's nomination contest. This indicates that the switch to the blanket primary format has the potential to change outcomes.
At the district level, the results were often similar. There were sometimes substantial differences between those in a party and those outside of it, and most of these were large enough to have swung the outcome of a close election. However, the difference was never large enough in these instances to have swamped the preferences of a major party and throw the nomination to a Democrat or Republican who did not have a plurality of support
| Minor-Party Members | Nonpartisans | Democrats | Republicans | |
|---|---|---|---|---|
| NOTE: Figures are percentages of all votes cast in minor-party contests. | ||||
| Assembly | 10.1% | 13.9% | 45.5% | 30.5% |
| U.S. House | 8.4 | 14.6 | 46.6 | 30.4 |
| U.S. Senator | 13.9 | 15.3 | 45.9 | 24.9 |
| State Senator | 8.5 | 15.2 | 43.2 | 33.0 |
| Governor | 13.8 | 15.0 | 40.5 | 30.7 |
| Lieutenant Governor | 10.5 | 14.8 | 51.6 | 23.1 |
| Secretary of State | 14.6 | 8.9 | 51.8 | 24.7 |
| Controller | 11.0 | 14.4 | 40.7 | 34.0 |
| Treasurer | 11.0 | 14.7 | 45.2 | 29.1 |
| Attorney General | 7.1 | 10.5 | 52.7 | 29.6 |
| Insurance Commissioner | 9.2 | 14.3 | 41.4 | 35.2 |
CROSSOVER INTO MINOR-PARTY CONTESTS
Minor-party candidates also appear frequently on the California ballot, and they may potentially be the most affected by the consequences of crossover voting. Discrepancies between the preferences of voters in and outside a party can readily have consequences for minor parties. As these parties draw very few total votes relative to the major parties, it takes far fewer crossover votes to swing an election within them. With this in mind, table 7.9 shows that the bulk of votes cast in a given minor party's race did not always come from voters registered in that party. In this election votes for minor-party candidates came very heavily from outside the minor parties. Only 13.8 percent of the minor-party vote for Governor, for example, came from registered members of a minor party, while most of it came from registered Democrats and Republicans. The pattern is very similar for all the statewide offices. This is not a concern when a minor-party candidate is uncontested, as was the case with most of the minor parties and most of the offices. However in some instances, such as in the Peace and Freedom party's contest for Governor, there were two candidates competing. It is entirely possible
CONCLUSION
One of the objectives of the blanket primary was to allow voters to split their tickets. It is very clear that voters took much advantage of that opportunity. The overall pattern of voting in this primary was in fact comparable to a general election. Voters seemed to treat this primary as if it were simply round one in a two-round election: they selected candidates of either party, just as they do in November. We found no evidence that crossover voting involved deliberate strategic attempts by voters to raid the opposite party by aiding its weakest candidate.
When registered partisans supported candidates from outside their own party, their voting patterns indicate that they did so for one of two reasons: to vote for an incumbent from another party, or to vote in a competitive race—especially when there was no contest in their own party. Both actions have potentially serious electoral consequences. The former could give incumbents an additional advantage, not just in the primary contests but also in fund-raising for the general election. The latter certainly has the potential to swing a close primary election.
A contest held under a blanket primary system is not assured the same outcome as would have occurred under a closed primary: we showed that the vote choices of those registered within a party can vary substantially from those of voters who cross over. This crossover effect may be of particular concern for minor parties, whose low vote totals make it easy for crossover voters to affect a nomination if the race is contested. In fact, so much of the minor-party vote came from crossover voting that this effect could be greatly magnified. The major parties are surely susceptible to this influence as well. It is entirely possible that with enough crossover voting, or a close enough election, a different nominee could emerge under blanket primary rules than would have under a closed primary system. We cannot say for certain that it happened in 1998, at least not in the areas studied here—but it could undoubtedly happen in some other election held under blanket primary rules. The potential may rise higher still as candidates learn from California's experience with this primary and become more savvy at targeting voters from other parties.
This potential change in outcomes is undoubtedly what the proponents of the blanket primary wanted to see happen, in the hope that the winning nominees would be more moderate. However, this also confirms some of the worst fears of the blanket primary's critics, as the ability of a party and its members to choose their own nominees could most certainly be influenced
APPENDIX: DATA AND SAMPLES
Through 1998 Los Angeles County used a punch-card tabulation system that could save digital images of each absentee ballot as it was processed. (The system has since been retired.) The Registrar of Voters was kind enough to make the absentee ballot images from the primary available to us for this project. These images, however, were stored by the computer only in raw hexadecimal form and had to be decoded back into voters' actual punches, while accounting for the candidates' positions rotating on the ballot. This extensive process was accomplished with a computer program that one of this chapter's authors, Anthony Salvanto, developed for the task. The authors are grateful to Vern Cowles and the staff at the Los Angeles County Registrar's Office, who made the images available to us and without whom this study would not have been possible. We are also indebted to Bill Detlof for his technical advice and to Chad Rosenberg for his programming assistance and tireless work during the decoding process.
REFERENCES
Bowler, Shaun; Todd Donovan; and Trudi Happ. 1992.
“Ballot Propositions and Information Costs: Direct Democracy and the Fatigued Voter.”
Western Political Quarterly45: 559–68. 
Darcy, R., and Anne Schneider. 1989.
“Confusing Ballots, Roll-off, and the Black Vote.”
Western Political Quarterly 42: 347–64.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.
Jacobson, Gary. 1992. The Politics of Congressional Elections. New York: HarperCollins.
Vanderleeuw, James M., and Glenn H. Utter. 1993.
“Voter Roll-off and the Electoral Context: A Test of Two Theses.”
Social Science Quarterly74: 664–73.
| Contest | Percentage of Popular Vote in Absentee Sample | Percentage of Popular Vote in County |
|---|---|---|
| Governor | ||
| Davis | 34.0% | 36.9% |
| Harman | 11.4 | 12.0 |
| Checchi | 14.5 | 18.1 |
| Lungren | 30.6 | 27.8 |
| Lieutenant Governor | ||
| Bustamante | 37.9 | 43.3 |
| Hentschel | 15.0 | 14.2 |
| Leslie | 10.9 | 9.1 |
| Mountjoy | 15.3 | 12.2 |
| Secretary of State | ||
| Alioto | 41.9 | 45.1 |
| Jones | 49.7 | 44.0 |
| Controller | ||
| Connell | 65.1 | 65.3 |
| Barrales | 29.5 | 27.9 |
| Treasurer | ||
| Angelides | 26.0 | 28.2 |
| Robles | 22.7 | 23.2 |
| Goldsmith | 14.2 | 12.0 |
| Pringle | 27.5 | 24.8 |
| Attorney General | ||
| Calderon | 13.9 | 17.3 |
| Lockyer | 25.3 | 25.5 |
| Schenk | 10.2 | 12.3 |
| Capizzi | 13.6 | 12.8 |
| Stirling | 33.3 | 29.5 |
| Insurance Commissioner | ||
| Brown | 21.3 | 22.6 |
| Martinez | 26.5 | 29.6 |
| Quackenbush | 46.5 | 40.3 |
| U.S. Senator | ||
| Boxer | 51.3 | 52.0 |
| Fong | 21.6 | 18.9 |
| Issa | 17.5 | 16.5 |