This paper presents two arguments. One argument is that the 1968 election bears witness to the importance of issues at critical points in political history. American surveys show that issues vary greatly over time in their total impact on elections. From 1936 to 1948 the question of the role of the federal government had a major effect on the flow of voters between the parties. [1] In contrast, studies conducted during the rather apolitical period of the 1950s demonstrated both a lack of ideological structure in people's beliefs and a marginal impact of issues in influencing their votes.' With the Goldwater and Wallace insurgencies of 1964 and 1968, issue became again important components of electoral decisions' This was particularly true OF Vietnam, urban unrest and race, social welfare, and Johnson's performance as president--the dominant issues of 1968.A second argument is that issues having a similar impact on the election outcome can have very different consequences for popular control of public policy. On some issues the electorate exercises no effective constraints on leader's policy choices. On others, the electorate permits political leaders a wide array of options when a policy is adopted but passes a retrospective judgment on the policy in subsequent elections. On still other issues, the public severely limits the options of leaders at the time policy is made. Race and public order, it is argued, are examples of issues on which the electorate places the most significant restraints on the policy initiatives of political leaders.

Applying the concept of a normal vote to a political issue requires a calculation of normal (expected) votes for each group of persons holding different positions on the issue. Computing these expected votes is quite straightforward: [6]

1. Responses to the issue are cross-classification against the Survey Research Center measure of party identification.
2. The scores (+2, +1, 0, -1, - 2) are assigned to the five classes of identifiers (Strong Democrat, Weak Democrat, Independent, Weak Republican, Strong Republican).
3. From the scores, a mean party identification score is calculated for the distribution of party identifiers within each response category of the issue item.
4. The obtained mean (M ) is then inserted into the prediction equation:

As the equation makes clear, the only factor that enters directly into the calculation of an expected vote for a group of people is the distribution of party identifiers within it. This equation, while cheap in data costs, does presume that the expected Democratic vote is the same for all subgroups of the electorate having. the same distribution of party identifiers. This assumption is not as questionable as it might first appear. Shanks has begun an extension of Converse's procedures, adding to the prediction equation a more theoretically pleasing set of variables. Since Shanks's estimates of the expected Democratic vote for several subgroups are almost identical to those of Converse for the presidential elections of 1952-1964, I have used Converse's more parsimonious equation. [7]

The utility of the normal vote calculation is that it permits the analyst to separate issues into their partisan (long-term) components and their short-term components. Consider Figure 1 as an example of a normal vote analysis of an important issue in 1968 the speed with which civil rights leaders have been pushing their cause.

Party identification is both stable over the lifetime of most votes and a major determinant of their electoral choices. [9] Therefore, an issue on which mass party members are polarized sinks its roots into a stable, long-term division in the electorate. Knowing the balance of party loyalties of people holding some new on a particular issue, one can predict how these persons would vote in a normal election. These predictions are called Expected Democratic Votes (EDV) .

The partisan component of an issue is manifest in changes in the expected Democratic proportion of the vote across issue positions--the greater the changes, the more partisan arc people's attitudes on the issue. In Figure 1 the balance of party identification among persons believing that civil rights people have been pushing too slowly so favors the Democrats that the Democrats would expect to receive 68 per cent of the votes in a normal election. In contrast among those who believe civil rights people are pushing too fast, the Democrats would normally expect to receive 54 per cent of the vote. This difference of 14 percentage points (68 minus 54) reflects the relative degree of polarization of the mass parties on the issue.

(2) A second aspect of poleration is the proportion of people advocating each of the issue positions. Polarization is born of conflict of opinions; as consensus is approached, an issue ceases to divide the parties. In Figure 1, nine times as many people think civil rights people are pushing too fast as think they are pushing too slowly. In Stokes's terms this is almost a valence issue, an issue on which there is no disagreement; [10] as such, it cannot be a partisan issue.

The Short-Term Component of a Civil Rights Issue

What can we conclude from the observation of the deviations from the normal Democratic vote? Certainly not that the issue was causally related to the vote or that the issue made some net contribution to the observed Democratic vote. The reason is that from survey data it is difficult to determine with precision the issues (if any) that lead each person to cast his vote for a particular party. Lacking a method of categorizing the electorate, normal vote analysis cannot ascertain the net gain or loss of an issue to a candidate.

What we can say is this: For respondents in this sample the distribution of party identification is such that the Democrats would expect to receive 57 per cent of a normal vote. [11] In fact Humphrey received 41 per cent, 16 percentage points below a normal Democratic vote. If Humphrey had lost this same 16 percentage points across every issue position, then this issue would not even be statistically related to short-term defections from the normal vote. Such is not the case with the civil rights issue in the previous example. Humphrey exceeded his expected vote among some civil rights groups and fell short in others. The variation of the actual gains and losses; each issue category (+2, -10, and-21) from the overall deficit (16 points) is a measure of the issue's strength of association with short-term defections from the normal vote. Thus, even if we cannot speak in causal terms of an issue's contribution to the vote, we can compare issues statistically in terms of their association with long-term party loyalties and short-term defections from the normal party vote. Appendix II provides the formulas for such comparisons.

With this explanation of normal vote analysis, we can now decompose a set of 15 issues in the 1968 SRC election study and compare them in terms of their long and short-term components. This set of issues falls into four broad groupings: Vietnam, social welfare, urban unrest and race, and civil liberties and public protest. A final issue orientation, the electorate's evaluation of Johnson's performance as president; does not fit neatly into any of these categories; it will be considered separately.

The 1968 election study includes two very useful items probing attitudes toward the Vietnam war: escalation vs. de-escalation, and immediate withdrawal vs military victory. Both reveal similar patterns, illustrated by Figure 2. A presentation of preferences for escalation vs. de-escalation.

As one would predict on a matter of foreign policy, there are no long-term, partisan aspects of people's attitudes about Vietnam. [12] Republicans and Democrats are almost equally likely to endorse solutions raging from pulling out of Vietnam entirely to taking a stronger stand even if it means invading North Vietnam. This is reflected in the similarity of the expected Democratic votes in the three categories of policy alternatives. [13]

In contrast, the short-term component of at attitudes on escalation vs. de-escalation is evident. Among those who wanted to pull out entirely, Humphrey's vote is 12 percentage points lower than expected. Among those who would consider invading North Vietnam (34 per cent of the total sample), Humphrey's vote is almost 24 percentage points below its expected level. Humphrey's best showing is among the advocates of a policy of keeping American soldiers in Vietnam while trying to end the fighting. This one might interpret as a status quo position, a position with which the public apparently identified Humphrey because he was a member of an administration pursuing a similar policy. [14]

Obviously, Humphrey fared more poorly among the proponents of escalation than he did among those of de-escalation. His showing was weakest among the advocates of escalation, who outnumbered the advocates of de-escalation by nearly 2:1 among the voters. In addition, the proponents of escalation gave both Nixon and Wallace the greatest proportion of votes they received from any of the other opinion groups (51 and 18 per cent respectively).

Figure 1, the example used earlier to explain the concept of a normal vote, illustrates the loss of Democratic support among people who thought that civil rights leaders were pushing their cause too swiftly. The same voting pattern is manifested by people who advocated the use of all available force to quell urban unrest. Figure 5 presents a normal vote analysis of people who were asked to place themselves upon a seven-point continuum concerning alternatives for dealing with the urban problem. One end of the continuum is demarcated by "correct the problems of poverty and unemployment that give rise to the disturbances;" the other, by "use all available force to maintain law and order---no matter what results." [15] As Figure 5 demonstrates, Humphrey captured a substantial majority of the votes of those endorsing a "correct-the-problerns" approach. In contrast, he suffered massive losses (34 percentage points) among the advocates of the polar position of force as the solution. Humphrey lost the bulk of these normally Democratic votes to Wallace rather than to Nixon. Nixon pulled his strongest support toward the widrange of the scale, while Wallace's strength peaked at the polar right of the continuum. Among the people of the polar right, Wallace captured so many normally Republican votes that Nixon also fell short of his party's expected strength.

The 1968 SRC election study is rich in items on racial attitudes. In addition to the previous items on speed of movement on civil rights and urban unrest, other questions probe attitudes on segregation in general, attitudes on residential segregation in particular, and perceptions of Negro violence. Since all yield a consistent result under normal vote analysis, Figure 6 presents a composite picture of the vote on racial issues. Each line is a plot of the deviation of the expected Democratic vote front the observed Democratic vote.

In sum, in an election in which many issues held the publics' attention, race and urban attitudes were indisputably important In competition with a number of salient issues, urban un-rest ranks second, belief in segregation ranks fourth, and speed of movement on civil rights ranks sixth in the magnitude of their relationship to defections from the normal vote. Herein may be read the Wallace story. As Converse and others note, in both the North and the South, the Watlace vote came predominantly from Democrats. Nixon rather than Humphrey was the second preference for most of these Democratic Wallace voters. For a third it was at least their second consecutive defection from their national presidential ticket. [16]

In the South racial issues dominated all others in 1968. The items on belief in segregation. speed of movement on civil rights, and urban unrest rank first, second, and third, respectively, in their relation to short-term deviations from the normal party vote. [17] Despite that fact the South also leads the nation in party polarization on racial issues, with Southern Republicans being located farther to the right of Southern Democrats than Northern Republicans are to the right of Northern Democrats.

The 1968 election study includes a series of three items concerning the legitimacy of public protests. With respect to each form of protest the respondents are asked to state whether or they approve its use. The series begins with "protest meetings and marches that are permitted by the local authorities" and ends with "sit-ins, mass meetings, and demonstrations," presuming "all other methods have failed." Figure 7 presents a normal vote analysis of these attitudes.

Surprisingly, the relationship of these issues to the vote is marginal--marginal at least in comparison with others. Of the 15 issues analyzed here, the items for legal meetings and marches, civil disobedience, and mass demonstrations rank respectively twelfth, fourteenth, and fifteenth in the magnitude of their relationships to defections from the normal vote. As one can see from Figure 7, Humphrey fared somewhat better among those who approved of public protest than among those who did not. In the main, however, he lost consistently across an categories of approval and disapproval of protest methods (the gap between the expected Democratic vote [ED] and his observed vote [OD]. The implication of this uniform loss is that other issues were responsible for his defeat

The failure of civil liberties to surface as a short-term in= in 1968 parallels the fate of internal communism in the early 1950s. In 1952 the Republican party attacked the Democrats with the slogan of ---Communism,Corruption and Korea." Yet, Campbell et al. found that only three per cent of the population mentioned in their evaluations of the candidates and parties the argument that the Democratic administration had been "soft on communism." [20] Similarly in a survey conducted in 1954, Stouffer reported that less than one per cent of the population said they were worried about the threat of Communists in the United States. [21] This evidence from surveys is supported by Louis Bean's analyses of aggregate voting returns and Joseph McCarthy's campaigning. Bean concluded that McCarthy's campaigns in 13 states outside the South in 1952 had no effect on the presidential contest and, if anything, hurt the Republican senatorial candidates he had endorsed.[22]

In all likelihood, the reason that these civil liberties and protest items are unrelated to defections from the normal vote is that they have rarely if ever been partisan issues. Democrats are slightly more willing to tolerate these acts than are Republicans. On the whole, however, the public speaks with a single voice against the protests. Not even 10 per cent of the college educated--the principal supporters of tolerance of civil liberties-approve, if asked, of mass demonstrations. Since attitudes on these valence issues are not linked to either parties or candidates, their impact does not extend to the vote itself, though the political system surely feels their force in other ways.

Figure 8 underlines the Iong-term, partisan nature of these federal role issues. Again, the measure of the partisan component of an issue is the degree of change in the expected Dernocratic vote across various positions on the issue (line ED in Figure 8). For the four issues, the difference in the expected Democratic vote between those who would expand the federal government's role and those who would not averages almost 20 percentage points. From a 1958 sample, McClosky observed that the mass membership of the two parties did not differ on issues." This was clearly not the case for federal role issues a decade later.

One might predict the partisan component of these traditional issues; more surprising is their uniformly high relationship to defections from the normal vote. Note, in Figure 8, Humphrey's consistent losses among voters endorsing a limited role for federal activities. In every case the difference between the expected (ED) and the observed Democratic vote (OD) exceeds 20 percentage points. in contrast, Humphrey did quite well among those supporters of a broad scope of federal activities barely falling short of his expected strength and, in every case, capturing a comfortable majority of the vote. In an election in which race. law and order, and the war loomed on the scene, the more traditional issues seemed reluctant to leave quietly.

The presence of the short-run component of the federal role issues suggests the possibility of spurious association. The conservative positions on these federal role-social welfare issues are associated with four other positions: opposition to civil rights, use of force in dealing with urban unrest escalation of the war, and a harsh evaluation of Johnson's performance as president. [24] All of the latter issues were highly related to short-term defections from the normal vote in 1968. To check against the possibility that the federal role issues manifest a shortterm component only because of their association with the others- separate normal vote analyses of the federal role issues were run within categories of the other issues. Uniformly, the short-term component of the federal role issues remained as strong as in the uncontrolled case.

In sum, the traditional set of attitudes relating to the scope of the federal government persisted as a significant force in the 1968 election. In an election supposedly dominated by urban unrest and the war, why was this so? Perhaps one clue lies in the votes that Humphrey did not lose rather than these he did.

In late September, Humphrey's cause looked hopeless. A Gallup poll showed him trailing Nixon by a margin of 4-3, to 28 per cent and leading Wallace by only seven percentage points. Scarcely a month later, he lost the election by less than one per cent of the popular vote. As Theodore White chronicles the last month of the campaign, Humphrey attacked the Republicans with an echo of the 1964 campaign, the alleged Republican indifference to the traditional issues of social security and social welfare. [25]

Meanwhile, the AFL/CIO had mounted a two-pronged assault on Wallace and Nixon. First, the AFL/CIO carried a message--Wallace and Nixon were no friends of labor. Second, COPE took over much of the effort of mobilizing the eleclorate, particularly the Black vote. In White's words,

"The dimension of the AFL/CIO effort, unprecedented in American history, can be caught only by its final summary figures; the ultimate registration, by labor's efforts, of 4.6 million voters; the printing 'and distribution of 55 million pamphlets and leaflets out of Washington and 60 million more from local unions; telephone banks in 638 localities using 8,055 telephones manned by 24,611 union men and women and their families; some 72,225 house-to-house canvassers; and, on election day, 94.457 volunteers serving as car-poolers, materials-distributors, baby-sitters, poll-watchers, telephoners." [26]

A cross-sectional sample such as this one cannot establish who changed their minds during that last mouth of the campaign. Yet the appeals of Humphrey were the traditional ones of the old Democratic coalition. Herein may rest the importance of the short-term effects of social welfare-federal role issues. Humphrey and organized labor used-them to counter the new issues---successfully so among those Democrats for whom the traditional social welfare issues remained important. [27] In this respect the trend in 1968 bore a notable similarity to Truman's late surge in 1948. In their Elmira study Bcrelson et al. document the basis of Truman's "Fair Deal Rally." They found the swing to Truman to be sharpest among those for whom class issues were most salient. [21]

People who characterized Johnson's performance as "very good" gave Humphrey even more than his expected share of the vote. From that point Humphrey's vote plummeted in an almost linear fashion as the severity of judgment of Johnson mounted. From those who judged Johnson's performance as "poor", or "very poor," Humphrey received only a small fraction of the normal Democratic strength. Of an the issues considered in this analysis, this one bears the greatest relationship to defections from the normal vote. Humphrey's fate may represent a case of a more general phenomenon--that the unpopularity of leaders transfers to their successors more readily than does their popularity. For example, those who thought well of Johnson's performance out numbered those who thought poorly of it by a comfortable margin. However, Johnson's admirers gave Humphrey only a normal Democratic vote, while his detractors voted heavily against him. Indeed. it may be a contribution to party responsibility and competition that blame's retribution survives credit's reward. [30] Parties are more likely to be sensitive to public opinion, if they believe popularity dissipates quickly, while political blunders endure in the public mind. In the process of governing, the party in power collects enemies faster than friends. In this vein the authors of The American Voter rely on the persistence of negative attitudes to explain the inexorable decline of majorities. [31] Thus, we have an explanation of the restoring forces that seem to create an equilibrium among competing American parties over time. [32]

In a more speculative vein, it is interesting to ponder the second major inquiry of the paper--the relationship of public attitudes to popular control of governmental policy. The problem is posed by V. 0. Key's classic question: If governmental policy corrodes the vitality of a society, who is the villain--the public or its political leaders? [32] To conclude that it is the public requires evidence, Key argues, that the public severely limits the discretion of policy makers. Yet, from survey data Key notes that substantial portions of the public are unconcerned and uninformed about most concrete policies of government. Whatever limits opinion fixes, governments still enjoy wide latitude in the determination of whether to act, when to act, and what exactly to do. Thus Key concludes, leaders cannot excuse their actions by pleading the hampering restrictions of public opinion.

Key's formulation of the problem of responsibility for policy has its difficulties. One is the factual problem of distinguishing different degrees on a continuum of popular constraints on leaders' policy options. The difficulty of this judgment is lessened if the continuum is divided into three broad categories: (1) issues on which people exercise no important constraints and thus, have no voice; (2) issues on which people have voice, but permit political leaders a wide array of policy options; and (3) issues on which voters do severely restrict the options of leaders.

Regardless of the extent of popular control over policy, a second, ethical issue remains. If we can agree that the public should not be held responsible for policy when leaders possess an array of options (categories 1 and 2), what of the case in which leaders' options are severely restricted? Many would surely argue that leaders are morally responsible for public policies no matter how much the public seems to force their hands. 1 will not take up this issue except to assert that severe public constraints on policy options is a necessary condition for holding the public responsible for a policy. Whether or not that condition is sufficient remains problematic. Caveats considered, let us turn to examples in the 1968 election of the three categories of popular control over policy.

(1) On many, perhaps most, governmental policies, the public as an electorate has no voice. On these policies, only the few have influence because the many have no interest and no information. Issues that do not impinge upon people's economic needs or ego defenses fall into this category, as do, no doubt policies of abstruse complexity. Foreign policies in peace time strategic decisions, esoteric regulatory rules--these sorts of issues are not likely to engage popular attention even though they may have a substantial impact on people's lives in the long run. Certainly political leaders must bear full responsibility for failures in policy in these areas.

(2) On other issues the public does find its voice, but it still leaves political leaders with so many options that leaders retain responsibility for the success of their choices. This interpretation of the relationship of popular preference and public policy is consistent with the views of Key and many other voting analysts. [34] In the main, policy initiatives lie with the leaders, not voters. After a policy is enacted, voters make a retrospective evaluation of its success. The electorate's influence over such a policy is the leaders' anticipation of voters' reactions in the following election, [35] when voters will ask themselves, "How did the party in office do, and how likely is it that the opposition party would have done better?" An archetypal example is Humphrey's loss of support among voters who ranked Johnson's presidential performance as poor. It is in this sense that Key remarked, "The vocabulary of the people consists mainly of the words 'Yes' and 'no'; and at times one cannot be certain which word is being uttered." [36]

Vietnam may be another example of an issue on which the electorate retrospectively said ..no" to governmental policy.. This. interpretation is necessarily tentative because people make voting decisions across a range of issues. As a result, a majority may vote for a man in spite of his position on an issue, not because of it. [37] Thus, even when one limits the interpretation of the vote to "Yes" and "no," it is difficult to infer conclusively what the judgment is on any particular issue. On the matter of Vietnam policy alternatives, one can only speculate about the nature of constraints posed by past and future elections.

On the eve of the 1964 election, only 20,000 American soldiers were in Vietnam, and President Johnson was campaigning on pledges against fighting Asian wars with American boys. What could the 1964 vote have told him about popular support for various war options? In brief, it could have told him anything he cared to believe.

Figure 10 presents a normal vote analysis of the escalation/de-escalation issue in 1964. The data could be interpreted as support for "pull [ing] out of Vietnam entirely." Johnson received 62 per cent of the votesfrom the proponents of this option--a substantial majority, and three percentage points better than a normal vote would have predicted. On the other hand, the data also show strong support for "keep[ing] our soldiers in Vietnam but try[ing] to end the fighting." From this group he received a startling 82 per cent of the vote, 18 percentage points more than a normal vote would have predicted. Finally, the data also reveal popular support for the option of "tak[ing] a stronger stand even if it means invading North Vietnarn." Although he failed to win the expected Democratic vote within this group, he still claimed a majority. And, it should be noted. the advocates of escalation easily numbered the largest of the war opinion groups. Looking at the 1964 data and noting Pomper's evidence that 1964 Johnson voters were still supporting Johnson's policies in 1966, one must conclude, as Pomper did, that Johnson did not violate an electoral mandate in spite of the fact that he ignored his own campaign pledges. [35] The 1964 election foreclosed none of Johnson's war options. He and other political leaders formulated the escalation policy and rightly should have borne full electoral responsibility for it.

By 1968, escalation had changed the context of the options confronting the electorate. Rather than 20,000 men in Vietnam, there were 530,000. Sixteen thousand men had died, and America's prestige was at stake. As we have seen, the public's reaction to a continuation of the existing war policy was negative. Humphrey won a majority of the votes of people endorsing a centrist status quo alternative: "keep our soldiers in Vietnam but try to end the fighting." He fared less well among those who wanted out entirely; he fared disastrously among those who would escalate the war, a group of voters who gave both Wallace and Nixon the greatest proportions they received from any Vietnam attitude category. (See Figure 2.) The common thread of the views of those who deserted Humphrey--people who would withdraw rapidly or who wanted to sharply escalate the fighting--would seem to be a desire to end the engagement quickly. As in the Korean war, an official Vietnam Policy of a war of limited aims and lengthy duration confronted a public unwilling to support war on those terms. [39]

Though one cannot marshal evidence about events that never happened, the Johnson Administration appeared to lose more electoral support by its pursuit of an extended limited war than it would have lost either by a policy of withdrawal (even after the buildup of men) or by a precipitous escalation of the conflict. This view of the electoral dangers of limited conflict is consistent with the views of Key, Kahn, and Brzezinski and Huntington. [40] The public does not necessarily prescribe a particular policy, but it can proscribe a policy--e.g., the option of extended limited war. Even after the 1968 election, Rosenberg, Verba, and Converse argued that policy makers retained a wide array of options for ending the war.

"The President is not particularly pressured by public opinion to engage in one kind of policy rather than another-to escalate or to de-escalate the war. Rather, the public would support a variety of initiatives. What the public wants is an end to the war. Thus it will judge the policies of the President retrospectively in terms of how successful they are in this direction." [41]

If political leaders possessed a wide array of options in Vietnam, then are there issues on which the electorate does severely limit the options of political leaders? If there are, what must the structure of political attitudes on such an issue be? For this close relationship of opinion to policy to exist, the following conditions are surely necessary:

• A substantial number of people must have intense views about specific policy alternatives.
• People must have information about the views of competing candidates on the issue.
• These attitudes must be fairly impervious to short-run manipulation by political leaders.
• The balance of opinion on the issue must be such that the winning candidate's choice of an issue position significantly affects his probability of being elected.

Clearly, these are necessary, not sufficient, conditions for influencing policy through the electoral process. There are other channels for influencing policy in our system, all with their own conditions.

What contemporary issues might satisfy these four conditions? The issues of race and public order seem the obvious candidates. even in the quiescent period of the late 1950s, there was evidence of a close relationship between opinion and policy on racial attitudes. Miller and Stokes, for instance, found a correlation of .65 between congressional roll-call votes and constituency opinion on civil rights issues. [42] Furthermore, the link between the two was the representatives' inclination to vote their perceptions of their constituencies' opinion: they were not simply voting their own attitudes, which happened to coincide with the majority opinion in their districts. Let us see how the race issue in the elections of 1964 and 1968 might satisfy the four conditions:

1. A substantial number of people must have intense views about specific policy alternatives. Survey questions are rarely framed in a manner that presents reliable evidence on either the direction or the intensity of views on specific policies. One exception in the 1964 SRC election study is a set of items probing knowledge of and support for the 1964 Public Accommodations Law. All but 23 per cent said they knew of the new law (and as shown below, their knowledge of the candidates' stands indicates people were not simply giving the socially desirable answer). Of those having an opinion on the law, 90 per cent said their minds were made up. No such questions were asked in the SRC election studies on the 1965 Voting Rights Law or the 1968 Open Housing Law. Nevertheless, we can be reasonably confident that the controversies over their passage served to crystalize intense views on these policies Negroes voted more than 90 per cent Democratic in 1964 and 1968, a display of cohesion that reflects the interests they saw at stake in the two elections. [43]

2. People must have information about the views of competing candidates on the issue. In 1964 the level of information on the public accommodations law was remarkably high. Of the 77 per cent who said they had heard of the law, 95 per cent knew Johnson favored it and 84 per cent knew Goldwater opposed it. Even in 1968, when differences between the Republican and the Democratic candidates were less striking, voters had little trouble sorting out the candidates' stands on racial issues. Figure 11 presents a distribution of people's perceptions of the candidates' positions on the urban unrest issue. The modal perceptions of Humphrey, Nixon, and Wallace were distinctly left, center, and right, respectively-perceptions that were quite accurate in view of the different positions of the parties and the candidates on issues of race and urban violence. [44] If one limits the sample to those who viewed the race issue as important in determining their vote, the distinctness of the perceived differences among the candidates is even more pronounced.

3. Attitudes on the issue must be fairly impervious to short-run manipulation by elites. Almost every survey of the literature on mass persuasion emphasizes the difficulty of changing popular attitudes when attitudes are well formed and when the time available to the elites is reasonably short (as in a campaign). [45] Certainly, racial attitudes should be among the most difficult for opinion leaders to influence, for they are learned early in life and are often central to people's ego defenses. [46] Thus, we can be reasonably sure that if we discover a close relationship between popular attitudes and the enactment of public policies on race issues, the link cannot be explained away by arguing that political leaders persuaded voters to support policies the voters previously resisted.

4. The balance of opinion on the issue must be such that the winning candidate's choice of an issue position significantly affects his probability of being elected. It is easier to state this condition than to present evidence appropriate to its test. In the first place, the importance of a particular issue in an election differs for each candidate (for example, whether, in Downs' terms, a candidate's victory is based on a "coalition of minorities" or on a "passionate majority"). [47] One illustration of this problem is the difference between the importance of the race issue for Nixon and Humphrey in 1968. Nixon won the election even though he lost more than 90 per cent of the Negro vote. Presumably, Humphrey needed all of those votes to have had any prospect of victory. In the second place, we have no analytical technique for determining the effect of alternative issue positions on each candidate's electoral coalition. [48] Our only approach to the analysis of the fourth condition is an indirect one examining the statistical relationship of beliefs on issues to votes.

In 1964 the relationships between civil rights issues and votes were substantial. The gamma association of the vote with support for the public accommodations law was .57; with the establishment of a fair employment practices Commission for Negroes, .47; with views on federal enforcement for school integration, .37. These links between civil rights positions and votes continued in 1968, as I argued in the first section of this paper.

The position that racial attitudes are closely related to public policy involves an apparent contradiction. If Humphrey's losses in 1968 were statistically related to racial attitudes, why did Congress in that same year pass an open housing law covering 80 per cent of all housing units? One answer is that there is no survey evidence of an increase in the number of white people opposed to traditional civil rights goals such as public accommodations, voting rights, and open housing. [49] There is greater fear of Negro crime and urban violence. This attitude is mirrored in the fact that the open housing law contains an anti-riot section. [50] Most people apparently believe that Negroes now enjoy legal equality and that the remaining social and economic inequalities are the responsibility of Negroes themselves. [51]

The evidence on racial attitudes and civil rights laws is undoubtedly insufficient to support a contention that Congress scarcely had time to leap to the lead of a mob bent on liberal changes in civil rights laws. On the other hand, it is so commonly assumed that changes in public policy have preceded changes in public opinion on civil rights issues that a second look is warranted. [52] The sum of the evidence indicates that racial issues were voting issues in 1964 and 1968, and that majorities of the voters supported the major civil rights laws passed in those years (though we note the warning of Converse and Schuman that the wording of an item has a major effect on which response category attracts majority support). [53] In my view the following argument merits attention: Congress ran few risks in its support for civil rights laws in the 1960s---not because an apathetic public paid little attention to policy proposals, but because on these issues Congress carefully stayed within the relatively narrow constraints posed by a watchful electorate.

The time to test the argument on existing civil rights laws is gone, for we cannot reconstruct a public past. Yet, determining the linkage between public opinion and policy is so critical a problem that it behooves us to ask policy questions of the public. We need to ask more than general questions about political attitudes. We need to know what people think about particular policies---the levels of information, intensity, and support for specific options. Without succumbing to hopeless naivete about the interest and information of the electorate, the opinion-policy linkage on the issues of race and public order is evidence for the rich payoff of such an approach.

Converse developed the concept of the normal vote in order to interpret the force of the religious issue in 1960. His challenge was to analyze voting changes between 1956 and 1960 without using 1956, a deviating election, as a baseline for the measurement of change. The normal vote, an empirical estimation of an abstract concept, became the baseline, and the impact of the religious issue was measured as a deviation from the normal vote.

The logic of the normal vote presumes that the force of issues constantly changes. This logic has been too often ignored by those who assume the immutability of political behavior. The electorate of 1956 was presented in The American Voter as being little moved by ideology or public issues. But 1956 was a deviating election in a quiescent. political era, and it should be of little surprise to find a minimal amount of issue voting in that apolitical period. The importance of issues in 1968 stands in striking contrast to 1956. Beliefs about Vietnam, race and urban unrest, and Johnson's performance as president were all highly related to the vote in 1968. Time has told us of the unusual nature of the 1956 election. Undoubtedly, the future will reveal the atypicality of 1968, as American parties seemed to be in a stage of realignment or disintegration. Surely we should expect issues to be more important in these times than in periods of stability in party strength. After what appears to be- a transition period, issue voting may once again decline to the level of the 1950s.

A second, more speculative argument of the paper is that issues similar in the magnitude of their relationship to the vote can pose very different restrictions on the discretion of leaders in policy making. On some issues the electorate exercises no effective constraints on leaders' policy choices. On others, the electorate permits political leaders a wide array of options at the time of the adoption of policy, while passing a retrospective judgment on such choices in subsequent elections. It is argued that the issues of Vietnam and Johnson's performance as president are examples of this type of public control of policy. Finally, there may be issues on which the public rather severely limits the options of leaders at the time of the adoption of policy. It is difficult, surely, to amass the evidence necessary to determine which issues could properly be placed into this final category. If any issues of 1964 and 1968 meet the tests, race and public order do. It may be, then, that the electorate must assume final responsibility for the success with which the United States solves its racial problem.