Arkansas Tech University The Culture Wars & Political Polarization in Perspective

relationship between defense spending and defense spending opinion. As defense spending becomes extreme relative to the average defense spending level, the preferences of the American public on defense spending moves in the opposite direction. Contrary to some cynics who believe the public is unresponsive to the actions of their government and budgetary decisions, the conflict over defense spending and the average opinion on defense spending levels indicates that the public is paying attention and adjusting their views dependent upon changes in philosophy, policy and spending.

Figure 7.12 illustrates the bimodality trends in the three ANES social issues. The largest substantive change in bimodality for any of the issues considered in this or other chapters is readily apparent in the kurtosis change on women’s equality. It reflects a profound shift in opinion of the American public on the right of women to fully participate in all the rights and responsibilities that men have traditionally enjoyed, and the consequent implication that government should, at minimum, be gender neutral in its treatment of women and men if not actively working to provide opportunities for women. In the 1970’s the polarization of society on women’s rights was relatively indistinguishable from that on aid to Blacks, with an average kurtosis for female equality sitting below -1 for the decade, just as was the case for aid to Blacks. However, the two trends break off in the 1980’s with aid to Blacks remaining playtkurtic while opinion on women’s equal role in society takes an extreme leptokurtic turn. In 1992, labeled by many in the media as “The Year of the Woman” given the number of successful Senate seat bids made by women that year including Patty Murray of Washington state and Diane Feinstein in California, the kurtosis of opinion on women in society crosses the ‘normal’ distribution threshold. From that point on, there is a steady and steep incline in the trend towards unimodality (consensus) in opinion on female equality public opinion. By 2008, opinion on this social issue is by far the most consensual issue of all the issues analyzed here.

While the shift in opinion on women’s equal rights is remarkable, a more subtle but just as interesting aspect of this trend is its separability from the abortion issue. Proponents of Second-Wave feminism and the women’s rights groups that have organized to lobby on issues related to it, such as the National Organization for Women (NOW), have linked the equal rights of women and abortion (reproductive rights) for some time. This conceptualization of the abortion decision as integral to equal rights for women was adopted by the Supreme Court in the land mark case of Roe v. Wade, where the court applied the right to privacy discovered in Griswold v. Connecticut to the abortion decision. As Justice Blackmun wrote in his majority opinion, “We, therefore, conclude that the right of personal privacy includes the abortion decision…” (Roe v. Wade 1973). However, there is little evidence that the general public has adopted this viewpoint. Indeed, it appears that public opinion on equal rights for women is distinct separable from that of abortion both in terms of the global status of the mean opinion on the two social issues, the dispersion of opinion on the two issues, and the bimodality of opinion on them. Furthermore, they display distinct and opposite trends across the two primary polarization

measures. While opinion on equal rights for women has become very leptokurtic, abortion has evidenced a significant playtkurtic trend (Figure 7.13). While the dispersion of abortion has remained relatively steady over the period (declining somewhat in the 1980’s but recovering to previous levels in the 2000’s), there has been a steady decline over the past three decades in the equality for women ANES item (Figure 7.14). Only in the mean is there a trend that suggests a connection between abortion and equal rights, with the average opinion on abortion becoming more “Pro-Choice” at the same time that there is a significant increase in support for an equal role for women in society relative to men.

Figure 7.13: Bimodality Trend in Mass Opinion on Abortion, 1972-2008


Webster v.

Reproductive Health Services

Planned Parenthood v. Casey

Roe v. Wade

Bush Admin


Figure 7.14: Dispersion Trend in Mass Opinion on Equality for Women, 1972-2008


ERA Ratification Fails

Year of the Woman

However, the average position on abortion remains much closer to the center of the distribution than the equality for women average placement. The average opinion on equality for women has moved significantly towards the extremes of the distribution, with an average of 1.833 on a 7 point scale in favor of an equal role in 2008. The Abortion trend does not display a shift in the same ballpark as that of the equality for women item.

An illustrative example of the separability of abortion and equal rights for women is the campaign to ratify the Equal Rights Amendment to the constitution. When Americans were convinced the ERA was about providing equal rights for women, it enjoyed strong support. For example, in 1975, respondents to a Gallup poll were asked, whether they favored the ERA which gave women “equal rights and equal responsibilities.” Respondents supported the measure 58 to 24⁵⁵. However, when language was included suggesting that it could remove protections in the law specifically for women, such as the Harris survey conducted the same year, the issue was significantly more conflictual (45 against, 40 support).⁵⁶ Opponents of the ERA, such as Phyllis Schlafly, were able to successfully identify the act with the feminist movement (i.e. abortion and advocacy of non-traditional roles for women). A survey in 1979 by the National Federation of Business & Professional Women asked respondents whether ERA advocates were “women’s libbers” seeking to change traditional roles of women.⁵⁷ There was an even split of respondents on the question, with 44% agreeing and 51% in disagreement.

The ERA was first proposed in the 1940’s. It passed the Congress and was ratified by most states—well into the 1970’s there was significant momentum in favor of ratification. The rate of ratification was particularly rapid in 1972 and 1973. Indeed, all but three states had ratified the ERA as of the end of the 1970’s. But the tide had turned as of the late 1970’s. In 1980 the Republican Party removed it from their platform. The ERA, in the wake of Roe v. Wade, was increasingly identified with the abortion-supporting feminist groups such as NOW and was attacked for costing women on the legal front. By 1982, the official end-date of the extended ratification period (extended by Congress in 1978), five states who had passed ratification measures had rescinded their ratification votes: Idaho, Kentucky, Nebraska, Tennessee, and South Dakota. The ratification movement was stopped short and, despite efforts to circumvent the rescissions and ratification deadline, the ERA was never ratified. While the ERA was strictly about equal rights, it enjoyed consensus support. When it was associated with issues such as abortion, opinion on the ERA became polarized and killed the opportunity for attaining the super-majority threshold required for a constitutional amendment.

Public opinion on aid to Blacks hasn’t changed a great deal since the 1970’s. While there has been a significant shift away from support of government involvement in helping Blacks overcome past discrimination, the polarization of t he distribution of opinion on aid to Blacks has varied little over the time period. Furthermore, the global trend for the time series masks some important developments in polarization of opinion on aid to Blacks. While the overall trend in bimodality is leptokurtic for aid to blacks (Figure 7.12), note that since 2000 there has been a steady negative trend in the kurtosis of public opinion on this issue, with the most recent distribution of opinion on aid to blacks just as bimodal as it was in 1972 (Figure 7.12). Though bimodality declined in the 1980’s the overall opinion on aid to blacks (rather than the trend) is distinctly bimodal for the breadth of the time series. The least conflictual kurtosis score for aid to Blacks is the -0.40 in 1980, well below the score for a normal distribution (zero) and far from unimodal consensus. As can be seen in Figure 7.13, a similar trend is apparent in the dispersion of opinion on government aid to Blacks. Dispersion in opinion on aid to Blacks falls in the

1980’s, and then oscillates between 2 and 1.5 standard deviations until 2000, when it begins a steady trend upwards. Just as with the bimodality measure, the dispersion of aid to Blacks in 2000 is the same as it was in 1972.

One possible explanation for the resurgence of polarization in public opinion on aid to Blacks is the continued political relevance and salience of a specific aid to Blacks policy: affirmative action. In

1978 there was strong support among Blacks and Whites for affirmative action programs. In 1978, a survey commissioned by the National Conference of Christians & Jews asked respondents whether t hey supported affirmative action policies (as long as there were no rigid quotas. Figure 7.16 reports the breakdown of the response for blacks and whites.⁵⁸ As can be seen, overwhelming (near consensual) support existed for non-quote affirmative action programs. Thirty years later, however, affirmative action is a much more controversial public policy. In Figure 7.17 the results of a 2005 Gallup poll reflects a much more divided public on the necessity of affirmative action.⁵⁹ Support for affirmative action has declined among Blacks and Whites since the 1970’s. The level of support for affirmative action among Blacks in 2005 is equivalent to the level of support for affirmative action among Whites in 1978. While AA programs still have supermajority support among Black respondents, a majority of Whites now oppose affirmative action. Indeed, this has become a considerably controversial (polarized) public policy issue between 1978 and the present for White voters. For the full sample, 50% of Americans support affirmative action while 42% oppose it. Over the course of the time series and into the 2000’s, affirmative action public opinion shifts to state very close to the cut-point for pure conflict.

We see this emerging conflict on t he policy stage in the form of the state-by-state campaign to enact “Civil Rights Initiatives” designed to enshrine the outlawing of affirmative action plans in state constitutions. This movement was originally founded by political activist and former Regent of the University of California, Ward Connerly. Connerly formed the California Civil Rights Initiative Campaign, which worked to get Proposition 209 onto the 1996 state ballot for constitutional amendments in the California state elections that year. The language of Prop. 209 included: “The state shall not discriminate against, or grant preferential treatment to, any individual or group on the basis of race, sex, color, ethnicity, or national origin in the operation of public employment, public education, or public contracting.” While couched in the language of anti-discrimination laws, the law would have the effect of preventing governments from developing and initiating affirmative action programs. While supporters viewed this as a next step in establishing equality for all before the law, critics argued it rolled back the civil rights victories of the 1960’s and the 1970’s. Connerly and his supporters have taken the campaign beyond California, successfully getting a similar ballot resolution passed in Michigan in 2006 (Hadley 2005).

Figure 7.16: Public Opinion on Affirmative Action by Race, 1978



Figure 7.17: Public Opinion on Affirmative Action by Race, 2005

However, the more recent developments in affirmative action politics haven’t been uniform victories for opponents of AA. In the companion cases of Gratz v. Bollinger and Grutter v. Bollinger

(known collectively as the University of Michigan Cases), while ruling specific provisions of the university’s affirmative action admissions program unconstitutional, the Supreme Court refused to rule affirmative action itself unconstitutional. Justice O’Conner, writing for the majority in Grutter, held that the U.S. Constitution “does not prohibit the law school's narrowly tailored use of race in admissions decisions to further a compelling interest in obtaining the educational benefits that flow from a diverse student body.” These rulings left affirmative action ensconced firmly within the salient policy debate and it remains an issue upon which opinion in the United States is polarized. The fact that debate over affirmative action is current, lively, bimodal, and conflictual suggests that it is at least part of what has driven the dispersion and bimodality of opinion on Aid to Blacks to 1970’s levels. While the previous levels were almost certainly the product of former segregationists and racial animus, the current conflict over race appears to be of a different character. Rather the current question on which the American public is polarized is whether the pendulum has swung too far the other direction in regards to government efforts to ameliorate past discrimination and provide a ‘leg up’ to Blacks in jobs, admissions to schools, and other areas of society where Blacks were previously excluded.

Conclusion

In this chapter I examined a variety of measures of the primary political issue dimensions in which political conflict exists in the United States. I found increasing polarization on ideology, social issues, defense policy, and government spending. The polarization of public opinion along the issue dimension was apparent in the detailed analysis of public opinion variance on the ANES issue scales. For all but one of the issue dimensions considered, there is evidence of consistent polarization over the time series. Even among issues in which there are global depolarization trends (aid to Blacks), the average kurtosis for these distributions is playtkurtic, indicating public opinion is relatively split into two opposing camps.

The trends in political polarization over the past thirty years reveal comprehensible attitude adjustments to political events, the policies implemented by the parties in control of the branches of the federal government, and to exogenous shocks to the system from stock market crashes to terrorist attacks. On some issues, such as the equality of women in society, there was substantial evidence of depolarization. On all measures of political polarization: mean trend, dispersion, and bimodality, opinion on gender equality has become increasingly consensual. This consensus has emerged in support of equality between men and women. While average public opinion on government jobs had no apparent trend in polarization, opinion on economic issues are fairly dispersed and the distribution is bimodal.

On other issue dimensions, there was strong evidence of polarization. On ideology, a proxy for opinion across the range of political issues, there was evidence of increasing bimodality in its distribution. The American public has shifted towards the ideological poles. On government spending, there was evidence that opinion had become more bimodal over the time series. However, the trends in opinion on government spending did not track well with actual levels of government spending, indicating that what drives opinion on government spending is at least somewhat independent of the actual level of government spending. Defense spending public opinion, on the other hand, while evidencing little polarization in the global model, tracked strongly with actual levels of defense spending. The investigation into defense spending polarization revealed that, along with actual levels of defense spending, partisanship (the party of the presidential administration) and exogenous foreign policy events interacted with the trends in actual levels of defense spending to condition and influence public opinion on defense spending. One interesting relationship reveled in the analysis of defense spending trends is the inverse relationship between actual defense spending and public opinion on the level of defense spending. When defense spending becomes extreme—either due to extreme increases in defense spending or extreme cuts in defense spending (relative to the mean level of defense spending)—the American public trends in the opposite direction.

On the social issue dimension, there is the previously mentioned significant depolarization of opinion on general equality. However, there was substantial evidence of polarization in the social issue dimension as well. On abortion, the American public has steadily become more bimodal and more dispersed. Indeed, the fact that abortion opinion has remained playtkurtic (indeed, evidencing a significant trend towards bimodality), despite the remarkable unimodal trend on women’s equality reveals an important story about American politics, namely the failure of the feminist movement to successfully associate abortion with women’s rights in the eyes of the public. And finally, while there was a significant global trend of depolarization in the dispersion and kurtosis on government aid to blacks, it is apparent that this trend has reversed itself in more recent years. As the American public has become less supportive of ameliorative policies for minorities such as affirmative action, the opinion on aid to Blacks has become more conflictual, i.e. more polarized. This chapter presents strong and consistent evidence of political polarization, not just on the culture wars, but on the gamut of political issue dimensions in American politics. While several issue dimensions failed to exhibit polarization trends (i.e. economic issues), most of the issues—whether they were social, economic, or defense-related—exhibited dispersed and bimodal opinion distributions even at the beginning of the time series. There was significant social issue polarization on the abortion issue as well as increasing polarization late in the time-series on aid to blacks. The only issue that exhibited consistent (and strong) tendencies towards consensus was in public attitudes towards the equal role of women. Americans may increasingly agree that women should have a stature in society on par with women but they agree on little else. And, furthermore, they are increasingly conflictual on social issues and foreign policy.


Public sentiment is everything. With public sentiment nothing can fail; without it nothing can succeed. He who molds public sentiment goes deeper than he who enacts statutes or decisions possible or impossible to execute. – Abraham Lincoln
Chapter 8: Polarization Trends in Issue Dimensions in Partisan Affect and Perceived National Problems

As Larry Bartels once observed, different people care about different issues (Niemi and Bartels 1985). Issue salience has long been recognized as an important factor in determining the voting behavior of the electorate and evaluations of parties and candidates. As Converse pointed out in his classic study of belief systems, people generally have just a few issues that are of particular importance to them and thus only a few issues they pay much attention to even when they are attentive to politics (Converse 1964). To the degree that issues matter, those issues matter relative to how important the public perceives them to be. Shapiro found that salience was relevant in the spatial calculus of voting (Shapiro 1969). RePass found that the salience of issues condition the degree to which partisan evaluations condition electoral choice (RePass 1971). And Rabinowitz et al. argue that “any issue singled out as personally most important plays a substantially greater role for those who so view it than it does for others” (Rabinowitz, Prothro, and Jacoby 1982). Brody found a relationship between press coverage of issues and relates those changes to patterns in presidential approval (Brody 1991). Edwards et al. use issue salience weights to assess issue impact on presidential approval, finding that issues vary over time in their salience to the mass public and as to their impact on presidential approval. Salience is relevant in evaluating presidential performance (Edwards, Mitchell, and Welch 1995). Some scholars have found that accounting for salience reduced the explanatory power of issue-based voting models, but by and large issue salience is recognized as an important factor in political behavior.

Issue salience clearly relates to political polarization. The more relevant an issue is to current political debates and partisan competition, the more likely citizens are to polarize on the issue. The more important an issue is to the candidate and party evaluations as well as the electoral choice of the mass public, the more likely political polarization is to occur and the more likely that polarization is relevant to the policy process. As I noted in the previous chapter, polarization of issues are tied to objective conditions such as unemployment, international tensions, and racial conflict. Hence the relative importance of these issue dimensions in terms of evaluating parties and identifying national problems is an important part of the story on political polarization and public policy.

The open-ended responses to the “important national problem” and party “likes and dislikes” provide a unique opportunity to assess how the public views the parties in terms of their issue positions and how this has changed over time as well as trends in what the mass public identifies as important national problems. Unlike closed-ended items, the open-ended questions on partisan affect and important national problems do not impose a structure upon the respondent’s answers. Questions that specifically identify social issues and ask respondents their opinion may be valid and reliable measures of public attitudes on social issues, but it does not and cannot tell us how important respondents believe these issues to be nor does it tell us how those issues relate to the perceptions of the parties or perceptions of the problems themselves relative to other problems. Though rank ordering items would serve this purpose, we lack long and consistent time series that allow respondents to rank order several issues, let alone the issue dimensions themselves. The open-ended responses, however, allow respondents to spontaneously identify the one or two aspects of the parties that could be classified along issue dimensions as well as the one or two issues that they view as the most important problems facing the country. Respondents are not forced onto preconceived formulations of these issues, open-ended items are free from the pitfalls associated with issue descriptions and particular question wordings, and they are not primed with a mention of that issue, which might muddy the waters in assessing salience. The most significant drawback of these items is the low response rates on open-ended questions, however this problem is somewhat mitigated by looking at the issues in broad classifications rather than the specific issue mention itself. Furthermore, the aggregate analysis here examines the full sample, and thus permits the maximum number of responses for assessing the salience and thus relevance of the issue dimensions on partisan affect and in the national problems item. While a breakdown of these responses may reveal interesting relationships, increasingly parsing the data raises the specter of false findings.

In the aggregate, this can provide us an accurate measure of the importance of particular issues relative to other issues or issue dimensions. A rank ordering of the issues that emerges from an analysis of the open-ended items is independent of what researchers believe to be relevant and important political issues. Such an analysis can also paint a picture of what issues citizens are increasingly or decreasingly associating with what they like or dislike about the Republican and/or Democratic Parties. A significant increase in the number of unsolicited issue mentions indicates greater salience for that issue dimension and is a reflection of the perceived public agenda at the level of the mass electorate. Using the measure classifying the open-ended responses on the government philosophy, social, economic, and foreign policy dimensions, we can compare shifts in the perceptions of those dimensions over the course of the time series.

There are two important methodological issues to address in analyzing the count data from the ANES open-response set items. First is the limited number of survey years in which these questions were asked. With only 12 survey years with open-ended responses, the analysis is a small-N study. While ordinary least squares regression is robust, among the possible small-sample violations of the OLS assumptions are non-normality in the distribution of the data (biasing standard errors and significance tests) and heteroscedasticity (non-constant error variance). In order to remediate this problem, I use robust standard errors for the open-ended models, estimating the asymptotic covariance matrix of the estimates under the hypothesis of heteroscedasticity. The standard error obtained from the asymptotic covariance matrix is less sensitive to violations of OLS assumptions such as normality, homoscedasticity, and the absence of outliers and overly influential data points (a particularly vexing problem with small samples). With robust standard errors, we have more confidence in the reported significance tests and estimates of the standard errors in the small-N analysis.

The open-ended response items are count variables. As such, using OLS to estimate the polarization models may yield inefficient, inconsistent, and biased estimates (Long 1997). The data does not contain an undue number of zero-value observations and, given the relatively large number of counts, the distributions of the count variables should relatively normal. While the count data is relatively well distributed—suggesting OLS procedures may be appropriate—I estimate Poisson regressions for all of the count models to illustrate the robustness of the results. The Poisson regression model estimates the probability of a count determined by a Poisson distribution. Given t he inherent problems of using a simple Poisson regression where conditional variance exceeds the conditional mean (overdispersion), I model t he counts using the Negative Binomial Regression Model (NBRM). Where overdispersion is absent, the NBRM reduces to the PRM (Long 1997). The NBRM is estimated using a maximum likelihood function:

Equation 8.1: Maximum Likelihood Function for NBRM

I report the univariate statistics on the open-ended issue items in Table 8.1 for the time period between 1976 and 1998. For all of the variables except the “Racial Problems” variable, the mean is several times the size of the standard deviation, suggestive of the relative normality of these count distributions. Interestingly enough, the “Racial Problems” variable has the lowest average count of all the open-ended variables. While the closed-ended analysis indicated racial issues remain a subject of significant political polarization, it doesn’t rate highly as one of the most important national problems at this time. That said, of the three national problem categories, the “Racial Problems” category is the most restrictive (the public order and social welfare categories include more kinds of issues).

Figure 8.1 reports a histogram of the total open-ended responses on the partisan “likes and dislikes” item on the social issue dimension for the time series. It is immediately apparent that the number of social issue mentions for partisan affect has substantially increased since 1976. Hovering near 100 mentions through 1984, the frequency of mentions increases by 150% over the course of a decade on the social issue dimension. Note that this increase is consistent with the culture wars thesis. The low point for social issue mentions was 1980, coincident with Ronald Reagan’s ascendency to the White House. Given the economic angst over stagflation, the Iranian hostage crisis, and the advancement of communism by an internationally aggressive Soviet Union, this is not a surprising result. Between 1976 and 1986, the frequency of mentions for social issues in the partisan affect measures tops 100 mentions only once, in 1984 (102). The steady increase of social issue mentions throughout the 1980’s is coincident with the rising political schism in American society identified by Hunter in the original culture

wars thesis. While 1996 and 1998 social issue mentions fall off from the high in 1994 (253), both of those years have significantly higher frequencies of social issue mentions in comparison to any of the survey years prior to 1990. Clearly there has been a marked increase in the salience of social issues for the American public and they increasingly identify the parties— both in terms of what they like and what they do not like about the parties—based on their perceived positions on social issues.

Figure 8.2 illustrates the trend in issue mentions for the partisan “likes and dislikes” measure for all four of the issue dimensions. Government philosophy mentions have been relatively low and fairly constant over the time-period, with a minor blip upwards in the early part of the 1990’s. It peaks at 100 mentions in 1994, coincident with the Republican Revolution that resulted in a turnover of both the Senate and the House of Representatives to the Republican Party, the latter of which had not been in Republican hands since 1954. Newt Gingrich’s congressional campaign, taking advantage of public angst over Clinton’s failed health care plan and the perception that Democrats had tilted Left and were out of touch with the American people, was perhaps the most ideological off-year election in history. It was centered on specific, poll-tested legislative proposals labeled “The Contract with America.” It is thus not particularly surprising to see ideological and government philosophy issues gain in salience over this time period.

There is an apparent increase in economic issue mentions, but it isn’t a substantially large increase. Economic issue mentions go from about 90 mentions in 1972 to a peak of 180 mentions in 1988 (coincident with the 1987 “Black Monday” stock market crash) and settles in around 140 mentions for the rest of the time series. Foreign policy mentions peak in 1984 (230 mentions), undoubtedly a consequence of a presidential election centered on the United States posture towards the Soviet Union and an explicit dove (Walter Mondale) on the Democratic ticket. As established in Chapter 6, defense and foreign policy public opinion is sensitive to foreign policy events and the policy positions of

presidential administrations. The significant increase in defense spending advocated and implemented by Ronald Reagan was likely a significant factor in raising the salience of foreign policy in public opinion along with the Cold War. Foreign policy mentions steadily decline throughout the 1990’s. In 1988, economic issues displace foreign policy mentions as the most frequently mentioned issue by respondents. With the Cold War having ended and a Democratic president reluctant to commit troops abroad, this is not a surprising result, as was apparent in the assessment of defense policy opinion in the 1990’s in Chapter 7. While there is no data on the likes and dislike item for the first decade of the 21^st Century, there is little doubt that this measure would have spiked in the wake of 9-11 and the resulting wars in Iraq and Afghanistan. Foreign policy is an increasingly inviting area for partisan conflict and political polarization.

Standing out from the other dimensions is the significant polarization trend on the social issue dimension. In the 1970’s social issues were third of the four issue dimensions and lagged well behind foreign policy concerns. By the 1990’s, social issue mentions were the most frequent open-ended response and, despite the decline from the peak year of 1994, is the most frequent of the issue dimensions through the end of the time series. Social issues were more frequently mentioned even in 1992, the presidential campaign in which James Carville, a campaign advisor for Bill Clinton, famously quipped that “It’s the economy, stupid.” Apparently it was also social issues.

Table 8.2 reports OLS regression models with bootstrapped (robust) standard errors of the trend in issue dimension mentions on the “likes and dislikes” partisan affect item from the ANES.

Equation 8.2: OLS with RSS – Time Trends in Issue Mentions

Note there is a significant trend for all four of the political issue dimensions. Both frequency of government philosophy mentions (1.470) and economic issue mentions have been on the rise (2.150). The spike in mentions of government philosophy in the 1990’s clearly contributes to the positive coefficient for the model. However, the model is a relatively poor fit, likely a consequence of the return to the ‘equilibrium’ government philosophy mentions in 1996. Economic issue mentions have become more frequent as well, though the economic mentions fall off in the later part of the 1990’s as the economy was in recovery and we were on the cusp of the “dotcom” bubble. There is a significant trend for foreign policy mentions, but it is a negative trend: foreign policy mentions decline about 2.5 mentions per survey year. Again, the end of the Cold War is the best explanation for this result. Social issue mentions evidence the largest increase—7.154 mentions per survey year—tripling the coefficients of the other issue dimensions. The model-fit statistic indicates that the time series explains 68% of the variation in social issue mentions, and this is by and far the best fit model of the four issue dimensions. This is significant evidence of political polarization on the social issue dimension.

Model:	Trend: Polarization Y/N	Intercept (R.S.E.)	Parameter Estimate (R.S.E.)		R2	N
GOVERNMENT PHILOSOPHY	y	-2872.529 (1587.578)	1.470 (0.801)	*	.200	12
SOCIAL ISSUES	Y	-14065.000 (3002.236)	7.154 (1.515)	***	.686	12
ECONOMIC ISSUES	y	-4142.311 (1751.358)	2.150 (0.882)	**	.231	12
FOREIGN POLICY	N	5330.541 (2339.134)	-2.608 (1.176)	**	.268	12

Model:	Trend: Polarization Y/N	Intercept (R.S.E.)	Parameter Estimate (R.S.E.)	PE Z Pr > |Z|	N
GOVERNMENT PHILOSOPHY	Y	-56.384 (28.939)	0.030 (0.015)	.038	12
SOCIAL ISSUES	Y	-91.855 (19.377)	0.049 (0.010)	< .0001	12
ECONOMIC ISSUES	Y	-28.137 (14.010)	0.017 (0.007)	.018	12
FOREIGN POLICY	N	40.192 (15.894)	-0.018 (0.008)	.027	12

Equation 8.3: OLS with RSS – Time Trends in Issue Mentions

where:

y_i = survey year

The models in Table 8.3 confirm the robustness of the OLS-estimated models for partisan likes and dislikes reported in Table 8.2. The parameter estimate that maximizes the likelihood of observing the increase in issue mentions is largest for social issues, consistent with the OLS models. Furthermore, while all of the models are significant at the .05 level, only the model for social issues is significant at the .01 level (Z Pr > |Z| < .0001). The ML model coefficients for all of the issue dimensions are consistent with the betas reported for the OLS models. The ML models for social issues, economic issues, and government philosophy have positive coefficients, while the ML model for foreign policy has a negative coefficient.

Table 8.4 reports on trends in the other open-ended item asked consistently over the ANES time series: the “most important problem” question using OLS estimates with bootstrapped (robust) standard errors⁶⁰. Survey respondents were asked to respond with what they believed to be the most important problems facing our nation at the time. The Public Order category for the open-ended “most

Model:	Trend: Polarization Y/N	Intercept (R.S.E.)	Parameter Estimate (R.S.E.)		R2	N
PUBLIC ORDER	y	-23214.000 (10214.000)	11.794 (5.152)	**	.257	12
RACIAL PROBLEMS	Y	-1844.554 (953.620)	0.943 (0.481)	*	.355	12
SOCIAL WELFARE	N	-7730.536 (1198.000)	4.103 (6.021)		.035	12

Model:	Trend: Polarization Y/N	Intercept (R.S.E.)		Parameter Estimate (R.S.E.)			PE Z Pr > |Z|		N
PUBLIC ORDER			y		-104.381 (45.390)	0.055 (0.023)		.016		12
RACIAL PROBLEMS			Y		-137.292 (65.192)	0.070 (0.033)		.032		12
SOCIAL WELFARE			N		-13.275 (29.126)	0.010 (0.015)		.507		12

While the “racial problems” model is also significant, it increases at just short of one mention per survey year (R² = .355). It is not surprising—given the relationship between the mean and standard deviation in the raw data—that this was the only model for which the use of robust standard deviations resulted in an increase in the estimated standard error for the model (and thus a decrease in statistical significance). There is no apparent statistically significant trend in social welfare mentions, though the parameter estimate for the model is positive. The ML estimates for the three national problems models are reported in Table 8.5.⁶² As with the partisan affect models, the results of the maximum likelihood estimations of the national problems counts is consistent—both in terms of statistical significance and the direction / size of the coefficients—with the OLS models reported in Table 8.4. Both the “Public Order” and “Racial Problems” models are significant and in the expected direction. The Social Welfare ML model, as with the OLS model, is statistically insignificant. The substantial increase in social issue mentions on the national problems item is significant and substantively important evidence of political polarization on the social issue dimension and in support of the culture wars thesis.

I have examined the open-ended responses to two items from the ANES: partisan affect and national problem items. While I found significant increases for three out of the four issue dimensions (government philosophy, economic, and social), my analysis found that foreign policy had declined in salience for the American public. Again, as mentioned earlier, foreign policy opinion for this time period is distinctly non-linear, and the open-ended response survey years end short of 2001 and the September 11^th terrorist attacks. On both partisan affect and the identification of important national problems, there has been a significant increase in the absolute salience and relative position vis-à-vis the other issue dimensions for social issues. Indeed, the social issue dimension evidenced the largest increase in counts and was the most frequently mentioned issue dimension from 1988 forward. This chapter, as was the case in Chapter 6, presents strong and consistent evidence of political polarization, not just on the culture wars, but on the gamut of political issue dimensions in American politics.