Salience, Security, and Social Welfare:
The Rise and Demise of Security Moms

Randolph Horn and Tatyana Karaman
SamfordUniversity

Introduction
     The events of September 11, 2001 led to a profound surge of bipartisanship on Capital Hill and to a phenomenal rallying behind the president and his response to the attack by the American electorate. Though the display of bipartisan unity among politicians was a short-lived phenomenon (see for example, Jacobson, 2003), popular wisdom holds that events had a more lasting impact on the partisan attachment in the electorate resulting in a significant expansion of the Republican camp. More specifically, the popular press has argued that the events and the resulting war on terror produced a unique and pivotal subgroup of the electorate, the so-called “security moms.” As the name suggests, the “security moms” describe white suburban middle class married women with children or the demographic previously known as the “soccer moms.” However, unlike the “soccer moms” who have been traditional supporters of the Democratic Party, the “security moms” are more concerned with the war on terrorism and/or war in Iraq than all other voting issues and identify themselves with the Republican Party.
     The purpose of this paper is two-fold. First, we will analyze whether there has been a wide-spread movement towards the Republican camp among women since the attack. Then, we will investigate whether the security concerns were indeed determining factors for any new affiliation.
Partisanship Prior to September 11th
     The gender gap in partisan differences in voting behavior became an apparent reality of American politics in the mid 1980s. Throughout this period, women were more likely than men to identify with the Democratic Party and to support Democratic candidates. Early research on the gender gap attributed the differences in party identification to women shifting toward the Democratic Party away from Reagan and his conservatism (see Gilens 1988 for a review of this literature). Many factors were suggested as sources of women’s distaste for Reagan-era Republicanism: women’s disadvantaged socio-economic status (Piven, 1984; Smeal, 1984), concerns over Reagan’s views on the so-called “women’s issues” of abortion and equal rights (Abzug, 1984; Klein, 1984), and Reagan’s aggressive foreign policy (Frankovic, 1982; Lake, 1982). However, according to Gilens (1988), the gender gap was produced primarily by military and social welfare issues while either “women’s issues” or socio-economic status contributed little to the explanation of differences between men and women.
     The Reagan-centered explanations of the gender gap, however, were found to be insufficient when the differences in party identification and voting behavior between the sexes persisted and even increased throughout the 1990s. Indeed, in the 1996 presidential election, the gender gap resulted in a fourteen percentage point difference between women and men in support for Clinton’s candidacy. As a result, general models of the gender gap were developed. The Attitude Model explains gender differences by attitudinal differences between the sexes. Specifically, this model assumes that there are vital differences between men and women in their attitudes toward a broad range of political issues. Together, these differences result in the gender gap (see for example, Conover, 1988; Cook and Wilcox, 1991; May and Stephenson, 1994). Alternatively, the Salience Model, while allowing for attitudinal differences, focuses on the differences in weights that men and women assign to various political issues. Therefore, even when a male and a female voter have identical stances on a particular issue, the difference in weights that they assign to this issue can make their evaluation of a candidate quite different (Welch and Hibbing, 1992; Bedyna and Lake, 1994, Chaney, Alvarez, and Nagler, 1998).
     In a comparative analysis of the two models, Kaufmann and Petrocik (1999) concluded that both attitudinal differences between the sexes and the differences in weights that they assign to various political issues have been essential components of the gender gap throughout the 1990s. However, according to the authors, while attitudinal differences have a relatively stable nature, gender differences in issue salience are dynamic and context-dependent. As the result, Kaufmann and Petrocik (1999) conclude that sharp changes in the gender gap should be attributed to the changes in salience of some issues. Finally, the authors clearly demonstrate that the widening gender gap in the 1990s was produced by an shift of male voters to the Republican side while female political affiliation and voting patterns remained relatively unchanged. This conclusion agrees with Wirls’ (1986) argument that both men and women defected from the Democratic party, though the magnitude of defection has been much higher among men than among women.
     Researchers have identified a number of specific issues resulting in the different policy attitudes and issue salience between men and women. For example, studies pointed to social welfare issues (Kaufmann, 2002; Kaufmann and Petrocik, 1999) and cultural issues such as abortion, equal representation of women, and legal rights of homosexuals (Kaufmann, 2002). Specifically, using pooled American National Election Study data for the period from 1988 throughout 2000, Kaufmann (2002) concludes that social welfare policies and cultural issues play an important role in partisan identification of both sexes; however, they affect male and female partisanship in different ways. Overall, male voters tend to have more conservative views on social welfare issues than female voters. More importantly, however, not only do men tend to have more conservative attitudes toward social welfare policy, but they also tend to treat social welfare issues as the most significant determinants of partisanship. In contrast, women based their partisan identification more heavily on their attitudes toward cultural issues rather than toward social welfare issues.
     In addition, it has been shown that women react differently than men to changing economic conditions. Not only do women tend to be more concerned with the personal impact of worsening economic conditions than men (Piven, 1985; Stark, 1996), but they also are more likely than men to relate worsening economic conditions to social welfare issues (Schlozman, Burns, Verba, and Donahue, 1995; Welch, and Hibbing, 1992). Based on these factors, it is not surprising that the differences in Democratic partisanship between the sexes tend to broaden in periods of worsening economic conditions as measured by personal income growth (Box-Steffensmeier, De Boef, and Lin, 2004). Finally, gender differences in partisan identification tend to widen in response to broader changes in the political environment. Specifically, the gender gap increases as the overall political climate becomes more conservative and decreases as conservative macroideology ebbs (Box-Steffensmeier et al., 2004).
Partisanship Post September 11th
     The period from 2000 to 2002 saw a substantial shift among the pubic toward the Republican party. This shift has been especially pronounced among women in the electorate (See Figure 1). Here, the magnitude of the shift among women was more than double that of men. The period from 2002 to 2004, however, saw a slight reverse of the shift with both men and women showing net movement in the Democratic direction. Again, women’s net shift more than doubled that of men. However, these changes in partisanship were not uniform within each sex group. Specifically, the shift toward the Republican party in the period from 2000 to 2002 took place among females and males who already weakly identified with the Republican party in 2000 or those who thought of themselves as independent (See Figures 2 and 3). In contrast, the electorate who identified in 2000 with the Democratic party began to identify with this party even more. Therefore, the shift toward the Republican party from 2000 to 2002 has been mainly produced by female voters who in 2000 already weakly identified with the Republican party, yet thought of themselves as independent.
Model and Data
     To test the impact of security concerns on partisan identification after the events of September 11th among both men and women, we develop the following model:
Party Identification = f(Policy Attitudes, Personal Traits, Political Environment), that is, party identification is a function of a person’s attitudes toward various policy issues, his or her personal traits, and political environment.
Specifically we derive the following hypotheses:
Hypothesis 1: The greater a person’s concern over security issues, the more likely he/she is to identify with the Republican Party.
Hypothesis 2: Those with negative attitudes toward the traditionally female issues, are more likely to identify with the Republican Party
Hypothesis 3: Those with a positive retrospective evaluation of the economy, are more likely to identify with the Republican Party.
We test these hypotheses with the data from 2000-2004 American National Election Study (ANES) panel. Like previous ANES panel studies, the 2000-2004 study tracks a scientific sample of respondents over three electoral cycles. In each cycle respondents are interviewed both before and after the election. The questions include a standard set of items commonly used in ANES, questions developed for the particular election (e.g. candidate evaluations), and experimental questions that reflect researchers’ desire to assess the effects of question wording on responses. The panel, and the ANES typically, uses a split form design which allows for testing of experimental questions and to increase the total number of questions asked in the survey: Sometimes a battery of questions is asked in only one of the forms to reduce the cost of administration. The panel starts off with a large number of respondents, but as might be expected there is some sample mortality over the course of four years. Additional respondents are added to the sample to make it representative of any given year's electorate, but added respondents cannot be part of the panel because they were not respondents in earlier waves. In the 2000-2004 panel study 840 respondents completed all waves of the study.
     Party Identification is measured on a 7 point scale with strong Democrats assigned a value of 0 and strong Republicans a value of 6. Policy Attitudes include an individual’s attitudes toward the following public policies: security, welfare, childcare, public education, abortion, equal rights for women, and economic evaluations. An individual’s attitudes toward security are assessed by the following two questions: 1) should federal spending on homeland security be decreased, kept the same, or increased? and 2) should federal spending on the war on terrorism be decreased, kept the same, or increased? Attitudes toward welfare were assessed by the following question: should federal spending on welfare programs be decreased, kept the same, or increased? Attitudes toward childcare policies were measured by the following question: should federal spending on child care be decreased, kept the same, or increased? Attitudes toward public education were assessed by the two following questions: 1) should federal spending on public schools be decreased, kept the same, or increased?, and 2) should federal spending on big-city schools be decreased, kept the same, or increased? Respondents’ answers to all of the above questions were assigned the following values: decrease –- a value of 1, kept the same –- a value of 2, and increased -- a value of 3.
     Respondents’ attitudes toward abortion were measured by respondents’ self-placement on the following options: 1) by law, abortion should never be permitted, 2) the law should permit abortion only in case of rape, incest, or when the woman's life is in danger, 3) the law should permit abortion for reasons other than rape, incest, or danger to the woman's life, but only after the need for the abortion has been clearly established, 4) by law, a woman should always be able to obtain an abortion as a matter of personal choice. Respondents’ self-placements were assigned values of 1, 2, 3, and 4, accordingly. Attitudes toward women’s equal rights were assessed by respondents self-placement on the following options: 1) strongly agrees that woman’s place is in the home, 2) agrees that woman’s place is in the home, 3) other, 4) agrees with women’s equal role, 5) strongly agrees with women’s equal role. Respondents’ self-placements were assigned values of 1, 2, 3, 4, and 5, accordingly. Finally, respondents’ retrospective economic evaluations were assessed by the following question: would you say that over the past year the nation's economy has gotten better, stayed about the same, or gotten worse? The respondents’ answers were assigned the following values: 1) getting better, 2) stayed about the same, and 3) gotten worse. Respondents’ personal traits included the following variables: years of education, marital status (married is coded 1, else 0), number of children, age (in years), personal income, and race (non-white is coded 1, else 0).
     In addition to policy attitudes and personal traits, the model controls for the impact of the political environment on the state and district levels. Such aggregate measures may represent the immediate environment in which people form and express attitudes or a cultural context where values are transmitted and expressed despite changing material conditions (e.g. Wright, 1977). Geographic patterns of preference or values may be evident in sub-state units as well as in different regions of the country. Indeed, a long line of scholars has identified sub-state sections where current preferences reflect some previously established (typically) material reality (e.g. Key, 1984; Gimpel and Schuknecht 2004). Key’s work (1949) recognizes not only regional variation in the strength of political parties, but sectional differences in vote choice within a party across a state. Elazar (1962, 1984) attributed geographic variation in preferences to differences in political culture, itself a product of differential settlement patterns. Wright (1977) pioneered the use of continuous variables as measures of sub-state context. He merged county-level demographic data with public opinion surveys. For Wright, context was not only an important control variable, but a factor that sometimes had significant a direct effect on preferences. Conway (1989) substantiates the significance of contextual effects in a wide variety of circumstances and because of that importance urges the inclusion of contextual variables in models of individual behavior. Gimpel and Schuknecht (2004) find sectional differences within states in vote choice and that different policy questions may be salient in different states. While Gelman, Shor, Bafumi, and Park, (2005) find that income is positively correlated with Republican vote choice, they find that the slope varies depending on the state political context. In this paper, contextual variables measured at the congressional district level include the 2000 Republican presidential vote, median age, percent Anglo, and per capita income. The state-level contextual variables include percent Anglo, percent homeowners, percent of households in poverty, per capita income, percent of households with public assistance income. Finally, regional variations are controlled for by the dichotomous variable South that assigns a value of 1 to the eleven southern states, and a value of zero for all other states.
Discussion
     As the results in Table 1 show, partisan identification among women in 2002 was driven by the following primary concerns: security, economy, abortion, and women’s equal rights. As predicted, increased security concerns were associated with identification with the Republican party, while concerns for economic conditions and favorable attitudes toward abortion, and women’s equal rights were at the center of Democratic party identification. (Pearson’s correlation between abortion and equal rights is .27). It is notable, that none of the other traditionally “female issues”, such as welfare, childcare, and school policies, have a statistically significant impact on female partisan identification in 2002. This finding leads us to conclude that the newly emerged security concern was able to suppress the importance of these issues. Among the demographic characteristics, neither marital status nor having children has a statistically significant impact on partisan identification. Indeed, income, a measure of social economic status (SES), and race are the only statistically significant variables in this category. As expected, non-white women are more likely than white women to identify with the Democratic party. On the other hand, women of a higher social economic status, as measured by income, are more likely to think of themselves as Republican. None of the district level variables have a statistically significant impact on party identification among women during this period. However, three state level variables, percent of white population, percent of families below poverty level, and state per capita income have a positive and statistically significant impact on Republican identification. The direct effects of the state-level contextual variables on women’s partisanship are consistent with previous findings concerning women’s sensitivity to the economic context, but perhaps more importantly specify the context of individual choice (see Gelman, et al.): Richer and whiter environments might breed Republicanism (outside of the four exceptional states Gelman et al. identify), while the percentage of households in poverty might more accurately capture the conservative tendencies of the Sunbelt than the south indicator.
     Party identification among men in 2002 was based on concerns over security, welfare, and abortion. Men with a heightened concern over security tended to identify with the Republican party, while liberal attitudes toward welfare and abortion were associated with Democratic identification. Interestingly, men’s attitudes toward the economy did not have a statistically significant impact on their partisanship (the correlation between welfare and economic evaluation is .11 as measured by Pearson’s R) as well as their attitudes toward women’s equal rights. Among the demographic variables, only race has a statistically significant impact on male party identification, with white men being more likely than non-white men to identify with the Republican party. The contextual variables measured at the district level are consistent with previous findings. Those residing in more conservative political environments are more likely to identify more strongly with the Republican party, while those with greater exposure to families poverty are more likely to hold the opposite view. There is an inverse relationship between the “whiteness” of the district and Republican proclivities. This is consistent with Tolbert and Hero’s (1996) work on diversity and voting behavior. None of the state level variables has a statistically significant impact on male partisanship.
     As the results in Table 2 show, in 2004 female partisanship was based on a much broader variety of issues. While an increased concern for security drove identification in a Republican direction, Democratic identification was based on concerns for economic performance, favorable attitudes toward abortion, women’s equal rights, and welfare, and toward increased spending on childcare. Among the demographic and SES variables, only education and race have a statistically significant impact on partisanship. White women and women with a higher level of education are more likely to identify with Republican party, while non-white women and women with a lower level of education are more likely to identify with the Democratic party. Similar to 2002, none of the district variables has a statistically significant impact on female partisanship in 2004. Also, analogous to the 2002 results, three state level variables, percent of white population, percent of families below poverty level, and state per capita income have a positive and statistically significant impact on Republican identification.
     Like female partisanship in 2004, male partisanship is associated with a variety of issues. Analogous to women, men with heightened security concerns are more likely to identify with Republican party. In contrast Democratic partisanship is associated with concerns over economic performance, favorable attitudes toward abortion, and welfare, and toward increased spending for childcare. Indeed, male partisanship and female partisanship appear to be associated with the same variables with the exception of attitudes toward women’s equal rights, which have no statistically significant impact on male partisanship. Among the demographic and SES variables, race and income have a statistically significant impact on male partisanship. White men and men with a higher personal income are more likely to identify with Republican party than non-white men and men with a lower personal income. For women, the contextual variables have the same impact in 2004 that they did in 2002. For men only one state level variable has a statistically significant impact on male partisanship: State per capita income has a positive and statistically significant impact on Republican identification among men. In other words, for men the salient contextual factors seem to have narrowed from the ideological (district-level) to economic (state-level) environments.
     As the findings show security concerns, indeed played a significant role in both female and male party identification during 2002 and 2004 elections. So far, however, it remains undetermined whether, indeed, as the media has claimed, security concerns had a greater impact on female than male partisanship. A straightforward way to test this claim is to compare regression coefficients between female and male voters. The results of this analysis are presented in the last columns of the tables. Here, the dummy variable security*female is used to capture the difference in the magnitude of security concerns between women and men in 2002 and 2004. If the difference in the security concerns between the sexes is statistically significant then the dummy variable will reach statistical significance as well. As the results show, the dummy variable fails to reach the level of statistical significance in both years. Therefore, contrary to the media claim, security concerns did not have a greater impact on party identification among women than men.
     However, in order to assess the notion of security moms publicized by the media, additional tests are necessary. The results of these tests are presented in Table 3. In 2002, being married with children had a positive and statistically significant effect on security concerns among women. That is, married women with children were more likely than nonmarried women with children to support additional spending on homeland security and on the war on terrorism. This finding offers a partial support to the notion of security moms perpetrated by the media. Another statistically significant variable is age. That is, older women were more likely than younger women to support additional spending on homeland security and on the war on terrorism. However, education, and race had no statistically significant impact on security concerns among women in 2002. Therefore, the results show that even though security concerns were dominant among married women with children, these women did not necessarily represent the media’s portrayal of the security moms – white suburban middle class married women with children. The results for 2004 are even more telling. As the test indicates, contrary to the public perception and the media depiction, in 2004, none of the variables associated with the notion of security moms had a statistically significant impact on concerns with security issues among women. This lead us to conclude that even though security concerns were heightened among married women with children in 2002 by 2004 the phenomenon of security moms as portrayed by the media was nonexistent.
Conclusion
     The events of September 11th led to a remarkable, though short-lived, display of bipartisanship among the politicians and to a significant shift toward conservatism among the electorate. While this shift has occurred among both male and female voters, the press has attributed it to the most part to the changing attitudes among white suburban middle class married women with children or the so called security moms. It has been argued that this category of women who were traditionally concerned with such issues as child-care, education and welfare and, hence, identified and voted along Democratic lines became predominantly concerned with security issues and consequently switched both their party identification and vote choice to the benefit of the Republican party.
     This paper shows that this view is faulty. While indeed, concerns over security issues have had significant impact on partisan identification of women since the attacks, there is no indication that those concerns were especially heightened among white suburban middle class married women with children. Moreover, while in 2002, being married with children increased the probability that women will have security concerns, in 2004 security moms were little more than a media myth. In addition, these findings show a interesting impact of security concerns on issues traditionally associated with women such as welfare, childcare, and school policies. In 2002, none of these issues has an impact on female partisan identification. However, in 2004, women returned to their traditional concerns as determinants of their partisanship. These findings suggest that however remarkable the shift toward conservatism and, hence, the Republican party among women in 2002 was, it did not constitute a new electoral realignment.

References:

Abzug, B. (1984). Gender Gap. Boston: Houghton Mifflin.
Bedyna, M., & Lake, C. (1993). Gender and voting in the 1992 presidential election.
       In E. Cook, S. Thomas, & C. Wilcox (Eds.), In the year of the woman: Myths and Realities
       (pp.237-254). Boulder, CO: Westview.
Box-Steffensmeier, J., De Boef, S., & Lin, T. (2004). The dynamics of the partisan gender gap.
       American Political Science Review, 98, 515-525.
Chaney, c., Alvarez, R.M., & Nagler, J. (1998). Explaining the gender gap in U.S. presidential elections, 1980-1992.
       Political Research Quarterly 51, 311-340.
Conover, P. (1988). Feminists and the gender gap. Journal of Politics, 50, 985-1010.
Conway, M. (1989). The political context of political behavior. Journal of Politics, 51(1), 3-10.
Cook, E., & Wilcox, C. (1991). Feminism and the gender gap – a second look.
       Journal of  Politics, 53, 1111-1122.
Frankovic, K. (1982) Sex and politics – new alignments, old issues, PS, 15, 439-448.
Gelman, A., Shor,B., Bafumi, J. & Park, D. (2005)
       Rich state, poor state, red state, blue state:What’s the matter with Connecticut? Mimeo. Columbia University:
       http://www.stat.columbia.edu/~gelman/research/unpublished/redblue11.pdf
Gilens, M (1988) Gender and support for Reagan: A comprehensive model of presidential approval.
       American Journal of Political Science 32, 19-49.
Gimpel, J. G. & Schuknecht, J. E. (2004). Patchwork nation: Sectionalism and political change in American politics.
       Ann Arbor, MI: University of Michigan Press.
Jacobson, G (2003) Terror, terrain, and turnout: Explaining the 2002 midterm elections.
       Political Science Quarterly, 118(1), 1-22.
Kaufmann, K. (2002). Culture wars, secular realignment and the gender gap in party identification.
       Political Behavior, 24, 283-307.
Kaufmann, K., & Petrocik, J. (1999). The changing politics of American men: Understanding the sources of the
       gender gap. American Journal of Political Science, 41, 270-283.
Key, V. O. (1984). Southern politics in state and nation. Knoxville, TN: University of Tennessee Press.
Klein, E. (1984) Gender Politics. Cambridge: Harvard University Press.
May, A., & Stephenson, K. (1994) Women and the Great Retrenchment: The political economy of gender in
       the 1980s. Journal of Economic Issues, 28, 533-542.
Piven, P. (1984) Women and the state: Ideology, power, and the welfare state. Socialist Review, 14, 11-19.
Schlozman, K., Burns, N., Verba, S., & Donahue, J. (1995). Gender and citizen participation: Is there a different
       voice? American Journal of Political Science, 39, 267-293.
Smeal, E. (1984) Why and How Women Will Elect the Next President. New York: Harper and Row.
Stark, S. (1996). Gender politics. The Atlantic Monthly, July, 71-80.
Tolbert, C., & Hero, R. (1996). Race/ethnicity and direct democracy: An analysis of California's illegal immigration
       initiative. Journal of Politics, 58(3), 806-818.
Welch, D., & Hibbing, J. (1992) Financial conditions, gender, and voting in American national elections.
       Journal of Politics, 54, 194-213.
Wirls, D. (1986) Reinterpreting the gender gap. Public Opinion Quarterly, 50, 316-330.
Wright, G. C. (1977). Contextual models of electoral behavior: The Southern Wallace vote.
       The American Political Science Review, 71(2): 497-508.

Table 1. 2002 Partisan Identification regressed on individual, district, and state characteristics.
Entries are unstandardized regression coefficients with t-values in parentheses.

a p<0.10
* p<0.05
** p<0.01
***p<0.001

Table 2. 2004 Partisan Identification regressed on individual, district, and state characteristics.
Entries are unstandardized regression coefficients with t-values in parentheses.

a p<0.10
* p<0.05
** p<0.01
***p<0.001

Table 3. Support for increased security spending regressed on individual characteristics among women.
Entries are unstandardized regression coefficients with t-values in parentheses.

a p<0.10
* p<0.05
** p<0.01
***p<0.001

--------------------------------------------------------------------------------------

1. According to the National Election Study