net Spending Support for Different Programs, by Income Level

Transcription

1 Figure 1.1 net Spending Support for Different Programs, by Income Level 1 8 Low Middle High 6 Net Spending Support Crime Defense Education Health Environment Welfare 8 1 Source: Authors calculations based on data from the General Social Surveys (Davis, Smith, and Marsden ).

2 Figure 1.2 net Support for Welfare Spending, by Income Level, 1973 to 28 Net Support for Welfare Spending Middle income High income Year 1991 Low income Source: Authors calculations based on data from the General Social Surveys (Davis, Smith, and Marsden ).

3 Figure 1.3 tax Preferences, by Income Level, 1976 to 28 Percent Saying Own Taxes Are Too High Low income High income Year Middle income Source: Authors calculations based on data from the General Social Surveys (Davis, Smith, and Marsden ).

4 Figure 1.4 stimson s Policy Mood, by Income Level, 1956 to 26 Percent Liberal Low income Year High income Middle income Source: Authors calculations based on data from the General Social Surveys (Davis, Smith, and Marsden ), American National Election Studies (Sapiro, Rosenstone, and the National Election Sudies 24), and the ipoll Databank (Roper Center Public Opinion Archives, various years).

5 Figure 1.5 stimson s Policy Mood, by Education Level, 1956 to Percent Liberal High education Year Low education Middle education Source: Authors calculations based on data from the General Social Surveys (Davis, Smith, and Marsden ), American National Election Studies (Sapiro, Rosenstone, and the National Election Sudies 24), and the ipoll Databank (Roper Center Public Opinion Archives, various years).

6 Figure 1.6 stimson s Policy Mood, by Party Identification, 1956 to 26 Percent Liberal Independents Democrats Year Republicans Source: Authors calculations based on data from the General Social Surveys (Davis, Smith, and Marsden ), American National Election Studies (Sapiro, Rosenstone, and the National Election Sudies 24), and the ipoll Databank (Roper Center Public Opinion Archives, various years).

7 Figure 1.7 Kellstedt s Racial Policy Liberalism, by Race, 1962 to African American respondents Percent Liberal All respondents 4 3 White respondents Year Source: Authors calculations based on data from the General Social Surveys (Davis, Smith, and Marsden ), American National Election Studies (Sapiro, Rosenstone, and the National Election Sudies 24), and the ipoll Databank (Roper Center Public Opinion Archives, various years).

8 Figure 2.1 predicting Ideology and Partisanship as a Function of Demographics Versus Thermometer Scores R-Squared Ideology-d Ideology-t PID-t PID-d Year Source: Authors compilation, based on the National Election Studies (Sapiro, Rosenstone, and the National Election Studies 24). Note: Ideology-d and PID-d reflect values based on demographic variables. Ideology-t and PID-t reflect values based on thermometer scores.

9 Figure 2.2 perceptions of Candidate and Party Ideology, by Ethnic-Racial Group 7 Liberal-Conservative Scale Whites Blacks Latinos Asians Ethnic-Racial Group Kerry Democrats Republicans Bush Source: Authors compilation, based on the 24 linked data, as discussed in the text. Figure 2.3 opinions on the Role of Government, by Ethnic-Racial Group Mean Response to 7-Point Scale Whites Blacks Latinos Asians Ethnic-Racial Group For spending on services Against provision of jobs Against provision of health insurance Source: Authors compilation, based on the 24 linked data, as discussed in the text.

10 Figure 2.4 opinions on Government Aid to Minorities, by Ethnic-Racial Group Mean Response to 7-Point Scale Whites Blacks Latinos Asians Ethnic-Racial Group Against aid to blacks Against aid to Latinos Source: Authors compilation, based on the 24 linked data, as discussed in the text.

11 Figure 2.A1 nes 24 Survey Party ID Differences, by Thermometer Matching Versus Demographics Matching.4.3 Thermometer matched N = 459 Density.2.1 Random matched N = 583 Demographic matched N = Difference in Party ID Source: Authors compilation, based on the 24 National Annenberg Election Survey (Romer et al. 26) and the 24 National Election Study (Center for Political Studies 24).

12 Figure 2.A2 nes 24 Survey Liberal-Conservative 7-Point Scale Differences, by Thermometer Matching Versus Demographics Matching.3.25 Thermometer matched N = Density Random matched N = 583 Demographic matched N = Difference in Liberal-Conservative 7-Point Scale Source: Authors compilation, based on the 24 National Annenberg Election Survey (Romer et al. 26) and the 24 National Election Study (Center for Political Studies 24).

13 Table 2.1 distribution of Demographic Indicators in the Linked, Annenberg, and NES Data Linked Annenberg NES Gender (percent) Male Female Education (percent) Eight grade or less Grades 9 to High school diploma Some college Two-year college Bachelor s degree Advanced Race-ethnicity (percent) Black Asian White Latino Age (mean) N 61,98 81,422 1,212 Source: Authors compilation, based on the 24 National Annenberg Election Survey (Romer et al. 26) and the 24 National Election Study (NES) (Center for Political Studies 24).

14 Table 2.2 Comparing the Linking Procedure with Demographic Matching Linking Procedure Matching Procedure (Age, Education, Gender) Political Percent Number Percent Number Variable Correctly of Exact Correctly of Exact Classified Matches Classified Matches Vote choice Kerry Bush Party ID Democrats Independents Republicans Ideology Liberal Moderate Conservative Source: Authors compilation, based on the 24 National Annenberg Election Survey (Romer et al. 26) and the 24 National Election Study (Center for Political Studies 24).

15 Table 2.3 Group Thermometer Evaluations, by Ethnic-Racial Group Group Thermometer Scores Toward... R s Latinos Blacks Whites N Race Linked NES Linked NES Linked NES Linked NES Latino (14.) (15.5) (2.1) (18.7) (19.6) (18.3) Black (17.5) (18.2) (14.7) (15.5) (23.) (2.) White (19.1) (19.3) (18.8) (18.4) (19.2) (19.2) Source: Authors compilation, based on the 24 National Annenberg Election Survey (Romer et al. 26) and the 24 National Election Study (Center for Political Studies 24). Note: Standard deviation in parentheses.

16 Table 2.4 Mean Responses to 7-Point Issue Scales, Linked Versus NES Data Government Spending Defense Spending Government Jobs Government Aid to Blacks Linked NES Linked NES Linked NES Linked NES Race Latino Black White Gender Men Women Vote choice Bush Kerry N 53,232 1,6 53,877 1,61 56,384 1,13 54,459 1,73 Source: Authors compilation, based on the 24 National Annenberg Election Survey (Romer et al. 26) and the 24 National Election Study (Center for Political Studies 24).

17 Environment vs. Jobs Women s Role Government vs. Private Health Insurance Government Aid to Latinos U.S. Intervention Linked NES Linked NES Linked NES Linked NES Linked NES ,536 1,19 59,117 1,157 56,16 1,112 48, ,13 141

18 Table 2.5 distribution of Linked, Annenberg, and NES Data on a Common Issue Question Approve of the way the president is handling the economy (percent approving) Linked NES Annenberg Race Latino Black White Gender Men Women Aggregate N 61,948 1,121 84,122 Source: Authors compilation, based on the 24 National Annenberg Election Survey (Romer et al. 26) and the 24 National Election Study (Center for Political Studies 24).

19 Table 2.6 opinions on Social Welfare Issues, by Ethnic-Racial Group Whites Blacks Latinos Asians Federal spending on welfare programs Increase Keep the same Decrease Cut out entirely Investing social security in the stock market Favor Neither favor nor oppose Oppose Government should give parents in low-income families money to help pay for their children to attend a private or religious school instead of their local public school Favor Neither favor nor oppose Oppose N 4,88 7,36 2, Source: Authors compilation, based on the 24 linked data, as discussed in the text.

20 Figure 3.1 theoretical Expectations of Income and Education in Shaping Response to Elite Polarization High Low Education/High Income High Education/High Income Economic Security Awareness of Social Welfare Policy Cues Income Greatest increase in importance of cultural dimension Low Education/Low Income No increase in importance of either dimension Conflict extension : response on both dimensions High Education/Low Income Greatest increase in importance of scope-of-government dimension Low Low Education Ability to Discern Hard Issue Cues Reliance on Easy Issue Cues High Source: Figure generated by authors.

21 Figure 3.2 expected Impact of Issue Variables on Partisanship 3.5 Expected Impact on Partisanship Scale Scope-of-government preferences Cultural preferences Year Source: Authors compilation based on the General Social Surveys (Davis, Smith, and Marsden ). Note: All predictors have been scaled to a range of to 1. Lines represent the expected impact on partisanship of moving from the most conservative possible position to the most liberal on each issue dimension.

22 Figure 3.3 expected Impact of Issue Variables on Partisanship, by Education Level 4 Expected Impact on Partisanship Scale Scope-of-government preferences, high education Cultural preferences, low education Year Scope-of-government preferences, low education Cultural preferences, high education Source: Authors compilation based on the General Social Surveys (Davis, Smith, and Marsden ). Note: All predictors are scaled to a range of to 1. 26

23 Figure 3.4 expected Impact of Issue Variables on Partisanship, by Income Level 4 Expected Impact on Partisanship Scale Scope-of-government preferences, high income Cultural preferences, high income Year Scope-of-government preferences, low income Cultural preferences, low income Source: Authors compilation based on the General Social Surveys (Davis, Smith, and Marsden ). Note: All predictors are scaled to a range of to 1.

24 Figure 3.5 expected Impact of Issue Variables on Scope-of-Government Preferences Among High- and Low-Income Cohorts, by Education Expected Impact on Partisanship Scale High income, high education Low income, high education 199 Year High income, low education Low income, low education Source: Authors compilation based on the General Social Surveys (Davis, Smith, and Marsden ). Note: All predictors are scaled to a range of to 1.

25 Figure 3.6 expected Impact of Issue Variables on Cultural Preferences Among High- and Low-Income Cohorts, by Education Expected Impact on Partisanship Scale High income, high education Low income, low education 199 Year High income, low education Low income, high education Source: Authors compilation based on the General Social Surveys (Davis, Smith, and Marsden ). Note: All predictors are scaled to a range of to 1.

26 Table 3.1 Context-Dependent Predictors of Partisanship, Baseline Impact Economic/scope-of-government liberalism 1.31 * (.16) Cultural-issue liberalism.9 (.9) Real income (tens of thousands).61 * (.9) Ideological self-identification (7-point scale) 1.56 * (.13) Urban.7 (.5) Rural.11 (.6) Catholic.67 * (.5) Jewish 1.14 * (.13) Religious fundamentalist.28 * (.6) Black 1.6 * (.7) Female.1 *.4 Southern white.7 (.6) Polarization (in DW-Nominate scores).7 * (.32) R 2.2 Variable * Polarization 1.76 * (.33).78 * (.18).17 (.16) 1.99 * (.26).2 (.12).24 (.14).6 * (.11).44 (.33).46 * (.12).28 * (.14).26 * (.9).5 (.12) N 15,341 Source: Authors compilation based on the General Social Surveys (Davis, Smith, and Marsden ). Note: Table entries are OLS coefficients (cluster-corrected standard errors in parentheses). All predictors have been scaled to a range to 1. Baseline impact taken when polarization is at the lowest level. * p <.5, two-tailed tests

27 Table 3.2 Baseline and Context-Dependent Effect of Policy Preference Variables, by Income and Education Low Education High Education Low Income Middle Income High Income Scope-ofgovernment issues (Baseline) Scope-ofgovernment issues (Context effect) Cultural issues (Baseline) Cultural issues (Context effect) 1.15 * (.19) 1.41 * (.27).69 * (.3) 1.53 * (.27) 1.58 * (.28).84 * (.43) 2.24 * (.49) 1.13 * (.61) 1.51 * (.56) 2.28 * (.56).35 * *. (.11) (.14) (.17) (.15) (.15) 1. * * 1.24 *.66 * (.22) (.26) (.3) (.3) (.3) R 2 (full model) N 1,627 6,242 5,147 5,264 4,958 Source: Authors compilation based on the General Social Surveys (Davis, Smith, and Marsden ). Note: Table entries are OLS coefficients (cluster-corrected standard errors in parentheses). All predictors scaled to a range of to 1. All other variables included in table 3.3 models are included in these models, but not shown in this table. * p <.5, one-tailed tests

28 Table 3.3 Baseline and Context-Dependent Effect of Policy Preference Variables, by Education Level Within Income Low Income Middle Income High Income Low Education High Education Low Education High Education Low Education High Education Scope-of-government issues (baseline).94 * (.34) Scope-of-government issues (context effect).16 (.73) Cultural issues (baseline).12 (.21) Cultural issues (context effect).56 (.38).45 (.63) 3.9 * (1.17) R 2 (full model) (.35).73 (.64) N 3,937 1,196 3,42 1,836 2,236 2,716 Source: Authors compilation based on the General Social Surveys (Davis, Smith, and Marsden ). Note: Table entries are OLS coefficients (cluster-corrected standard errors in parentheses). All predictors scaled to a range of to 1. Demographic control variables included in these models, but not shown in this table. * p <.5, one-tailed tests.95 * (.33) 1.72 * (.76).58 * (.19) 1.4 * (.41) 2.4 * (.48).68 (.88).2 (.26).46 (.46) 1.37 * (.4) 2.46 * (.91).2 (.22) 1.19 * (.47) 1.64 * (.41) 2.16 * (.74).36 (.21).7 (.39)

29 Figure 4.1 state-level Variation in Presidential Election Results, 1972 to 28 Mean State Deviation from National Popular Vote 1% Unweighted by electoral vote Weighted by electoral vote Year Source: Authors compilation based on publicly available electoral data.

30 Figure 4.2 declining Scope of Competition in Presidential Elections, 1972 to 28 Percentage of Electoral Votes Cast by States Within Each Threshold of Deviation from the National Two-Party Popular Vote 1% Within plus or minus 1 percentage points Within plus or minus 5 percentage points Within plus or minus 3 percentage points Year Source: Authors figure based on publicly available electoral data.

31 Figure 4.3 potential State-Level Patterns Behind Rising Party Identification Vote Association Predicted Presidential Vote Panel 1 Each slope =.25 Average residual state difference =.3 Party Identification Predicted Presidential Vote Panel 2 Average slope =.25 Average residual state difference =.3 Party Identification Predicted Presidential Vote Panel Party Identification Each slope =.5 Average residual state difference =.3 Predicted Presidential Vote Panel 4 Party Identification Each slope =.5 Average residual state difference =.1 Source: Figure generated by authors.

32 Figure 4.4 Relationship Between Party Identification and Presidential Vote Across States Average Slope Coefficient Year Source: Authors compilation based on data from the National Election Studies (21).

33 Figure 4.5 Intercept Variation Across States.1 Standard Deviation of Intercepts Year 28 Source: Authors compilation based on data from the National Election Studies (21). Note: From random effects model predicting presidential vote index from PID.

34 Figure 4.6 Variation Across States on Presidential Index, and Percent of that Variation Explained by Party Identification (PID) Variation Across States on Presidential Index Variation across states Percent explained by PID Percent Explained by PID Year Source: Authors compilation based on data from the National Election Studies (21).

35 Figure 4.7 perceived Ideological Differences Between the Parties Across States and Time.7.6 Average Difference Year Source: Authors compilation based on data from the National Election Studies (21). Note: The thick line reflects the nine states with no trend in the link between PID and presidential vote. The thin line reflects the remaining states.

36 Figure 4.8 partisan Divide on the Cultural-Issues Index over Time.6 Average Standardized Score Democrats Republicans Year 28 Source: Authors compilation based on data from the National Election Studies (21). Figure 4.9 percentage Cross-Pressured on Cultural Issues 25 2 Democrats Republicans Percentage Year 28 Source: Authors compilation based on data from the National Election Studies (21).

37 Figure 4.1 effect of Party Identification Among Liberals and Conservatives on Cultural Issues Predicted Presidential Index Score Liberals Conservatives Year 28 Source: Authors compilation based on data from the National Election Studies (21).

38 Figure 4.11 effect of Cultural Issues on Party Loyalty in Presidential Elections, 1972 to 28 Predicted Presidential Index Score Predicted Presidential Index Score to 1984 Republican Independent Democrat Race, Gender, and Cultural-Issue Index (Liberal to Conservative) Republican Independent Democrat 1988 and Later Race, Gender, and Cultural-Issue Index (Liberal to Conservative) Source: Authors compilation based on data from the National Election Studies (21).

39 Table 4.1 National Election Studies (NES) Sample Sizes Within States Over Time State NES ID no Alabama Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Idaho 63 5 Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana 64 1

40 Nebraska Nevada New Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont 6 1 Virginia Washington West Virginia Wisconsin Wyoming Source: Authors compilation based on data from the National Election Studies (21). Note: Boldface indicates states included in this analysis.

41 Appendix 5.A descriptions of Groups Observed and Municipalities in Which They Met Municipality Description Group Type Municipality Population (2) Median Household Income, in Dollars (1999) Central hamlet Daily morning coffee klatch, local gas station (men) 5 38, Northern tourist loation North western hamlet Weekly breakfast group, local restaurant (women, primarily retired) Weekly morning coffee klatch, local church (mixed gender, primarily retirees) 5 32, 5 35, North central village Group of library volunteers at local library (mixed 5 34, gender, retirees); also, daily coffee klatch of male local leaders meeting in the local municipal building North eastern resort village Group of congregants after a Saturday evening service at 1, 41, Immanuel Lutheran Church (mixed gender) North western village Daily morning coffee klatch, local gas station (men) 1, 32, Northern American Indian reservation South central village Group of family members, during a Friday fish fry at a local gas station restaurant (mixed gender) Daily morning coffee klatch, local gas station (mixed gender, working and retired) 1, 35, 1,5 31, North central village Daily breakfast group, local diner (men) 2, 38, South central village Women s weekly morning coffee klatch at local diner; 3, 43, also, group of male professionals, construction workers, and retirees meeting later there Central western village Two daily morning coffee klatches, one at a local gas 3, 3, station, the other at a local diner (men) Central eastern village Kiwanis meeting (mixed gender, primarily retirees); also 3, 45, daily morning coffee klatch of male retirees at local fastfood restaurant Western Minneapolis suburb Daily morning coffee klatch, local diner (male localbusiness owners, lawyers, retirees) 9, 51,

42 South eastern city on northern edge of Milwaukee metropolitan area Daily morning coffee klatch, local diner (men) 1, 54, South central city Central city East central city Middle-aged man and woman taking a midmorning break at a local café Daily morning coffee klatch, local café (middle-aged professionals, mixed gender) Daily morning coffee klatch, local gas station (retired men) 1, 36, 38, 37, 42, 41, Milwaukee suburb, west of the city Group of teachers and administrators at local high school 47, 55, (mixed gender); daily lunch group of middle-aged men; mixed-gender breakfast group of retirees Western city Daily morning coffee klatch, local café (middle-aged 52, 31, professionals, retirees, mixed gender) South eastern city Weekly breakfast group, local diner (mixed gender, 82, 37, retirees, and currently employed) North eastern city Daily breakfast group, local diner (men) 1, 39, Madison Middle-aged female professionals book club; also, daily 2, 42, morning coffee klatch of male retirees at bakery; female resident volunteers in food pantry in low-income neighborhood North Milwaukee neighborhood AIDS/HIV activism group meeting after services in a 6, 32, Baptist church (mixed gender) South Milwaukee neighborhood Group of Mexican immigrants, waiting at a pro bono 6, 32, health clinic (mixed gender) Source: Authors compilation. Note: Population and income figures have been rounded to preserve the anonymity of the groups observed.

43 Table 6.1 Impact of Public Opinion Data on Domestic or Foreign Policy Positions (Domain Effect Model) Dependent Variable: Presidential Policy Positions Independent Variables Domestic Policy Foreign Policy Public s Ideological Identification.7 (1.8) 3.39** (1.55) Public s Policy Opinions 1.13** (.14).16 (.21) Presidential Policy Positions, t 1.73** (.2).76** (.2) Constant.44 (.65) 1.13 (.93) R N 1, Source: Authors compilation. Note: OLS coefficients, standard errors in parentheses. **p. 5, *p.1, one-tailed test.

44 Table 6.2 Impact of Policy Preferences of Independents on Domestic Policy Positions Independent Variables Dependent Variable: Presidential Policy Positions on Domestic Policy Public s Ideological Identification.14 (1.18) Public s Policy Opinions.68** (.23) Policy Opinions of Independents 1.6** (.22) Presidential Policy Positions, t 1.56** (.3) Constant.49 (.71) R 2.74 N 847 Source: Authors compilation. Note: OLS coefficients, standard errors in parentheses. **p.5, *p.1, one-tailed test.

45 Table 6.3 Impact of Policy Preferences of High-Income Americans on Economic Policy Positions Independent Variables Dependent Variable: Social Security Reform, Taxes, and Government Spending Public s Ideological Identification.45 (1.58) Public s Policy Opinions.7** (.38) Policy Opinions of Higher Income 4.6** Americans (.85) Presidential Policy Positions, t 1.5** (.7) Constant.79 (1.6) R 2.84 N 173 Source: Authors compilation. Note: OLS coefficients, standard errors in parentheses. **p.5, *p.1, one-tailed test.

46 Table 6.4 Impact of the Policy Preferences of Baptists and Catholics on Social-Conservative Policy Positions Independent Variables Dependent Variable: Family Values and Crime Public s Ideological Identification.1 (.53) Public s Policy Opinions.24 (.2) Policy Opinions of Baptists 1.38* (.85) Policy Opinions of Catholics.27 (.7) Presidential Policy Positions, t 1.11 (.1) Constant 4.78** R 2.7 N 14 Source: Authors compilation. Note: OLS coefficients, standard errors in parentheses. **p.5, *p.1, one-tailed test.

47 Table 6.5 Impact of Policy Preferences of Republicans on Defense- Spending Policy Positions Independent Variables Dependent Variable: Increased Defense Spending Public s Ideological Identification 3.31* (2.11) Public s Policy Opinions.32 (.43) Policy Opinions of Republicans 5.** (1.74) Policy Opinions of Independents 3.88** Policy Opinions of Democrats (1.51) 2.6* (1.29) Presidential Policy Positions, t 1.3 (.1) Constant 3.15** (1.57) R 2.15 N 9 Source: Authors compilation. Note: OLS coefficients, standard errors in parentheses. **p.5, *p.1, one-tailed test.

48 Figure 7.1 Comparison of Partisan Gap Versus Income Gap in Policy Liberalism 1.75 Partisan Gap & Liberalism (Self-Identified) Economic Policy Liberalism Partisan gap Social Policy Liberalism Income gap Source: Authors compilation based on the 2 and 24 Annenberg National Election Surveys (Romer et al. 26). Note: Bars represent the absolute value of the difference in liberalism scores for Democrats versus Republicans (partisan gap) and those with the lowest (less than $1,) and highest (more than $15,) incomes (income gap).

49 Figure 7.2 Income-Opinion Relationship Across Poor and Wealthy States Policy Liberalism Economic Issues (2) Economic Issues (24).6 Income Policy Liberalism Income Policy Liberalism Social Issues (2) Income Policy Liberalism Social Issues (24) Income Poor state ($3,) Wealthy state ($45,) Source: Authors compilation based on the 2 and 24 Annenberg National Election Surveys (Romer et al. 26). Notes: Lines represent the predicted policy liberalism for each income group, ranging from less than $1, (very left) to more than $15, (very right), estimated from multilevel models in which the intercept is estimated as a function of average state income and the slope is estimated as a function of both individual-level income (at level 1) and state-level income (as a cross-level interaction). Solid line presents predicted means for a poor state, such as West Virginia, which is two standard deviations below the mean state income (from Annenberg data), while the line with dots presents predicted means for wealthier states, such as Maryland, whose average state wealth is two standard deviations above the mean.

50 Figure 7.3 states Economic Policy Responsiveness to Different Income Groups Responsiveness Coefficient Responsiveness Coefficient 12 Panel A: All States Model 5 (*) Model 6 (*) Model 5 (*) Model 6 (*) Economic Policy (2) Economic Policy (24) 12 Panel B: Poor States and Rich States (*) 24 (*) 2 24 Poor States Rich States Low Middle High Source: Authors compilation based on the 2 and 24 Annenberg National Election Surveys (Romer et al. 26), as well as the policy measure presented in table 7.3. Notes: N = 48 states in panel A, N = 24 poor states and 24 wealthy states in panel B. Bars represent the coefficient for each group s policy liberalism in panel A, from model 5 (no controls) and model 6 (controlling for state wealth), and in panel B, from model 7 (poor states) and model 8 (rich states) controlling for state wealth. * p <.5, + = p <.1

51 Figure 7.4 states Social Policy Responsiveness to Different Income Groups Responsiveness Coefficient Responsiveness Coefficient Panel A: All States Model 5 (+) Model 6 (*) Model 5 (*) Model 6 (*) Social Policy (2) Social Policy (24) Panel B: Poor States and Rich States 2 24 (*) 2 (+) 24 (*) Poor States Rich States Low Middle High Source: Authors compilation based on the 2 and 24 Annenberg National Election Surveys (Romer et al. 26), as well as the policy measure presented in table 7.3. Notes: N = 48 states in panel A, N = 24 poor states and 24 wealthy states in panel B. Bars represent the coefficient for each group s policy liberalism in panel A, from model 5 (no controls) and model 6 (controlling for state wealth), and in panel B, from model 7 (poor states) and model 8 (rich states) controlling for state wealth. * p <.5, + = p <.1

52 Table 7.3 policy Indicators Used to Estimate States Economic and Social Policy Liberalism Data Source Year Mean SD Range Factor Loading Eigenvalue Proportion of Variance Economic policy indicators Eligibility for SCHIP, percent of FPL (1) Capital gains tax rate (2) Corporate income tax rate (2) Minimum wage (3) $.83 $5.15 $ Per pupil expenditures in K 12 (4) 26 9,75 $2,16 $5,437 $14, Income eligibility for TANF/ welfare (5) $37 $269 $1,59.22 Health insurance mandates index (3) Economic policy score Social policy indicators Gun control index (3) Abortion index (3) Has death penalty (3) No discrimination for sexual orientation (6) Requires (2) or allows (1) school prayer (7) Social policy score Source: Authors compilation of data from Kaiser Family Foundation (26); Institute on Taxation and Economic Policy (23); Soren, Muedini, and Ruger (28); Education Week (28); Urban Institute (26); Human Rights Watch (28); and Education Commission of the States (28) Note: N = 48.

53 Table 7.4 Income Differences in Ideology Versus Policy Liberalism Scores N 2 Annenberg Survey 24 Annenberg Survey Self-ID Liberalism Economic Policy Liberalism Social Policy Liberalism N Self-ID Liberalism Economic Policy Liberalism Social Policy Liberalism Less than $1K 3, , $1K to $15K 3, , $15K to $25K 7, , $25K to $35K 8, , $35K to $5K 11, , $5K to $75K 11, , $75K to $1K 6, , $1K to $15K 4, , More than $15K 2, , F Source: Authors calculations from the 2 and 24 Annenberg National Election Surveys (Romer et al. 26). Notes: All three measures of liberalism are standardized scores with M = and SD = 1.

54 Table 7.5 economic Policy Responsiveness to Different Income Groups All States Poor Rich All States Poor Rich (1) (2) (3) (4) (5) (6) (7) (8) Economic policy liberalism (2) Opinion: average 5.62** 4.49** 3.22** 5.62** (.78) (.8) (1.5) (1.2) State wealth.79** * (.26) (.56) (.53) (3.9) (3.36) (7.9) Opinion: low (1.9) (2.5) (2.17) (5.31) Opinion: middle 6.98** 8.78** 7.77* 7.44 (2.12) (2.47) (3.14) (5.4) Opinion: high 6.36** ** 1.64 (2.21) (2.6) (3.25) (5.7) Percent low income (4.2) (11.2) (14.2) (24.2) Percent high income (4.11) (7.72) (8.98) (18.) Constant ** * (.9) (1.3) (2.67) (2.79) (2.49) (18.1) (2.9) (4.5) Observations R

55 Economic policy liberalism (24) Opinion: average 6.17** 4.81** 3.23* 6.24** (.94) (.96) (1.24) (1.44) State wealth * (.27) (.59) (.54) (2.44) (2.74) (6.7) Opinion: low (3.23) (3.14) (3.3) (8.19) Opinion: middle 8.86** 8.88** 8.4* 1.36 (3.1) (3.) (3.4) (8.35) Opinion: high 7.82** 8.6** (2.8) (2.73) (4.9) (5.96) Percent low income (4.18) (9.7) (11.7) (22.3) Percent high income (4.4) (6.48) (9.11) (16.8) Constant ** * * (.9) (1.36) (2.81) (2.87) (2.51) (14.3) (16.9) (38.1) Observations R Source: Authors compilation based on analysis of the 2 and 24 Annenberg National Election Surveys (Romer et al. 26), as well as the policy measure presented in table 7.3. Notes: N = 48 states, excluding Alaska and Hawaii. Coefficients from OLS regression models. Group opinion measures are weighted for the proportion of each grop within each state.

56 Table 7.6 social Policy Responsiveness to Different Income Groups All States Poor Rich All States Poor Rich (1) (2) (3) (4) (5) (6) (7) (8) Economic policy liberalism (2) Opinion: average 3.9** 3.51** 3.47** 3.79** (.39) (.55) (.8) (.79) State wealth (.33) (.73) (.51) (2.67) (3.23) (7.73) Opinion: low (2.42) (2.43) (2.98) (5.89) Opinion: middle * (2.2) (2.23) (3.49) (4.28) Opinion: high (3.68) (3.69) (7.46) (5.19) Percent low income (5.16) (1.1) (14.2) (27.5) Percent high income (5.13) (7.6) (12.6) (16.5) Constant (.8) (1.7) (3.51) (2.66) (3.3) (15.7) (19.7) (45.4) Observations R (Table continues on p. 214.)

57 Table 7.6 (Continued) All States Poor Rich All States Poor Rich (1) (2) (3) (4) (5) (6) (7) (8) Economic policy liberalism (24) Opinion: average 3.49** 3.92** 4.37* 3.79** (.42) (.59) (.85) (.88) State wealth (.32) (.64) (.53) (2.46) (2.94) (6.8) Opinion: low (2.9) (2.93) (4.41) (5.78) Opinion: middle (3.3) (3.35) (5.) (5.94) Opinion: high 13.1** 13.3** 12.8* 17.44* (3.93) (4.) (5.97) (6.51) Percent low income (4.12) (9.16) (13.) (2.6) Percent high income (4.6) (6.47) (1.2) (15.6) Constant (.8) (1.64) (3.1) (2.78) (2.44) (14.5) (18.4) (34.7) Observations R Source: Authors compilation based on analysis of the 2 and 24 Annenberg National Election Surveys (Romer et al. 26), as well as the policy measure presented in table 7.3. Notes: N = 48 states, excluding Alaska and Hawaii. Coefficients from OLS regression models. Group opinion measures are weighted for the proportion of each grop within each state. ** p <.1, * p <.5, + p <.1

58 Table 7.7 summary of Findings Smaller income differences Larger income differences Source: Authors compilation. More Balanced Responsiveness Economic policymaking in rich states Social policymaking in rich states More Skewed Responsiveness Social policymaking in poor states Economic policymaking in poor states

59 Table 7.1 policy Issue Items Used to Generate Economic and Social Liberalism Scales Liberal Position (pre-imput) Less More Than Than Raw N Range $1, $15, F Economic policy liberalism (2) Inheritance tax should be cut (Q113a & Q113b) 18, % 75% Should spend on health care for uninsured (Q111b) 55, Should spend on Medicare (Q111g) 24, Favor universal health care for children (Q91d) 29, Should spend on Medicaid (Q111h) 24, Should reduce income differences (Q136e) 23, Should spend on aid to mothers with young children (Q111e) 24, Social policy liberalism (2) Favor restricting abortion (Q91b & Q38c) 54, Should ban abortion (Q136b) 24,

60 Economic policy liberalism (24) Favor eliminating estate tax (Q48 & Q74 & Q75) 13, Favor spending more on health insurance (Q38) 19, Favor health insurance for children (Q62 & Q77) 19, Favor health insurance for workers (Q63 & Q78) 18, Favor assistance to schools (Q22) 28, Should reduce income differences (Q22) 35, Social policy liberalism (24) Favor banning all abortions (Q2) 56, Favor banning all late-term abortions (Q25 & Q26) 22, Favor stem cell funding (Q65 & Q66 & Q83 & Q84) 16, Favor marriage ammendment (Q17) 55, Favor allowing same sex marriage (Q656 & Q657) 17, Favor gun control 31, Source: Authors calculations from the 2 and 24 Annenberg National Election Surveys (Romer et al. 26). Note: Descriptive statistics from datasets prior to imputing for missing values. Income-group responses indicate the percent of each income group selecting the most liberal response option for each item. F statistics drawn from one-way ANOVA tests across all nine income categories. F statistics for each question are significant at p <.5

61 Table 7.2 economic and Social Policy Liberalism Individual Level State Level N Mean SD Min Max N Mean SD Min Max All respondents Economic (2) 59, Economic (24) 72, Social (2) 59, Social (24) 72, Low income: less than $35, Economic (2) 23, Economic (24) 24, Social (2) 23, Social (24) 24, Middle income Economic (2) 22, Economic (24) 26, Social (2) 22, Social (24) 26, High income: more than $75, Economic (2) 12, Economic (24) 21, Social (2) 12, Social (24) 21, Source: Authors calculations from the 2 and 24 Annenberg National Election Surveys (Romer et al. 26).

62 Table 8.1 predicting Senate Roll-Call Ideology from Mean State Ideologies, 11st to 13rd Congresses Mean ideology for voting-age population Republican senator Intercepts Standard error of regression Adjusted R 2 Mean Ideology = 1 to +1 Scale 1.41*** (.24).95*** (.4) Congress-specific Mean Ideology = 1 to 7 Scale.47*** (.8).95*** (.4) Congress-specific N Source: Authors compilation based on Poole and Rosenthal (1997) and Miller et al. (1993). Note: Dependent variables in both regressions are senator-specific W-nominates. The coefficients are the unstandardized regression coefficients. Standard errors clustered by senator in parentheses

63 Table 8.2 predicting Senate Roll-Call Ideology from Income-Specific Ideologies, 11st to 13rd Congresses Bartels Mean Ideology = 1 to +1 Scale Replication, Mean Ideology = 1 to +1 Scale Replication, Mean Ideology = 1 to 7 Scale Wgt. low-income ideology ( X L P L ) Wgt. middle-income ideology ( X M P M ) Wgt. high-income ideology ( X H P H ) Republican senator dummy Intercepts Standard error of regression Adjusted R 2.33 (.44) 2.66*** (.6) 4.15*** (.85).95*** (.4) Congressspecific (.41) 2.52*** (.53) 4.91*** (.72).92*** (.4) Congressspecific *** (.9).43*** (.13).5*** (.14).96*** (.4) Congressspecific N 33 Source: Authors compilation based on Poole and Rosenthal (1997) and Miller et al. (1993). Note: Dependent variables in all regressions are senator-specific W-nominates. Wgt. lowincome ideology, wgt. middle-income ideology, and wgt. high-income ideology are the raw mean ideologies for the respective income groups times the proportion of that group. The coefficients are the unstandardized regression coefficients. Standard errors clustered by senator in parentheses

64 Table 8.3 predicting Senate Roll-Call Ideology from Income-Specific Ideologies, 11st to 13rd Congresses Wgt. low-income ideology ( X L P L ) Wgt. middle-income ideology ( X M P M ) Wgt. high-income ideology ( X H P H ) Republican senator dummy Proportion low-income (P L ) Proportion high-income (P H ) Intercepts Standard error of regression Adjusted R 2 Replication, Mean Ideology = 1 to +1 Scale 1.6** (.39) 2.26*** (.56) 4.58*** (.75).92*** (.4).75 (.39).14 (.35) Congress-specific Replication, Mean Ideology = 1 to 7 Scale.35** (.13).75*** (.19) 1.52*** (.25).92*** (.4) 5.18*** (1.3) 2.97* (1.35) Congress-specific N Source: Authors compilation based on Poole and Rosenthal (1997) and Miller et al. (1993). Note: Replicated results with proportions added. Dependent variables in both regressions are senator-specific W-nominates. Wgt. low-income ideology, wgt. middle-income ideology, and wgt. high-income ideology are the raw mean ideologies for the respective groups times the proportion of that group. Proportion low-income and proportion high-income denotes the proportions entered separately. The coefficients are the unstandardized regression coefficients. Standard errors clustered by senator in parentheses. *** p <.1, **.1 < p <.1, *.1 < p <

65 Table 8.4 predicting Senate Roll-Call Ideology from Income-Specific Ideologies, Defined Statewise, 11st to 13rd Congresses Low-income ideology ( X L ) Middle-income ideology ( X M ) High-income ideology ( X H ) Republican senator dummy Intercepts Standard error of regression Adjusted R 2 Mean ideology = 1 to +1 scale.21 (.17).57* (.26) 1.24*** (.22).94*** (.4) Congress-specific Mean ideology = 1 to 7 scale.7 (.6).19* (.1).41*** (.7).94*** (.4) Congress-specific N Source: Authors compilation based on Poole and Rosenthal (1997) and Miller et al. (1993). Note: Dependent variables in both regressions are senator-specific W-nominates. Low-income ideology, middle-income ideology, and high-income ideology are the mean ideologies for each group where the group is defined statewise (one-third in each state), not nationally. The coefficients are the unstandardized regression coefficients. Standard errors clustered by senator in parentheses. *** p <.1, **.1 < p <.1, *.1 < p <

66 Table 8.5 predicting Senate Roll-Call Ideology from Mean State Ideologies, 16th to 18th Congresses (Annenberg Study Data) Mean ideology for voting-age population ( 1 to +1 scale) Republican senator Intercepts Standard error of regression Adjusted R 2 N 1.99*** (.35) 1.31*** (.4) Congress-specific Source: Authors compilation based on Poole and Rosenthal (1997) and Romer et al. (26). Note: Dependent variables are both senator-specific W-nominates. The coefficients are the unstandardized regression coefficients. Standard errors clustered by senator in parentheses. *** p <.1, **.1 < p <.1, *.1 < p < Table 8.6 predicting Senate Roll-Call Ideology from Income-Specific Ideologies, 16th to 18th Congresses, Defined Nationally (Annenberg Study Data) Wgt. low-income ideology ( X L P L ) Wgt. middle-income ideology ( X M P M ) Wgt. high-income ideology ( X H P H ) Republican senator dummy Proportion low-income (P L ) Proportion high-income (P H ) Intercepts Standard error of regression Adjusted R 2 Mean Ideology = 1 to +1 Scale 1.2 (1.14) 2.6 (1.99) 3.72* (1.57) 1.3*** (.5).2 (.79).56 (.82) Congress-specific N Source: Authors compilation based on Poole and Rosenthal (1997) and Romer et al. (26). Note: Dependent variables are senator-specific W-nominates. Wgt. low-income ideology, wgt. middle-income ideology, and wgt. high-income ideology are the raw mean ideologies for the respective income groups times the proportion of that group. The groups are defined nationally. Proportion low-income and proportion high-income denotes the proportions entered separately. The coefficients are the unstandardized regression coefficients. Standard errors clustered by senator in parentheses. *** p <.1, **.1 < p <.1, *.1 < p <.5 291

67 Table 8.7 predicting Senate Roll-Call Ideology from Income-Specific Ideologies, 16th to 18th Congresses, Defined Statewise (Annenberg Study Data) Low-income ideology ( X L ) Middle-income ideology ( X M ) High-income ideology ( X H ) Republican senator dummy Intercepts Standard error of regression Adjusted R 2 Mean Ideology = 1 to +1 Scale.59 (.41).4 (.62) 1.14* (.5) 1.31*** (.4) Congress-specific Mean Ideology = 1 to +1 Scale EIVREG 1.16 (.76).95 (.96) 1.58* (.71) 1.3*** (.4) Congress-specific N Source: Authors compilation based on Poole and Rosenthal (1997) and Romer et al. (26). Note: Dependent variables are senator-specific W-nominates. Low-income ideology, middle-income ideology, and high-income ideology are the mean ideologies for each group where the group is defined state-wise (one-third in each state), not nationally. The coefficients are the unstandardized regression coefficients. Standard errors in parentheses are clustered by senator in column 1. Because the Eivreg procedure in STATA does not allow for clustering, we also estimated the model in column 2 with only one observation per senator/cluster. That is, the dataset was collapsed at the individual senator level to preclude statistical dependence due to senators holding office in multiple sessions. This did not alter the results substantively. *** p <.1, **.1 < p <.1, *.1 < p <

68 Table 8.8 predicting Senate Roll-Call Ideology from Income-Specific Ideologies, Defined Nationally (24 Exit Poll Data) 1st Dimension of DW- Nominates 2nd Dimension of DW- Nominates Composite Measure Wgt. low-income ideology ( X L P L ) Wgt. middle-income ideology ( X M P M ) Wgt. high-income ideology ( X H P H ) Republican senator dummy Proportion low-income ( P L ) Proportion high-income ( P H ) Intercept Standard error of regression Adjusted R * (.99) 1.61* (.62).47 (.61).79*** (.4).22 (.63).4 (.5).68* (.3) (2.8) 1.7 (1.31) 1.59 (1.29).55*** (.7) 1.4 (1.33).43 (1.6).8 (.64) ** (1.2) 1.47* (.65).6 (.64).44*** (.4).43 (.66).18 (.52).35 (.71) N 11 Source: Authors compilation based on Poole and Rosenthal (27) and Edison Mitofsky Research (24). Note: Dependent variables are different versions of senator-specific DW-nominates. The composite measure is.74 times the 1st dimension score plus.26 times the 2nd dimension score. Wgt. low-income ideology, wgt. middle-income ideology, and wgt. high-income ideology are the raw mean ideologies for the respective income groups times the proportion of that group. The groups are defined nationally. Proportion low-income and proportion high-income denotes the proportions entered separately. The coefficients are the unstandardized regression coefficients. Standard errors in parentheses. *** p <.1, **.1 < p <.1, *.1 < p <

69 Table 8.9 Influence of General Opinion on Three Versions of DW- Nominates (24 Exit Poll Data) 1st Dimension of DW- Nominates 2nd Dimension of DW- Nominates Composite Measure Mean ideology Republican senator dummy Intercept Standard error of regression Adjusted R 2.79*** (.14).78*** (.3).46*** (.2) *** (.32).51*** (.7).8 (.5) *** (.16).42*** (.4).39*** (.3) N Source: Authors compilation based on Poole and Rosenthal (27) and Edison Mitofsky Research (24). Note: Dependent variables are different versions of senator-specific DW-nominates. The composite measure is.74 times the 1st-dimension score plus.26 times 2nd-dimension score. The coefficients are the unstandardized regression coefficients. Standard errors in parentheses. *** p <.1, **.1 < p <.1, *.1 < p <.5