washington, dc

The Democratic Strategist

Political Strategy for a Permanent Democratic Majority

Brace Yourself for Arcane Methodological Details

by Scott Winship
At some point, I am going to rue the day that we decided to call this blog “The Daily Strategist.” Maybe we should have gone with “The Occasional Strategist” or “The Sometime Strategist”. Perhaps “I’ve Got Your $*^#@ Strategy Right Here!”
Oh dear, pardon my French. Anywho, I promised I would return to my post from Monday examining the number of liberals and conservatives. This is going to be a longer and very Mystery Pollster kind of post today. But even if you don’t want the arcane stuff, there are some interesting findings here on public opinion in a number of areas.
The biggest question I received from folks was how I defined the four issue areas I created: security and foreign policy; values; economic and social policy; and fiscal policy. You may recall that I defined people as liberal or conservative on a number of individual items from the 2004 National Election Study, gave weights to the items depending on how well they predicted the presidential vote, and then defined people as liberal or conservative depending on whether their (weighted) liberal responses outnumbered their (weighted) conservative responses.
Here is a summary of the issue area components, with weights included and a breakdown of how many people answered one way or another on the individual items:
Security and foreign policy:
     1. human rights (weight=1.4)
          • liberal = promoting human rights is a very important foreign policy goal (43%)
          • conservative = somewhat important or not important at all (57%)
     2. support for the United Nations (weight=1.7)
          • liberal = strengthening the U.N. is a very important foreign policy goal (48%)
          • conservative = somewhat important or not important at all (52%)
     3. democracy promotion (weight=1.3)
          • liberal = promoting democracy is somewhat important or not important at all as a
               foreign policy goal (78%)
          • conservative = very important (22%)
     4. neoliberalism (weight=1.4)
          • liberal = promoting market economies is somewhat important or not important at all
               as a foreign policy goal (76%)
          • conservative = very important (24%)
     5. soft vs. hard power (weight=4.2)
          • liberal = prefer diplomacy and international pressure to threatening the use of force
               (i.e., placed self as being between 1 and 3 on a 7-point scale) (39%)
          • conservative = prefer threatening the use of force (i.e., placed self as being
               between 5 and 7) (39%)
     6. military strength (weight=1.6)
          • liberal = very or somewhat important for the US to have a strong military, or not
               important at all (43%)
          • conservative = extremely important (57%)
     7. war on terror (weight=6.3, decreased from 23.7)
          • liberal = disapprove of Bush’s handling of the war on terror (44%)
          • conservative = approve (56%)
OK, let’s pause for some comments here. First let me say something about the weights. These were produced using what’s known as logistic regression. Essentially I predicted whether a person voted for Bush or for Kerry using these seven items. The weights are “odds ratios”. So the weight of 4.2 for “soft vs. hard power” indicates that the odds of voting for Bush rather than Kerry were 4.2 times higher if a person leaned toward military force than if she leaned toward diplomacy, controlling for the effect of the other six items. Intuitively, I’m giving items greater weight the more important they are in predicting how people voted.
I arbitrarily lowered the weight on the war on terror item to be 1.5 times the next-highest weight because otherwise this single question completely determined whether one was coded as a liberal or conservative on security and foreign policy. That’s worthy of reflection – the “effect” of a person’s response to this item on her vote is in some sense over five times the effect of the next-most important item.
The war on terror item itself isn’t ideal – it reflects both attitudes toward Bush’s policies and attitudes toward his competence. But this is the best I could do given the questions available, though I am exploring other approaches. As for other items, it is perhaps problematic to define the liberal and conservative positions on democracy promotion, but since the weight on this item is so small, it doesn’t really have any effect on the results.
Finally, note that some people are neither coded as liberal nor conservative on the hard vs. soft power item. That’s because they could give a neutral response (4 on a scale from 1 to 7). When a person didn’t have liberal/conservative scores on every item in an issue area, I compared weighted liberal and conservative responses for the items she did have a score on. I required that the person have scores on a majority of the items, otherwise she received no overall liberal/conservative designation for that issue area.
Next: values:
     1. Constancy of moral values (weight=2.8)
          • liberal = we “should adjust our moral views to a changing world” (agree strongly or
               somewhat) (47%)
          • conservative = should not (disagree strongly or somewhat) (43%)
     2. Feminists vs. fundamentalists (weight=1.6)
          • liberal = feel more warmth (on a scale of 0-100) toward feminists than toward
               Christian fundamentalists (35%)
          • conservative = feel more warmth toward fundamentalists (41%)
     3. Legality of abortion (weight=1.6)
          • liberal = “a woman should always be able to get an abortion as a matter of personal
               choice” (36%)
          • conservative = a woman should only be able to get an abortion in the case of rape,
               incest, or danger to her life (if at all) (46%)
     4. Public funding for abortion (weight=1.8)
          • liberal = favor a state law providing public funds for poor women to have abortions
          • conservative = oppose (62%)
     5. Gay marriage (weight=2.4)
          • liberal = “same-sex couples should be allowed to marry” (or get civil unions if this
               option was volunteered by the respondent) (36%)
          • conservative = “should not be allowed to marry” (61%)
     6. Job protection for gays (weight=1.0, increased from 0.9)
          • liberal = favor “laws to protect homosexuals from job discrimination” (74%)
          • conservative = oppose (26%)
     7. Patriotism (weight=1.5)
          • liberal = love for country is very, somewhat, or not very strong (45%)
          • conservative = extremely strong (55%)
     8. Capital Punishment (weight=5.2)
          • liberal = oppose the death penalty (29%)
          • conservative = favor (71%)
     9. Gun control (weight=2.1)
          • liberal = “the federal government should make it more difficult to buy a gun” (55%)
          • conservative = should not (45%)
Note that one can be neither liberal nor conservative on the abortion item if she believes that abortion should be allowed in cases other than rape, incest, or danger to the mother’s life, but only once the need for an abortion has been established. I increased the weight on the gay job discrimination item because I didn’t want any of the weights to be less than one.
Economic and social policy:
     1. Government activism vs. individualism (weight=3.1)
           • liberal = the federal government “should see to it that everyone has a job and a
               good standard of living” (34%)
          • conservative = it “should just let each person get ahead on their own” (47%)
     2. Jobs vs. environment (weight=1.9)
          • liberal = we “should protect the environment even if it costs some jobs or reduces
               living standards” (scored self as 1-3 on a 7-point scale) (45%)
          • conservative = “protecting the environment is not as important as maintaining jobs
               and living standards” (scored self as 5-7) (27%)
     3. Illegal immigration (weight=1.6)
          • liberal = “controlling and reducing illegal immigration” is only somewhat important
               or not at all important as a foreign policy goal (42%)
           • conservative = very important (58%)
     4. Unions vs. corporations (weight=5.5)
          • liberal = feelings toward labor unions are warmer than feelings toward big business
               (on a scale from 0 to 100) (44%)
          • conservative = feelings toward big business are warmer (36%)
     5. Poor people vs. business people (weight=1.5)
          • liberal = feelings toward poor people are warmer than feelings toward business
               people (41%)
          • conservative = feelings toward business people are warmer (26%)
     6. Opportunity (weight=6.0)
          • liberal = agree that “one of the big problems in this country is that we don’t give
               everyone an equal chance” (strongly or somewhat agree) (49%)
          • conservative = disagree (strongly or somewhat) (34%)
     7. Civil rights enforcement in employment (weight=2.8)
          • liberal = the federal government “should see to it that black people get fair
               treatment in jobs” (54%)
          • conservative = it “is not the federal government’s business” (43%)
     8. Outsourcing (weight=1.3)
          • liberal = the federal government should discourage U.S. companies from “hiring
               workers in foreign countries to replace U.S. workers” (66%)
          • conservative = it should encourage them or do nothing (34%)
     9. Health care system (weight=1.4)
          • liberal = a government insurance plan should “cover all medical and hospital
               expenses” (placed self as 1-3 on a 7-point scale) (47%)
          • conservative = all such expenses should be “paid by individuals through private
               insurance plans” (5-7 on a 7-point scale) (34%)
     10. Social security privatization (weight=3.1)
          • liberal = oppose allowing people to “put some of their Social Security payroll taxes
               into personal retirement accounts” (25%)
          • conservative = favor (43%)
     11. School vouchers (weight=1.9)
          • liberal = oppose having the government “give vouchers to low-income families so
               their kids may attend private or religious school instead of local public
               school” (67%)
          • conservative = favor (31%)
Note that respondents can be neither liberal nor conservative on most of these items.
Finally, fiscal policy:
     1. Expansion of government (weight=1.7)
          • liberal = “government has become bigger because the problems we face have
               become bigger” (59%)
          • conservative = government has become bigger “because it has gotten involved in
              &nbspthings that people should do for themselves” (41%)
     2. Current role of government (weight=1.5)
          • liberal = “there are more things that government should be doing” (58%)
          • conservative = “the less government the better” (42%)
     3. Taxes vs. spending vs. deficit reduction (weight=1.5)
          • liberal = against cutting domestic spending, against tax cuts, and for increasing
               domestic spending (25%)
          • conservative = either a) against increasing spending and for either cutting spending
               orfurther tax cuts, or b) for leaving taxes, spending, and the deficit as is (31%)
     4. Spending vs. services (weight=2.5)
          • liberal = want more government services even if it means more spending (39%)
          • conservative = unsupportive of more government services if it means more
              &nbspspending (61%)
     5. Tax progressivity (weight=3.9)
          • liberal = a) agree that either one’s own taxes or those of rich are too low and
              &nbspb) disagree that the taxes of the rich are too high and that the taxes of the poor
               are too low (61%)
          • conservative = a) disagree that one’s own taxes and those of the rich are too low
               and b) agree that the taxes of the poor are too low, agree that either one’s own
               taxes or those of the rich are too high, or supported the Bush tax cuts and agree
               that one’s own taxes or those of the rich are about right (29%)
     6. Spending on social programs (weight=3.3)
          • liberal = increase spending on public education, child care, and aid to the poor
          • conservative = decrease spending on one of the three or leave it as is on two of the
               three (38%)
These items are perhaps the most arbitrary, and a couple of them leave many people with no liberal/conservative designation. Incidentally, I separated fiscal policy from economic and social policy because people may, for instance, take conservative positions on the latter but be in favor of progressive taxes or increased spending. Or they may philosophically favor liberal positions on economic and social policies but in practice be unwilling to pay for them (or may disagree about how to distribute the costs).
Once I had everyone coded on these four issue areas (with some people missing one or more codes due to nonresponse), I compared the weighted number of liberal and conservative codes for the four areas to get an overall classification as being operationally liberal or conservative. The weights were again based on logistic regression and again are odds ratios. Security and foreign policy received the highest weight – 10.1, which was actually reduced from 14.5 to equal 1.5 times the next-highest weight. That next-highest weight was for economic and social policy (6.7). Values came next (5.4) followed by fiscal policy (2.7).
And that’s all I have to say about that. Simple, huh? I am certain that this won’t be the last set of estimates I produce, so I would gladly accept feedback and suggestions for improvement. Go nuts….

3 comments on “Brace Yourself for Arcane Methodological Details

  1. Chris Glaze on

    Interesting analysis. I would take this to the next level of geekiness by doing clustering analysis (more on that below), which I would argue is actually essential to how one should interpret these data.
    But first, there’s an issue with including the approval of Bush’s handling of terrorism question, and conclusions about overall public opinion on foreign policy may be tainted by it. To begin with, whether a respondent approved of Bush’s terrorism policies in 2004 was very related to how they prioritized diplomacy over military intervention (t-test, p=0). So by including both questions you are double-counting a shared factor, which I would argue is actually the intervention question itself, a pretty direct measure of hawkish-dovish attitudes. On that question, the NEC data on their website indicate that 34% of valid respondents rated themselves 5-7 (hawkish), and 44% rated 1-3 (dovish). Perhaps your numbers are different because of who you had to exclude.
    Inclusion of the approval question is also problematic because it is also related to perceptions of Bush (at the time), which I would hypothesize are influenced by how conservative one is to begin with. So the question may really be a measure of (1) how much one values diplomacy, the first question, (2) how effective Rove’s strategy was in playing up Bush as a “strong” leader, and (3) opinions on other issues. What you really want to due is extract (3), which the other questions in that category may do, I don’t know. This kind of general weighting scheme works especially well if you can assume that responses to each question are not significantly related to each other (linear independence, for those who speak math).
    Simply put: I would not conclude that Americans are “hawkish” from these data.
    On clustering: I did a rough analysis of the same dataset, using gay marriage as a proxy for social conservatism, the military intervention question for hawkishness, and either support for increases on spending and services or support for raising taxes on the rich for fiscal outlook. I did 3-D plots of subset sizes and color coded by how each group voted for Kerry, how Democratic each group is and lib-con self-ratings.
    What I see is a number of untapped voters who are economically populist, but are social conservative, and neutral on foreign policy: they support raising taxes on the rich, neutral on increasing government services, oppose gay marriage and in the middle of the diplomacy-military scale, and vote Republican. On most of these plots this group looks to be about 10% of the electorate, although there’s a lot of error in these estimates because of sample sizes. This is not my job, but if you aren’t already doing this kind of clustering analysis perhaps it would be helpful to see if these conclusions hold and how nuanced opinions on social issues are in this group.
    If the conclusions do hold, you need to figure out how to include these voters in the Democratic coalition. However, as indicated by a lot of frustration expressed in blogs (and in parts of this site), you risk alienating your base if you go too far to the right. Fortunately, social issues like gay marriage may not be as intractable as they seem. Obama sounds to me like he’s on the right track here, in his message that one need not be an evangelical conservative to have American values, and it would be hypocritical for liberal secular voters (myself included) to think there’s something wrong with being religious, it’s not exactly an open-minded attitude. Perhaps the DP should consider making (constitutionally correct) faith-based programs a big part of their social policy, not only for votes but because they may actually make sense given the role churches play in many communities.
    In sum, appealing to economically populist voters may actually require attention to issues in other areas.


Leave a Reply

Your email address will not be published. Required fields are marked *

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.