Arecent issue of Perspectives on Politics includes two articles that focus on the federal tax cuts of 2001.1
Such attention is merited given the controversy surrounding the policy. Conservative and liberal opinion leaders disagreed about the motivation for—and long-term consequences of—the tax cut. Some claimed that the cut was nothing more than a scheme to coddle the rich while others saw it as a means for promoting economic growth.In “Homer Gets a Tax Cut: Inequality and Public Policy in the American Mind,” Larry Bartels (2005: 20) uses the 2002 American National Election Studies (NES) to examine public support for this policy.2
Bartels 2005, 20.
Bartels' conclusion about the “the public mind” is not charitable. The “Homer” to whom citizens are compared in “Homer Gets a Tax Cut” is Homer Simpson, the cartoon überdolt who Wikipedia describes as simple-minded. Bartels, in turn, characterizes the opinions of “ordinary people” as being superficial and based on “simple-minded and sometimes misguided considerations of self interest.”3
Ibid., 21.
Finally, and most importantly, better-informed respondents were much more likely to express negative views about the 2001 tax cut…. If we take this crosssectional difference in views as indicative of the effect of information on political preferences, it appears that the strong plurality support for Bush's tax cut … is entirely attributable to simple ignorance.4
Ibid., 24.
In what follows, we report results from an analysis of the same data. Our findings and conclusions are quite different than those attributed to voters in “Homer.” Like Bartels, we agree that it is important to understand why voters support or oppose particular policies and the role that information plays in such assessments. We also recognize that isolating the role of information in citizens' opinions is no simple task. For if an analyst wishes to claim that a particular opinion is “entirely attributable to simple ignorance,” he or she must be able to compare observed opinions to what people would have opined if they were not “simply ignorant.” Bartels measures the “effect of political information by comparing the views of better and worse informed respondents in the NES survey using a measure of political information based on interviewers' ratings of respondents.”5
Ibid., 23.
The 2002 NES was run on a smaller budget than those of the presidential election years. It was phone-based and much shorter than other NES surveys. As a result, many common NES questions were not asked in 2002, including well-known “political information” questions. Like Bartels, we use post-election interviewer ratings as an instrument for pre-election interviewer ratings in our analysis to minimize the probability that the rating is contaminated by a respondent's answers to the tax cut and inequality questions, which were asked in the pre-election survey.
In addition to presenting our result, we explain why Bartels achieved different results than we did. The explanation is not complicated. Bartels makes a very strong assumption about how information affects public opinion. He restricts all respondents—whether liberal or conservative, Republican or Democrat—to respond to increasing information levels in an identical way. Must changes in information levels have such universal effects? Or is it possible that on issues such as the merit of a tax cut proposal, reasonable people can disagree?
To see why Bartels' assumption is problematic, suppose for a moment that we gave respondents more information about the tax cuts. Suppose, moreover, that this additional information revealed that the tax cuts would spur long-term economic growth at the expense of greater economic inequality in the short-run. Bartels' analysis requires liberals and conservatives (and Republicans and Democrats) to react to this information in the same way. Even if the truth or relevance of additional information is contested (e.g., if leading economists disagree about how a tax cut affects economic growth), or if the matters on which experts agree involve a tradeoff between competing social goals, Bartels proceeds as though all respondents must respond identically to the added information. In this respect, he writes as if there is only one right answer—only one conclusion about the Bush tax cuts that a highly-informed respondent can reach. But often in politics, and perhaps distinctively in politics, there is a clash of perspectives, values, and views about how society works.7
We thank Paul Sniderman for suggesting this wording to us.
Our analysis differs from that in “Homer” in that we allow people of different partisan or ideological identities to react to higher information levels in different ways. In so doing, we can test Bartels' assumption. While our study could have verified that higher information ratings affect the tax cut opinions of all political ideologues or partisans in the same way, it did not. The assumption is simply untrue. Increasing information ratings affect liberals and conservatives (and Republicans and Democrats) very differently. While seemingly a technical point, this difference in assumptions is sufficient to make Bartels' conclusion erroneous.
In this comment we present our initial result which we follow by comparing our analysis to that of “Homer Gets a Tax Cut,” showing how the restrictive information assumption affects the results. We also assess Bartels' claim that people who reported not having thought about economic inequality in a particular way “lacked a moral basis” for their opinions. We then propose an approach that takes public policy questions seriously while respecting ideological differences in opinion.
A New Finding
It is important to understand why voters support and oppose the policies as they do and to discover what role information plays in their assessments. However, conclusions about how information affects opinion can be difficult to draw because we usually only observe citizens with the information they have. Nature seldom provides us with direct observations of how people might think or act if given more (or less) information. So to claim that ignorance is the cause of a particular outcome, one must be able to compare an observed behavior to what someone would have done if they had more information. While there are many ways to do this, we will keep things simple by following Bartels' procedures and assumptions—with one notable exception.
Bartels shows that among respondents who offered an opinion on the tax cut, an overwhelming majority supported it. He argues that much of this support was due to ignorance. His claim is based on a relationship he observes between support for the tax cut and the NES information rating taken at the end of the interview. The survey interviewer is asked to rate the interviewer's political information. They can choose one of five responses: very high, fairly high, average, fairly low, and very low.8
Since so few respondents were below average, we merged them into the “average” category in figures 1 and 2. In the pre-election interview, fewer than 9 percent of the respondents were rated below average (either “fairly low” or “very low). Just over 10 percent of respondents earned this rating in the post-election interview. Liberals and conservatives were not significantly different in this respect with conservatives receiving slightly higher ratings on average.
Zaller 1986 offers an explanation and defense of these interviewer ratings. His arguments are part of a continuing argument about the validity of survey-based political information measures (see, e.g., Althaus 2003, Delli Carpini and Keeter 1993, 1996; Iyengar 1986, 1990; Lupia 2005, Mondak 1999, and Mondak and Davis 2001). We use these measures to show that a simple variation in Bartels' analysis yields a very different substantive conclusion. To make the examination efficiently, we use Bartels' analysis as a foundation and develop our own result with a small number of changes from his analysis. Except where noted, therefore, our choice of variables and codings are identical to those of Bartels.
Support for the tax cut by NES post-election information rating
Differences in support for the tax cut by NES post-election information rating
Figure 1 shows this correspondence for the sample as a whole. It shows that as people reach higher information levels, they are less likely to support the tax cut. However, the effect of information is quite small for the sample considered as a whole, which cuts against the claim that “better-informed respondents were much more likely to express negative views about the 2001 tax cut.”10
Bartels 2005, 24.
Figure 2, while depicting a more substantial relationship between information and opinion, exposes an important qualification to Bartels' claim. The bars on the right-hand side of figure 2 depict this relationship for respondents who label themselves “conservative” when asked about their ideology (top) or “Republican” when asked about their partisanship (bottom). We display both groups to demonstrate that our result survives using either categorization and because Bartels alternates between these categories in “Homer.” For this large group of citizens, the relationship between information rating and tax cut opinion is just the opposite of what Bartels describes. As members of these groups achieve higher information scores, their support for the tax cut increases. Also noteworthy are the extremely high levels at which all members of these groups support the tax cut. For conservatives, 82 percent of respondents whose information rating was “average” or below supported the tax cut. This compares to 88 percent of those who rated “very high.” For Republicans, the corresponding statistics are 89 percent for respondents rated “average” or below and 96 percent for those rated “very high.” Among these groups, there was a clear consensus in favor of the tax cut—particularly for those who were coded as having the most information.
The left side of figure 2 depicts the relationship for self-identified liberals and Democrats.11
Of the 122 liberals placed into figure 2's average or below category, the tax cut was supported by 80 percent of the 5 rated “very low,” by 64 percent of the 14 rated “fairly low,” and by 48 percent of the 103 rated “average.” Of the 211 conservatives placed into figure 2's average or below category, the tax cut was supported by 40 percent of the 5 rated “very low,” by 81 percent of the 31 rated “fairly low,” and by 83 percent of the 175 rated “average.” Of the 168 Democrats placed into figure 2's average or below category, the tax cut was supported by 57 percent of the 7 rated “very low,” by 61 percent of the 23 rated “fairly low,” and by 53 percent of the 138 rated “average.” Of the 150 Republicans placed into figure 2's average or below category, the tax cut was supported by 50 percent of the 2 rated “very low,” by 94 precent of the 18 rated “fairly low,” and by 89 percent of the 130 rated “average.”
Twenty-six respondents identified themselves as neither liberal nor conservative and answered the tax cut question. Of these respondents, twelve were rated “very low” to “average” and seven (58 percent) of them supported the cut; nine were rated “fairly high” and five of them (55 percent) supported the cut; and five were rated very high with three (60 percent) supporting the cut. Fifty respondents identified themselves as neither Democrat nor Republican and answered the tax cut question. Of these respondents, twenty-seven were rated “very low” to “average” and 63 percent of them supported the cut; eleven were rated “fairly high” and 73 percent of them supported the cut; and twelve were rated very high with (50 percent) supporting the cut. In neither case is the relationship between information and tax cut support equivalent to the one described by Bartels.
Bartels claims that ignorance explains the overwhelming support for the tax cuts. Figure 2 suggests a different story. For liberals, higher information ratings correspond to more tax cut opposition. For conservatives, the same is not true.13
NES interviewers rated almost all respondents as having an average or above average level of political information. This fact presents an additional problem for Bartels' thesis—for if we treat the interviewer assessments as a valid measure of a respondents' political information, then we must conclude that almost all of the support for the tax cut was offered by respondents rated average or above average in their political information. The “simple ignorance” thesis is not consistent with this fact.
Can Reasonable People Disagree?
In this section, we replicate the core finding in the statistical analysis of “Homer Gets a Tax Cut” and show that its ignorance claim is due to an assumption about information that is questionable as a general matter and, in this case, is falsified by the data. While “Homer” includes a range of statistical analyses, its main conclusion derives from the final regression equation included in its table 4 and reprinted without changes here as the first column of table 1.
A replication and expansion of Bartels' information–Opinion analysis
The purpose of Bartels' analysis is to estimate how different information levels affect a respondent's likelihood of supporting the tax cut. Its dependent variable comes from the question,
“As you may recall, Congress passed [President Bush signed] a big tax cut last year. Did you favor or oppose the tax cut, or is this something you haven't thought about? Do you favor [oppose] the tax cut strongly or not strongly?
The main explanatory variable is the interviewer rating of the respondent's political information described above. Control variables in the analysis account for party identification, family income, and a question-wording experiment in which half of the respondents were told that “President Bush signed” the tax cut while the other half was told that “Congress passed” it. With the exception of party identification, none of the control variables produce large or statistically significant effects.
In the second column of table 1 we report our replication of Bartels' analysis.14
Like Bartels, we coded responses as 1 for those who strongly support the tax cut, .5 for those who support it, −.5 for those who opposed it, −1 for those who strongly opposed it, and as 0 for all others. Political information equals 0 for respondents who receive the lowest information rating from NES interviewers, 1 for those who receive the highest rating, and interim values of (.75, .5, .25) for those whose ratings are in the middle of the interviewers' 5-point scale. With the exception of running separate regressions for different partisans or ideologues, we use the same instrumental variables equation as Bartels. This includes using the post-election interviewer rating as an instrument for the pre-election rating (the two are correlated at .35). We also use the same NES sample weights in our regressions. For readers who are interested in our variable codings and regression equations, we have available a complete replication file.
As previewed in the introduction, however, Bartels' regression—both its original version and its replicate—restricts all respondent opinions about the tax cut to react to changing information ratings in an identical manner. Applying this same logic more generally, we would have to assume that Bush supporters and Kerry supporters would react to any new information carried in 2004's “Swift Boat” commercials in identical ways or that every American would respond identically to questions about the suitability of a nominee for the Supreme Court after learning that she or he was pro-choice. In many political settings, such an assumption makes no sense as politics often involve trade-offs between outcomes valued by one group and outcomes consistent with other values. In this case, we will show that Bartels' assumption is falsified by the data.
We present our analyses in the third through sixth columns of table 1. In contrast to Bartels' regression, we do not make any a priori assumption about how different partisan and ideological groups should react to different amounts of information. By running separate regressions for liberals and conservatives, and separate regressions for Democrats and Republicans, we let the data tell us whether more information affects members of the group in the same or different ways.15
While only Republicans and Democrats are explicitly referred to in the regression in question, elsewhere in “Homer,” results are characterized for liberals and conservatives. Since the two means of classifying respondents are not identical, we report results on both groupings in table 1 to demonstrate the robustness of our claim. We also ran regressions for respondents who categorized themselves as Bush supporters and opponents (the variable the Bartels uses as an instrument for Republican Party identification in the replicated regression). In neither of these cases did the political information coefficient achieve statistical significance.
Our regressions show differences in how information ratings relate to tax cut support. Were Bartels' assumption about information and public opinion correct, we would see very similar coefficients on the political information variable (i.e., his assumption would receive greatest support if all of the numbers in bold font were close to −.907 or −.721). Table 1 shows that this does not happen.
In columns 3 through 6, increasing information ratings makes liberals and Democrats less likely to support the tax cut. However, higher information ratings have no significant effect on conservative or Republican support for the tax cut. Put another way, reasonable people (e.g., highly informed liberals and highly informed conservatives) can disagree—and in this case they did. Only liberals react as Bartels predicts. Other respondents react quite differently.
Fueling Bartels' result was what some scholars call “within group variance” and what others might call “ceiling effects.” Put simply, in the present context, liberals vary in ways that other respondents do not. Conservatives or Republicans—it doesn't matter which categorization is used—supported the tax cut at extraordinarily high levels, regardless of their information ratings. Referring back to figure 2 reveals not only the conservative consensus, but also considerable differences among liberals. Bartels' analysis merges these effects into a single statistic. Our analysis reveals that the size and statistical significance of his “−.907” coefficient is driven largely by the large variance in the relationship between information and opinion among Democrats/liberals. Merging the effects masks the conservative consensus. Had conservatives/Republicans not been in such high agreement about the tax cuts, the “ignorance” conclusion would have been harder to refute. But “the right” was in agreement on this issue.
If we take our result as indicative of the effect of information on political preferences, we can easily reject the hypothesis “the strong plurality support for Bush's tax cut … is entirely attributable to simple ignorance.” For this conclusion to be true, the effect of political information on tax cut support would have to be large, negative, and statistically significant for all respondents. Table 1 shows that it is not. Instead, support for the tax cuts comes from a conservative consensus (which varies little with information rating or income) and quite a few liberals, including some who earned high information ratings.16
Moreover, the 2002 NES later asked respondents “How important is this issue to you personally—very important, somewhat, or not important at all? Almost 30 percent of respondents replied “very important” and another 57 percent replied “somewhat important.” Differences between Republicans and Democrats were neither large nor significant, with Republicans only slightly more likely to say that the issue was somewhat or very important. With respect to the question of whether Republican support for the tax cut was due to ignorance or something else, two statistics are worth noting. First, the more important Republicans rated the tax cut issue, the more likely they were to support it. Second, sixty-five respondents identified themselves as Republican, responded that the tax cut issue was either “very important” or “somewhat important,” and received the highest possible information rating. Sixty-two of them (over 95 percent) supported the tax cut.
How Other People Should Think
Elsewhere in “Homer,” citizens are described as “simple-minded,” “unenlightened,” “superficial,” and “lacking a factual or moral basis” because of the opinions they held. A point of emphasis for Bartels in drawing such conclusions is the fact that many NES respondents, when interviewed late in 2002, reported not having thought about a tax cut that was passed early in 2001. What conclusions can we draw about such citizens?
Since several events of great public relevance have occurred since the passage of the 2001 tax cut, it may be difficult to remember that it was supported not just by citizens, but also by many political elites. Among economists, the set of endorsements ranged from Federal Reserve Chair Alan Greenspan and William Niskanen to Princeton economist Alan Blinder, who served under President Clinton as the Vice Chairman of the Federal Reserve's Board of Governors and as a member of the President's Council of Economic Advisors. Blinder, in fact, argued that the tax cut in the first year was too small.17
http://marketplace.publicradio.org/shows/2001/06/07_mpp.html. Indeed, looking back at the web pages posted on economics and the tax cut reveals many strong claims about what economics had to say about the likely consequences of the tax cut. While many of these pages viewed in isolation suggested a consensus among economists, reading across pages revealed deep and important disagreements. Among the topics about which leading economists disagreed were about the extent to which the tax cuts were of sufficient size to generate a short-term economic stimulus and whether the nature of their targeting made such a stimulus more or less likely. Others argued that the tax cut's relatively large benefits to the rich would be more effective in boosting longer term growth by increasing the demand for investment rather than consumption, though these arguments were often countered by fears of increasing the deficit. More noteworthy is that advocates on both sides of the tax cut issue could, in the early months of 2001, find leading economists that would support their point of view. While we suspect that this information will tempt some to conclude that the economists on the “other side” of the tax cut issue were just ignorant, it is important to remember that economics is a complicated field of study. There is no universally recognized axiom sufficient to render the Bush tax cut right or wrong. Therefore, notions such as “simple ignorance” are not an accurate or constructive way to characterize the views of those economists with whom one disagrees.
It is also important to remember that expert and public discussions of the tax policy focused on more than economic inequality. The cuts were proposed as the economy continued to slow, as the collapse of the tech stock bubble and the dot-com era's “irrational exuberance” were being fully realized, and soon after the federal surplus came in far larger than most analysts expected. In 2001, the tax cut debate was about much more than inequality.
Of course, politics is filled with people who make claims about what others should know. They will assert that certain pieces of information should take precedence in decision-making. For example, pro-life advocates will encourage people to focus on the plight of the unborn child, while pro-choice advocates will encourage audiences to focus on the plight of pregnant women. In other cases, advocates claim that certain issues should be raised above all others in the public mind—claims about which matters others should weigh most heavily when contemplating social issues.
Bartels' descriptions of citizens as “simple-minded,” “unenlightened,” “superficial,” and “lacking a factual or moral basis” for not thinking about a particular policy (the 2001 tax cut) in a particular way are based on strong assumptions about the kinds of tradeoffs that people with different values should make. The “superficial” description, for example, is in reference to individuals who reported not having thought about whether the 2001 tax cut was good or bad. But what aspects of politics should citizens think about? The answer is less clear cut than “Homer” suggests.
While inequality is an important aspect of social life, it is not the only one. To make this point concretely, we went back to the 2002 NES pre-election study from which the tax cut question was drawn. We found three other questions that had a format similar to that question—similar in that the question allowed respondents to offer a response about a particular political issue or allowed them to say that they hadn't thought about it (or were not at all interested). The questions meeting this criterion covered a possible war with Iraq, corporate scandals (such as Enron), and interest in the 2002 campaign.18
The questions were as follows. Iraq: As you may know, President Bush and his top advisers are discussing the possibility of taking military action against Iraq to remove Saddam Hussein from power. Do you FAVOR or OPPOSE military action against Iraq—or is this something you haven't thought about? Corporate scandal: Over the last year, investigations into a number of large corporations such as Enron have found top executives to have exaggerated profits through shady accounting procedures. The executives received huge bonuses, but their companies went bankrupt and workers lost their jobs and retirement savings. How much attention would you say you've paid to those stories—QUITE A LOT, SOME, JUST A LITTLE, or NONE AT ALL? Political campaigns: Some people don't pay much attention to political campaigns. How about you? Would you say that you have been VERY MUCH INTERESTED, SOMEWHAT INTERESTED or NOT MUCH INTERESTED in the political campaigns so far this year? For this question, we coded people as not thinking about the issue if they said they were “not much interested.”
What 2002 pre-election NES respondents haven't thought about
Only 3 out of every 200 respondents report not thinking about any of the four issues. In other words, using Bartels' measure of whether or not citizens thought about the tax cut issue, over 98 percent reported having thought about one or more of the four issues. This means that almost all of the NES respondents who were characterized as simple minded for not thinking about the tax cut reported thinking about other political issues. Clearly, respondents varied in the political topics that interested them. Given that the 2002 NES was conducted so soon after the 9/11 terrorist attacks, six months before the widely anticipated war with Iraq, and over a year after the 2001 tax cuts had been passed, table 2's distribution of attention should not be surprising.
Though these four questions cover only a small fraction of the issues that make people interested in politics, they are sufficient to reinforce the long-standing finding that citizens vary in the political topics that engage them.19
The fact that many do not pay attention to an issue such as tax cuts may say little about their political competence more generally. Who are we to say that someone who thinks about matters of life and death such as war or terrorism, or someone who thinks more about unemployment or education or crime or abortion than they do about economic inequality or the tax cut, are “simple minded?”A similar argument holds for citizens who care about inequality but had not thought about the specific tax proposals described on the 2002 NES. The fate of the poor depends on many factors, some of which are a direct function of federal tax policy and others that are not. With this fact in mind, it is important to realize that most citizens lack the opportunity to act on individual policy proposals. The connection Bartels seeks regarding the 2001 tax cut is not one on which most citizens could act efficaciously.
Citizens do not get to vote from among a long list of possible tax policies. At the federal level, citizens' choices are very limited: they can choose a president, a House member, and a Senator who, to varying degrees, can affect tax cut debates. Given Bartels' evidence, it is impossible to reject the hypothesis that many people eschew investing in information upon which they can never act (i.e., federal tax policy) in favor of information on which they can. The same people who paid little attention to Bush's tax policy may have concluded that other issues were more important or that information about state, local, or neighborhood level civic activities (including, perhaps, local activities that help the poor) were a better investment of their time. That they respond to a survey question about the tax cut by saying that they haven't thought about it may just tell us that they have made a different set of trade-offs; that other social issues or venues are more relevant to them.20
Bartels notes that in his analysis of a related issue—the repeal of the inheritance tax—even highly informed people were split on the issue (Krupnikov et al. (2006) reveal that 88 percent of highly informed Republicans and nearly 60 percent of highly informed Democrats supported estate tax repeal). He argues (2006:25) that many respondents lacked a moral basis “for thinking of growing economic inequality as a problem that might be exacerbated by repealing the inheritance tax.” However, it is worth noting that like the tax cut, the inheritance tax had broad support among political elites—a fact that parallels Bartels' own findings about the opinions of the most highly informed respondents in his study. He concludes that “these results … highlight real and profound limits of political information as a transforming force when it comes to public opinion about complex policy issues.” If we read this claim correctly, it implies that even when people are well informed they still do not meet his preferred standard of decision making.
Indeed, at the time of our analysis, available surveys suggest that many people who voted for George W. Bush continued to support his economic agenda. This group includes many very intelligent people who are very knowledgeable about the actions of government, people who knowingly did not benefit from the tax cut, and people who have found his choices to be consistent with their notions of the public good. Of course, some in this group will base their support for Bush on very diffuse stereotypes of the consequences, but this fact will not distinguish them from many people who support Democratic candidates. Voters choose from the candidates they are offered with varying levels of awareness about the impact of a candidate's policy agendas. Some do so because, given how they weigh the large number of issues upon which a presidential vote can be cast, Bush is a far better fit than any Democrat. That Democrats reach a different conclusion is not sufficient evidence to render either group wrong. Difference need not imply ignorance.21
While readers are likely accustomed to arguments of the form that poor or middle class people who support Republicans are necessarily voting contrary to their true self-interest, they may be less accustomed to seeing a parallel argument applied to traditional Democratic voters. We offer an example of such an argument, itself presented as irony, to reinforce the idea that citizens can reach political conclusions for reasons that are not strictly economic.
The thirty-seven blocks of residential towers that line the western edge of Central Park, from its lower end at Columbus Circle to the age-old social barrier of Ninety-sixth Street, make up a self-contained world whose sprawling apartments, with their high-ceilinged living rooms, formal dining rooms, and unobtrusive maids' quarters, are home to investment bankers, corporate lawyers, and media executives. And yet in a baffling testament to the failure of Americans to grasp their economic self-interest, the residents of CPW (as locals colloquially call their street) overwhelmingly voted for John Kerry and the Democrats. This shouldn't be! … Why, then, does Central Park West cling so stubbornly to irrational Democratic Party loyalties? The most plausible explanation is that the prickly voters of CPW feel that their traditional moral values (getting into Yale on merit, reading books other than the Bible, cherishing things from France) are not fully embraced by President Bush … CPW is an insular and hidebound neighborhood, brimming with cultural resentments unfathomable to outlanders (Shapiro 2005).
Going beyond the issue of partisan differences, it is true that Bartels identifies a group of respondents who supported the tax cut at the same time that they said that economic inequality is a “bad thing.” Were they acting ignorantly? While the two opinions appear internally inconsistent, they are far from providing concrete evidence of ignorance. Indeed, we contend that the 2002 NES data is not well equipped to support the “simple ignorance” conclusion because it contains very little information about what people were thinking when they answered the questions at issue. While the study is a tremendous resource for demonstrating the strong patterns of public opinion that we offer above, it provides precious little data about the mental states of the people offering those opinions. To see the difference, suppose that you answer two questions. When asked if you would like to earn more, you say “yes.” When asked if you would like to see economic inequality reduced, you also say “yes.” Then, you are offered an opportunity to earn more. You take it. Your decision can increase economic inequality. Suppose that it does. In this situation, there are a range of conclusions we might draw about the relationship between your mental state and your actions. First, we might conclude that you are ignorant—that you do not understand that you just chose to earn more and, hence, increase inequality even though you said that you preferred less inequality. Second, we might conclude that you are not willing to sacrifice any pay increase, no matter how immediate or large, in exchange for a decrease in inequality, no matter how remote, uncertain, or small. Given our observation, however, we cannot reject the hypothesis that there are other circumstances in which you would sacrifice higher pay for less inequality. A third possibility is that you were not as committed to the principle as you said you were (because, perhaps, saying to someone else that economic inequality is “bad” is more socially acceptable than saying the opposite). These three cases describe three different descriptions of your mental state. Without more data, we cannot tell which description of you is correct.
So considering what economists and other experts have said about the 2001 tax cuts, perhaps we can agree to characterize it as analogous to the package “tax rebate now plus greater short term economic inequality plus debatable longer term economic effects.” That citizens opted for this particular tax cut over the pre-2001 tax code may, as Bartels implies, tell us that they were too “unenlightened” to see any kind of connection between the policy and the principle or it may tell us that they would have preferred other, more equal, tax proposals to the final Bush plan (had such proposals been offered) or it may tell us that while they prefer less inequality, this desire is less important to them than supporting a policy that was likely to provide an immediate economic stimulus or it may tell us that the respondents weren't so committed to inequality in the first place. While Bartels concludes that “public support for President Bush's tax policies derived in considerable part from ‘unenlightened’ considerations of self-interest,” the fact of the matter is that the data do not allow us to distinguish the possibilities listed above from one another.22
Bartels 2005, 25. The “simple ignorance” conclusion on which we focus our reanalysis is the final step in a more extended statistical argument about why people supported the tax cut. Bartels first shows that many people respond to a survey question about growing economic inequality by saying that it is a bad thing. He then contrasts these responses with their support for the tax cut and seeks an explanation. In addition, he contends most people should have opposed the tax cut because it was against their “material self interest” while admitting that the basis of his calculation of such interest is “obviously debatable.” He devotes the final statistical analysis to questions of why so many people supported the tax cuts nevertheless. To this end, Bartels analyzed how respondent opinions of their relative tax burdens related to their feelings about the tax cut. He found that people who perceived their own tax burdens as being too high were particularly likely to support the tax cut:
Respondents' subjective perceptions of their own tax burdens had a consistent and very substantial effect on their views about the 2001 tax cuts … [T]his effect of “unenlightened self-interest” is extremely consequential for the aggregate distribution of policy preferences accounting for more than three-fourths of the substantial net support for the 2001 tax cut in the NES sample” (Bartels 2005, 23).
After reaching this conclusion, Bartels (2005:23) asks, “One is left to wonder how these people would resolve the contradictions implied by their simultaneous antipathies toward inequality and taxation—if they recognized those contradictions.” This question leads to his final statistical analysis which yields the “simple ignorance” conclusion.
Having focused on the validity of the “ignorance” claim in the text, we offer three comments about Bartels' prior conclusion that his findings demonstrate “unenlightened self interest.” Each of the comments parallels an argument that already appears in our text. First, the conclusion is premature. As we have argued above, people can support or oppose a tax policy for many reasons. The fact that middle and low income people voiced support for the policy does not, as Bartels suggests, make them inferior thinkers nor (given continuing debates amongst economists about the impact of changes to the tax code) does it necessarily make them worse off. Second, the conclusion is speculative. The 2002 NES did not ask respondents to give an opinion about how the Bush tax cuts would affect their relative tax burden (or their well being)—it asked about their perceived tax burden as a general matter. While we could speculate as to what such responses might be, the data provide no direct evidence of how people relate the Bush tax cut to their own perceived tax burden. Third, the statistical result is not the same for conservatives and Republicans as it is for liberals and Democrats. We come to this conclusion by rerunning the regression described in the long quote above. Our re-analysis follows the same rules as in table 1: replicate Bartels's procedures as closely as possible and then run different regressions by party or ideology. The regressions contain the same dependent variable as in table 1 and the same control variables accounting for party identification, family income, and the question wording experiment. Unlike table 1, political information is not included, but several other independent variables appear. Bartels adds a conservative ideology measure along with responses to three questions about tax burdens. These questions ask about the respondent's own tax burden, the tax burden of “the rich,” and the tax burden of “the poor” respectively. The lowest values (−1) of these variables indicate that the named entity pays “less than they should” while the highest values (+1) indicate that, in the respondent's opinion, the named entity pays more than it should. Roughly half of ANES respondents (including 41 percent of Democrats and 54 percent of Republicans) believed that they pay more than they should. Bartels' regression (column 4 of his table 3) again restricts all respondents to react to changing perceptions of their own tax burdens in an identical manner. The coefficient on “own tax burden” is .575 and statistically significant (s.e. = .112). In our analysis, the same coefficient for regressions on Democrats and liberals are .892 (s.e. = .205) and 1.128 (s.e. = .313). When liberals and Democrats come to feel that they pay too much, they are significantly less likely to support the tax cut. The effect is much smaller, and not significant, for conservatives and Republicans. For Republicans the coefficient is .479 (s.e. = .364) and for conservatives it is .168 (s.e. = .199). So even if we put our first two objections to the side, Bartels' claim about the relationship between unenlightened self interest, perceived tax burden, and support for the tax cut would apply primarily to liberals and Democrats to the extent that it applies at all. A full description of this work is available upon request.
To use a study such as “Homer” to reach the conclusion that voters are “unenlightened,” “simple-minded,” or “simply ignorant” requires a great deal of speculation—speculation for which direct evidence is scant. Instead of characterizing people as misguided, it may be more instructive to conduct scholarship that attempts to better fit our analyses into their rationales—including the likelihood that they approach political problems from varying ideological perspectives and with different values in mind.23
Finally, Bartels (2005, 23 emphasis added) concludes from his analysis of tax burdens and enlightened self-interest “that liberals and Democrats were much less susceptible than conservatives and Republicans were to the simple allure of lower taxes.” This conclusion that such people supported the tax cut because of the “simple allure” is speculative, at best. Because nearly all Republicans and conservatives supported the tax cut and the fact that the 2002 NES does not include any questions about why people felt as they did about the tax cut, there is not sufficient variance in the dependent variable nor sufficient information in the available independent variables for Bartels (or us) to identify a single cause for why Republicans and conservatives supported the 2001 tax cut.
We suspect that many opponents of the tax cut were pleased to hear that public support for a policy with which they disagreed was due to the public's ignorance. It is heartening to think that the rationale for an outcome we dislike is that the people with whom we disagree are just uninformed. In this case, like many parallel explanations of political outcomes offered on the right as well as the left, such stories may be comforting and even entertaining, but it is important to know when they are untrue.24
Krupnikov et al. (2006), employ our reasoning and other methods to demonstrate the limited power of parallel arguments about the role of public ignorance in explaining for repeal of the estate tax. In one analysis, the same ANES data referenced in this article yields the finding that higher information levels correspond to a significant increase in support for estate tax repeal amongst Republicans.
A Constructive Way Forward
“Homer Gets a Tax Cut: Inequality and Public Policy in the American Mind” offers a critique of the conclusions that many citizens reached about the 2001 tax cut. Our essay offers a critique of the conclusions in “Homer.” Critiques, when valid, can be informative. But a better understanding of the American mind and its implications for public policy requires more than critiques. We end by offering a constructive way forward.
This alternative entails rethinking the relationship between the people who study social phenomena and the people who are being studied. Consider, for example, Andreasen's explanation of how skewed interpretations of this relationship inhibit the field of social marketing:
Customers are the problem… Here, the customers [or in the case of “Homer,” citizens] are “seen as the source of the problem.” The customer is seen as deficient in one of two ways.
Ignorance. Because the social marketer knows what a good idea it is to practice safe sex or put campfires out carefully, he or she assumes that the reason other people don't do this is that they simply do not know how desirable the marketer's favorite behavior is. Customers who are not complying are just too ignorant of the virtues of the proposed action.
Lack of Motivation. Every once in a while, social marketers who are convinced that customer ignorance is the main source of their lack of success are confronted by research data showing that customers are not all as ignorant as the marketers thought. They then turn to their backup explanation: the real problem must be a character flaw.25
Andreasen 1995, 42–43.
In this view (44), “customers are seen as ignorant, weak individuals.” The type of assessment permeates “Homer Gets a Tax Cut.” While it contends that people disagreed with its point of view because they were simply ignorant, it is hard to reject the hypothesis that many of them were deemed ignorant because they simply disagreed.26
The act of judging “out group” members “ignorant” because they cannot answer questions that the evaluator has deemed essential is a common fallacy in scholarship on civic competence. The fallacy occurs when an evaluator mistakes his peer group's consensus on the importance of a particular set of claims for a set of conditions that are necessary for other people to accomplish their tasks. While the history of intelligence testing in the 19th and 20th century is replete with such errors, the fallacy is avoidable. Avoidance requires a more careful analysis of the relationship between particular pieces of information and a person's ability to accomplish a well-defined set of tasks. For more, see Lupia 2005, 2006.
We have a different approach. It is consistent with Andreasen's remedy for the problem described above: “the customer is seen as someone with unique perceptions, needs, and wants to which the marketer must adapt.”27
Andreasen 1995, 49.
In this case, it is not clear how well these rationales were understood. In “Homer,” there was no concrete demonstration that people supported the tax cuts because they misunderstood critical facts. Indeed, no questions in which citizens could give true or false answers to questions about any concrete facts were analyzed. Instead, what was demonstrated is that citizens' responses to several questions about their opinions did not fit together in the way that one possible theory of what-should-go-with-what would have suggested. “Homer” blames citizens for the difference. We looked elsewhere and found an alternate explanation.
Members of politically relevant groups often disagree. We can learn a great deal about modern politics by studying why they reach different conclusions.28
With such knowledge we can match critical social needs (such as those of the poor) with citizens' desires. In so doing, a stronger and more credible foundation for progress in meeting these needs will be forged.
