Herding Cats

You've all heard that trying to get liberals united behind a common goal is like herding cats, or that when liberals go into battle, they form a circular firing squad. A new study appears to support these clichés, and attributes the problem to a liberal personality trait, the need for uniqueness.

The false consensus effect is a cognitive error in which we believe that more people agree with our attitudes than is actually the case. For example, both those who favor and those who oppose the legalization of marijuana might think that theirs is the majority view. One explanation for false consensus is that we tend to associate with others who share our views and are therefore exposed to a biased sample of opinions. The opposite of false consensus is a false uniqueness effect in which people incorrectly believe that most others disagree with them. However, most studies of the general population find false consensus rather than false uniqueness.

In the study by Chadly Stern and others, 292 participants recruited over the internet were asked their opinions on 41 topics, 22 non-political and 19 political. They were asked to classify themselves as liberal, moderate or conservative. Then they were asked to estimate the percentage of people participating in the study “who share your political beliefs” who agreed with them on each item. A true measure of false consensus or uniqueness was obtained by comparing the participants' estimates to the actual beliefs of the study participants in their political in-group.

It was found that all three groups were inaccurate. Both moderates and conservatives showed the usual false consensus effect, but liberals showed false uniqueness. They underestimated the percentage of fellow liberals who agreed with them. In a second study, 287 participants were this time asked to estimate the percentage of people who shared their political beliefs in the American population who agreed with them on each item. The same results were obtained.

The authors proposed an explanation for the false uniqueness effect among liberals: That they are higher on a personality trait called need for uniqueness, defined as a strong dispositional desire to feel unique. The authors used 11 items from this 34-item Need for Uniqueness Scale, but I'm not sure which ones. Liberals are indeed higher than moderates and conservatives in need for uniqueness. An analysis showed that the relationship between political ideology and overestimation of consensus was mediated by need for uniqueness, meaning that the correlation between ideology and overestimation was significantly reduced when the effect of their common relationship to need for uniqueness was statistically removed.

I have some concerns about this study. For example, I'm not sure how participants interpreted the phrase “people who share your political beliefs.” It would have been better if the instructions had specifically referred to fellow liberals, moderates or conservatives. There are good reasons to expect liberals to underestimate the percentage of people in the general population who agree with them. As Justin Lewis has demonstrated, the corporate media tend to exclude progressive opinions from public debate, leading liberals to feel outside the political mainstream. For example, the media continually identify support for modest cuts in Social Security and Medicare as the centrist political position when their own polls show that the public is overwhelmingly opposed to such cuts. If the participants in the Stern study were confused about the group whose attitudes they were estimating, this could partially explain the results. Liberals have been marginalized by the corporate media.

I am also unimpressed with the attempt to explain liberals' behavior as resulting from a need for uniqueness. In my opinion, the words “need” and “uniqueness” are both overstatements. The items in the scale largely reflect the respondent's willingness to engage in nonconforming behavior. All the data in this study are correlational, not causal. The apparent "need" for uniqueness could be as much a result of the illusion of uniqueness as its cause. In any case, the relationship between ideology and overestimation of consensus appears to still be statistically significant after the effect of need for uniqueness is removed, so it's not a complete explanation for the phenomenon.

In the article, the authors contrast the Occupy Movement to the Tea Party. They note that the Tea Party became a potent political force, while they claim the Occupy Movement broke down due to internal bickering, presumably a result of their individual needs to feel unique. This ignores a far more important reason for the Tea Party's greater success. They received behind-the-scenes financial and logistical support from wealthy conservatives and were welcomed into the Republican Party. The Occupy Movement, on the other hand, received almost no outside support, and was shunned by the Democrats, who chose to side instead with their wealthy political donors, mostly members of the 1%.  To attribute the failure of the Occupy Movement to the personalities of its members is politically naive. One of my pet peeves is the tendency for psychologists—even social psychologists, who should know better—to attribute behavior primarily to personal motives and overlook situational forces that are much more likely explanations for the behavior.

Regardless of why liberals underestimate their political strength, it is important that they examine public opinion polls directly, rather than trusting the corporate media to interpret them. (See, for example, my recent post on the public's “surprising” views on climate change.)

You may also be interested in reading:

The Myth of the Conservative Voter

Public Getting Warmer

Who Cares What They Think

Public Getting Warmer

Rep. Henry Waxman (D-CA) and Sen. Sheldon Whitehouse (D-RI), co-chairs of the Bicameral Task Force on Climate Change, released some surprising new public opinion data earlier this week. The data were prepared by social psychologist Jon Krosnick of Stanford's Woods Institute for the Environment. Krosnick did a secondary analysis of 21 random-digit telephone surveys and internet surveys conducted, mostly by Stanford, between 2006 and 2013, with a total sample size of almost 20,000 respondents. Krosnick's mathematical model looked at changes over time, and the data he presented are his best estimate of public opinion today.

The results show that there is broad consensus among the American public that climate change is happening and that government should do something about it. Here are the percentages in each state who believe global warming is happening.

The overwhelming majority agree, not only in liberal states like New York (84%) and California (82%), but also in conservative states like Mississippi (82%) and Texas (84%). At least 75% of residents in every state surveyed believe in the existence of climate change. (There were a couple of states for which Krosnick lacked enough data to make a reliable estimate.) Sixty-five percent or more in every state believe that global warming is caused by humans, and 58% or more believe that it poses a serious problem for the United States. Krosnick believes that personal experience with hot weather is primarily responsible for these opinion changes. 

There was only slightly less agreement on whether the U. S. government should take action to address climate change. In only four sparsely populated Western states (ID, MT, NV, UT) did the majority not agree that the government should do more. Here are the percentages by state who believe government should limit greenhouse gas emissions from U. S. businesses.

Similar majorities favor limiting emissions of greenhouse gases from power plants, tax breaks to produce renewable energy, and a cap-and-trade system to limit greenhouse gas production. You can go to this site to check out the maps for all the questions that were reported. If you would like to take a closer look at Pennsylvania's results, they are here.

Where did the public draw the line? In only five states did bare majorities endorse tax breaks for nuclear power. And in no state did a near-majority endorse consumption taxes on gasoline or electricity, or government support for all-electric cars.

These data are not new, but they seem to be sharply at odds with previous statements about American public opinion on climate change made by politicians and the corporate media. The consistency across states is particularly surprising. Here is Krosnick's take on how American public opinion is represented politically.

I have often heard legislators in Washington express the belief that there is considerable variation in opinions about global warming across parts of the country, and that most of the people in their state or district are skeptical about global warming. When I ask about the polling they have done that led them to this belief, I have routinely been told that they had not done polling and, instead, base their impressions on phone calls, emails and conversations with and from constituents on the issue. Our findings suggest that the balance of those direct communications from constituents to elected representatives may have created a misimpression of the public's opinions on the issue.

My guess is that Krosnick is trying to be diplomatic. This is not the first time politicians have been show to misperceive voter attitudes. Not only do Congresspeople not hear from a random sample of constituents, the ones they do hear from are primarily large campaign contributors.

Now that these data are available to Congress, we will have yet another opportunity to observe whether the American political system is responsive to public opinion. Would you like to estimate the chances of these surveys having an impact?

You may also be interested in reading:

Reality Shows a Slight Gain

The Myth of the Conservative Voter

Whose Opinion Matters?

The Myth of the Conservative Voter

The obvious fact that majority public opinion is unable to prevail in Washington has set off a flurry of research on why the political system is so unresponsive. The latest contribution comes from David Broockman of the University of California, Berkeley, and Christopher Skovron of the University of Michigan. They sent surveys by mail and email to all of the 2012 candidates for state legislatures in the country. The response rate was 19.5% (1907 candidates), which is about double the typical response rate for mail and email surveys. The candidates were asked to estimate what percentage of their constituents would agree with the following statements:

1. Same sex couples should be allowed to marry.
2. Implement a universal healthcare program to guarantee coverage to all Americans, regardless of income.
3. Abolish all federal welfare programs.

They were also asked their own opinions on the first two issues. A new technique called multilevel regression and poststratification (MRP) was used to estimate the attitudes of the citizens in each legislative district. It would obviously be very costly to survey the entire voting population each of the country's 6500 state legislative districts. MRP uses national survey data to determine level of agreement with the three statements based on education, race, gender, social class, religiosity and having voted for Obama. The percentage of people in each of these demographic subgroups in each legislative district is determined from the 2010 census. This generates an accurate estimate of the opinions of the people in each district. Although indirect, this method has been validated against real survey data.

There were positive correlations (rs = .43 to .51) across districts between the candidates' estimates and actual opinions, but they are not very impressive, accounting for only about 20% of the variance. The interesting thing about the data is that the candidates' perceived their constituents to be about 10% more conservative on these issues than they actually were, a difference roughly equal to the average difference between voters in California and Alabama. Stated differently, 60% of citizens would have to favor a liberal policy in order for their legislators to perceive them as a majority.

The responses were analyzed separately for liberal and conservative candidates based on their own answers to political questions. These data are shown in the table above. The horizontal axis represents the actual opinions in each district, and the vertical axis is the politicians' estimates. The black line is perfect accuracy. Anthing below it means the politicians are overestimating their constituents' conservatism. The blue line is the estimates of the liberal politicians, and the red line, the estimates of the conservatives. Although both groups believed their constituents were more conservative than they actually were, the conservatives seemed to be living in another world. They were off by over 20%. Nearly half the conservative candidates believed their district was more conservative than the actual opinions of the most conservative district in the entire country.

These findings may help us understand differences in the personal style of politicians. The well-known tendency of liberal politicians not to stand up for their beliefs contrasts sharply with the overconfidence of conservatives. Both may be explained by their fundamental misperceptions of public opinion.

Candidates were surveyed again after the election, but there was no evidence that they had learned anything. They were no more accurate than before, and their accuracy was unrelated to any of the activities they reported during the campaign, i.e., time spent talking with voters.

The first question about these data is their generalizability. The study should be replicated using federal and local as well as state candidates, and with a wider variety of issues. However, the differences are quite large, so I assume these findings are robust. How can they be explained? Broockman and Skovron do not speculate other than to suggest that political elites are not motivated to learn the attitudes of the people they represent.

Before giving them a political spin, we must be sure these errors are unique to politicians. It's possible that a random sample of the population might make errors of a similar magnitude and direction. If so, the data represent cultural influences to which we are all exposed. For example, the corporate media present an almost unrelentingly conservative view of the American public opinion, insisting that this is a “center-right” nation and referring to right-wing politicians as “moderates.” These data may reflect back what the media inaccurately tell us. The liberal-conservative difference may mean that the media outlets preferred by conservatives are even less accurate than those preferred by liberals.

Suppose the general public does not make errors that are comparable to these candidates—that is, these biases are unique to politicians. If these politicians are themselves more conservative than the general public, there might be a false consensus effect. False consensus refers to the tendency to assume that other people's opinions are more similar to our own than they actually are, to project our own beliefs onto others. There is reason to believe that political candidates are more conservative than the people they seek to represent, since so much money is required to run for office—either their own money or funds obtained from wealthy donors.

This shades into a second possibility, which we might call the “loudest voices” hypothesis. When considering this hypothesis, remember that money talks too. In fact, the Supremes have ruled that money is speech.

Suppose there are two types of political issues, economic and social. Economic issues are those issues that have major economic consequences, even they are sometimes debated using non-economic arguments. Therefore, they include not just taxation and public spending, but issues like climate change, health care, and foreign policy, including debates over war and peace. In the present study, health care and welfare are economic issues.

When the issue is economic, the loudest voices politicians are exposed to are those of campaign donors, whose views are well to the right of the general public. In fact, politicians may spend much of their time inside a bubble filled with rich contributors and corporate lobbyists. If you spend most of your time talking to these people, you are not likely to arrive at an accurate perception of what the general public thinks.

Social issues are those that have only minor economic consequences, such as abortion, gay rights and gun control. Although they are sometimes called “sideshow issues,” they are as important to some people as economic issues. For example, if you are pregnant, whether you have access to a safe, legal abortion is a matter of life and death. Social issues are represented in this study by gay marriage.

When the issue is social, political pressure comes primarily from grassroots organizations. Sadly, the loudest voices the politicians hear almost invariably come from the right. Conservatives are better organized, perhaps because they hold their attitudes with greater strength. Of course, what we call “grassroots” is sometimes actually astroturf. These groups receive financial and logistical help from the wealthy, since rich people know that they benefit when conservative candidates are elected.

Therefore, the loudest voices hypothesis suggests that, regardless of whether the issue is economic or social, the candidates hear mainly conservative voices. This is especially true of candidates who are themselves to the right of center. You can get a pretty distorted view of public opinion that way. Ask Mitt Romney.

Whose Opinion Matters?

On Monday, the Senate voted 51-45 (nine votes short of the supermajority needed) to block debate on the Buffett Rule that would have required millionaires to pay an income tax rate of at least 30%. A CNN poll showed that 72% of Americans favored the rule. This is just one of many policies—from ending the war in Afghanistan to providing a public option for health insurance—that were favored by a majority of Americans, but not by corporations or wealthy Americans, that have been defeated. How responsive is our political system to public opinion? And to whose opinions does it respond?

These questions are difficult to answer. Prevailing research can only tell us how well legislation corresponds to public opinion, and whose opinions it matches. In other words, we can look at the correlations between the attitudes of the population—or population subgroups—and political decisions. But if a positive correlation is found, that doesn't necessarily mean that the people are influencing their legislators' votes. The politicians could have persuaded the public to accept their policies. Or both the public and their legislators could be reponding to a variety of third variables, such as their shared background, real world events, or media coverage of political issues. Therefore, instead of talking about the public's “influence” on politicians, I will refer to the “consistency” between their opinions.

Early studies of consistency used the box-score method, comparing the majority preference on a particular issue, as indicated by public opinon polls, to subsequent legislative decisions. Chance alone would ensure that legislators would agree with the majority half the time even if they completely disregarded public opinion. Political scientist Alan Monroe found that in over 500 Congressional decisions between 1980 and 1993, the outcomes corresponded with public opinion 55% of the time. This was better than chance agreement (50%), but not much better. If our legislators follow the wishes of the majority 55% of the time, does this mean that we have a responsive government? Monroe reports that between 1960 and 1979, agreement was 63%, so whatever it means, it's going down.

Larry Bartels studied differences in the consistency of U. S. Senators' votes with the attitudes of their constituents of different social classes. He measured the relationship between participants' self-ratings on a 7-point liberalism-conservatism scale and all the decisions made by their Senators between 1988 and 1992. The respondents were divided into approximately equal thirds by income. (For convenience, I'll call them the “rich,” the “middle class” and the “poor.”) Through a complex (but not controversial) mathematical procedure, Bartels estimated the “weight” that should be given to upper, middle and lower class opinion in order to produce the best match with the Senators' decisions.

The opinions of both the middle class and the rich corresponded with the choices of their Senators, but the weight associated with the opinions of the rich was about 50% greater than the weight of the middle class. The opinions of the poor, however, were given no weight at all. (Their opinions were actually slightly negatively related to their Senators' votes.) These class differences were not as great on social issues, such as abortion, as on economic issues, but in all categories, rich people's views were most consistent with the outcome, and the views of the poor were unrelated to the outcome. Of the specific bills Bartels studied, the greatest inequality occurred on a proposal to increase the minimum wage. Here, neither the wishes of the poor nor the middle class had any weight at all, and the decision corresponded exclusively to the opinions of the rich. (I don't need to tell you what the Senators decided.)

Bartels divided the Senators by party affiliation. He found that the Elephants' votes were about twice as consistent with the views of the rich as the Jackasses' votes. Both parties were equally responsive to middle class opinion and equally unresponsive to the poor.

Martin Gilens compared the results of surveys (conducted between 1981 and 2002) asking people whether they favored 1781 proposed policy changes to subsequent votes on those policies. This allowed him to locate the best-fitting curve relating public opinion to Congressional decisions. He found a strong status quo bias. That is, as the percentage of people favoring a policy change increased, its probability of being enacted also went up, but even when a policy had 90% support, its probability of passing was only 46%.

He then divided the country into ten deciles according to income, and compared the opinions of people who were richer than 90% of Americans to the opinions of eight other groups, all the way down to the 10^th percentile of the income distribution. He found that the votes were consistent with the preferences of all ten economic subgroups, but as wealth increased, the correpondence between their preference and the decision increased. However, this analysis may have overestimated the consistency of the attitudes of the poor with politicians' votes, since in most cases they agreed with the rich. Gilens then did an analysis of the 887 cases in which there was significant diagreement on the policy among people at different income levels. Here are two charts showing these results.

The top chart compares the consistency of legislators' votes with people at the 90^th and 10^th percentiles of the income distribution. The positive slope of the 90^th percentile line indicates that as more rich people favor the policy, its probability of passage goes up. On the other hand, the 10^th percentile line is flat, indicating that the opinions of the poor are unrelated to the outcome. The bottom chart compares the rich with the middle class (the 50^th percentile). Unlike the poor, the middle class appear to have some influence, but of course not as much as the rich.

Why do the views of the rich carry more weight? Maybe they're more likely to vote, more likely to contact their Senators, or more knowledgable about current events. The surveys Bartels used asked participants to state whether they had voted or contacted their Senators. It also included a measure of political knowledge. Since all three of these measures were positively related to income, they were all associated with greater consistency between the respondent's views and those of the Senators. However, statistically controlling for turnout, contact and political knowledge reduced the difference between the apparent influence of the rich and poor by only 24%.

Unfortunately, the survey did not ask participants whether they had made political contributions. Looking at other data on campaign contributions by income level collected at that time, Bartels estimated that if the Senators had based their decision only on contributions, they would have given six times (600%) more weight to the views of the rich than the middle class and almost no weight to the poor. The actual disparity in weight between the rich and the middle class (50%) was not as great as the differences in contributions would have predicted.

I'm bothered by the fact that in these studies the "rich" are defined as either the top third of the income distribution (Bartels) or the 90^th percentile (Gilens). These subgroups include some people who might be considered upper middle class. It would be interesting to know how consistent the opinions of the top 1% are with those of the politicians. Gilens data seem to imply that as wealth increases, agreement also increases. Is there any upper limit to this effect?

Of course, these data are mostly from the 1980s and 1990s.  Over time, both the absolute amounts and percentages of campaign contributions coming from the rich have increased. This is probably especially true since the 2010 Citizens United decision, which allows unlimited contributions by corporations and the rich. But since the decision also allows these contributions to be anonymous, their impact will be difficult to measure.

Those who believe the U. S. political system is a plutocracy rather than a democracy will find these data to be at least consistent with their theories, particularly on economic issues, although the middle class's opinions are given almost as much weight as those of the rich when the issue is of little economic importance. The poor, however, have been effectively disenfranchised when their opinions differ from those of the rich. Given the secrecy with which political decisions are made, it will be difficult to find conclusive proof of quid pro quo corruption.

Reality Shows a Slight Gain

It's the first day of Spring in Pittsburgh, and a high of 80 degrees is predicted. Our magnolia tree is blooming a month ahead of schedule. Temperatures have been well above normal all winter—11.6 degrees so far this month. I would never cite any of this as evidence for global warming. This is weather, not climate. Yet for most Americans, this may be the most convincing evidence.

Photo by tumbleweed@yahoo.com

The sixth and latest National Survey of Public Opinion on Climate Change was conducted in December, 2011 by Christopher Borick of Muhlenberg College and Barry Rabe of the University of Michigan. It was a telephone survey with 887 respondents and a margin of error of +/-3.5%.

The good news is that belief in global warming seems to be on the rebound. In Fall 2008, 72% of Americans believed “there is solid evidence that the average temperature on Earth has been getting warmer.” This number dropped to 52% by Spring 2010. Since then, it has gradually increased and now stands at 62%. 26% say “no,” and 12% are “not sure.”

The survey finds the usual effect of political partisanship. (See table below.) The other three demographics they looked at—gender, race and education—had no effect. Yes, that's right, teachers, education has no effect!

	Earth Getting Warmer	Earth Not Getting Warmer	Not Sure
Democrats	78%	15%	7%
Republicans	47%	42%	11%
Independents	55%	30%	15%

Respondents who thought the Earth is getting warmer were asked the primary reason for their belief (out of nine possibilities that were suggested to them). It was a tie. 24% cited their personal experience of warmer temperatures, and another 24% said it was the presence of extreme or changing weather (floods, droughts, tornadoes, etc.) that convinced them. Therefore, almost half the respondents who believe climate change is real base their belief on the weather. Only 8% chose scientific research as their main reason for thinking the Earth is getting warmer.

The problem, of course, is that people are not distinguishing between weather and climate. Weather is far more variable. What randomness gives, it can also take away. Next year, when temperatures in the Northeast regress toward the mean—as they inevitably will—residents may again reduce their belief in global warming. (The authors point out that more people claim warm weather as evidence of climate change when their surveys are conducted in the Fall than in the Spring.) If we have to wait until a high percentage of Americans are convinced by the weather that global warming is real, we may fail to take action in time to save ourselves.

Finally, respondents were asked whether they thought “the media is (sic) overstating the evidence about global warming” and whether “scientists are overstating evidence about global warming for their own interests.” (Italics mine.) Not surprisingly, people who did not believe in global warming thought they were being lied to. 90% thought the media were overstating the evidence and 81% thought the scientists were.

I'm curious about the thinking that went into the wording of these two questions. Why are scientists presumed to have “an interest” in the outcome of global warming while the media are not? I wonder how much people think climate scientists are paid, whether they think scientists' salaries are contingent on finding evidence of global warming, and, if so, who is paying them this extra money. How does this compare to the incentive the media have for understating the evidence—namely, the millions of dollars they make from fossil fuel advertising?

The article does not report whether respondents attribute climate change to human activity. However, the authors promise a follow-up article looking at policy implications of the survey, so we may learn the answer to that question later.

Revisiting a Different World

In an earlier post, I suggested that liberals and conservatives live in two different worlds, and that conservatives, more than liberals, are lacking in epistemic rationality. That is, conservatives have a greater number of beliefs that do not correspond with objective reality. Public Policy Polling, a very credible polling organization, today reported a survey of 600 likely Republican voters in Alabama and 656 in Mississippi. The margin of error on these polls is plus or minus 4%. The main purpose was to predict the outcomes of tomorrow's primaries. Both are said to be too close to call, since the differences among the candidates are smaller than the margin of error.

What I found more interesting were the results of a couple of their throwaway questions--questions that were no doubt included simply to satisfy the pollsters' curiosity. In Alabama, 26% of Republicans said they believe in the theory of evolution, 60% do not, and 14% are not sure. In Mississippi, 22% believe in evolution, 66% don't, and 12% are not sure.

Another question that I believe measures epistemic rationality dealt with President Obama's religion. Only 12% of Mississippi Republicans believe Obama is a Christian, while 52% believe he's a Muslim, and 36% are not sure. In Alabama, 14% believe he's a Christian, 45% say he's a Muslim, and 41% are not sure.

One time, I had a conversation with a conservative student in which we were unable to reach a consensus about reality. When we finally decided to call it quits, his parting words were, "My vote cancels out yours."