Guest post by Luke Conway, Associate Professor of Psychology at the University of Montana.
The current consensus in psychology is that political conservatives are uniquely simple-minded. Indeed, even the famous critic of political bias and Heterodox contributor Jonathan Haidt (and colleagues) suggested that there is a “consistent difference between liberals and conservatives” on several measurements related to cognitive complexity (Joseph, Graham, & Haidt, 2009, p. 176).
When someone as open-minded as Jonathan Haidt implies that an array of evidence shows that conservatives are consistently more simplistic than liberals – even in an article largely devoted to exploring potential qualifiers to the effect – it suggests that the idea has achieved nearly axiomatic status in the minds of psychologists. And yet, in a recent Political Psychology paper, my colleagues and I showed that this “conservatives are simple” conclusion is, if not untrue, at least wildly premature.
You may reasonably wonder: If conservatives really aren’t simple-minded, how can so many smart people from all political viewpoints possibly be so fundamentally wrong about it? Let’s first talk about what we found, and then we’ll talk about how we ended up here.
What We Found
Our larger theoretical backstory was very simple and non-controversial: Complexity is domain-specific. You can be awfully complex in your views of steamed broccoli but still incredibly simple in your views of the Ku Klux Klan. This has implications for group differences. Because groups share attitudes and experiences that might lead them to different levels of complexity on different things, it is best to be cautious before drawing large-scale conclusions about any cultural group’s domain-general tendencies to be complex or simple – better to consider the relationship between culture and complexity as being moderated by topic domain. As a result, there is likely no such thing as a monolithic group of complex or simple thinkers.
Now this is a point we’ve made before in other contexts – a point so dull and obviously true that it hardly has seemed newsworthy to other researchers. But it turns out that its potential application to conservative simplicity is anything but non-controversial. (I’m pretty sure one of the ways you know that you’ve said something worth saying is when Steven Pinker tweets a link to your paper). Drawing from Peter Suedfeld’s and Phil Tetlock’s research and theory, what we argued in our paper is that it is possible, just possible, that prior work had not fully considered the topic domain when considering how simple conservatives might be. We set out to see what would happen when you offered a more comprehensive test of the conservative simplicity hypothesis with a large array of topic domains. The domains included:
- Climate change
- Death penalty
- Life without religion
- Bible truth
- Sex relations except in marriage are always wrong
- Drinking alcohol
- Smoking cigarettes
- Crossword puzzles
- Athletic activities
- Voluntary euthanasia
- Being the center of attention
- Separate roles for men and women
- Making racial discrimination illegal
- Getting along well with other people
- Playing organized sports
- Big parties
- Playing chess
- Public speaking
- Playing Bingo
- Wearing clothes that draw attention
- Easy access to birth control
- Organized religion
- Being the leader of groups
- Reading books
- Castration as a punishment for sex crimes
- Being assertive
- Roller coaster rides
- Loud music
And across four studies comprising over 2500 participants and covering over 40 topic domains, we found no evidence of the much-ballyhooed conservative simplicity effect. Rather, what we found instead was that for one set of topics, conservatives were less complex, and for another set of topics, liberals were less complex. And we barely came within a hint of anything like the now-famous effect suggesting that overall, conservatives are more simple than liberals. Comparing overall means that collapsed across multiple topics, the two groups of ideologues were consistently the same in their complexity levels.
In doing this, we examined two of the major arms of evidence often used to establish that conservatives are simple-minded: Self-reported dogmatism and open-ended scoring of participants’ integrative complexity. Integrative complexity – which is the most commonly used scientific method for scoring the linguistic complexity of written statements – involves assigning a score from 1 (very simple) to 7 (very complex) to source material based on the degree the author uses differentiated ideas and, at higher complexity levels, the degree that those ideas are integrated into a hierarchical structure. The studies on integrative complexity are especially noteworthy for at least two reasons: (1) In Jost et al’s (2003) famous meta-analysis that seemed to convince everyone (or at least the fourteen people who weren’t already convinced) that conservatives are unalterably more simple-minded than liberals, the total participant number for all integrative complexity studies combined was 307. The participant number in our three integrative complexity studies alone was over 2000. In other words, we had over six times the number of participants than all other work on the topic combined. (2) More importantly, the vast majority of that prior work had been in constrained elite contexts and had rarely made any effort to distinguish different topic domains. Thus, our work is the first systematic effort to incorporate a lot of different topics – and we basically found that liberals and conservatives are equally complex.
How Did We End Up Here?
What does this all mean? And I want to say up front that I don’t know for sure. I think that sometimes-awkward statement of doubt (“I don’t know”) is scientifically important. As we stated in the paper, I’m totally open to the possibility that conservatives are more simple-minded in a domain-general way, and I think there are reasons that it might be so, and there is certainly evidence (including from some of our own prior work on U.S. presidents and on the Obama/McCain election) suggesting that it’s true. The data from our most recent paper may turn out to be an anomalous and small counter-current that is overwhelmed by the larger sea of conservative simplicity. This is science; I’m open to that possibility.
But my point is that just as there are reasons why conservatives may be simple-minded, there are also reasons why they might not be – and we don’t totally know yet which is true. Some factors, such as those outlined in Jost et al (2003), may lead conservatives towards simplicity. But other factors – such as the fact that conservatives care about more value domains and thus may be more likely to have value conflict as discussed in Joseph et al. (2009) – might drive conservatives to have more complexity. I’m not pre-determining the conclusion – I just want to ask the question, honestly and openly.
And despite appearances, I’m not dismissing all available evidence in one fell swoop. I recognize that we still have only covered a small amount of the evidence for conservative simplicity in our paper. Yet we have another paper under review showing similar results for authoritarianism. After people have apparently been searching for Left-Wing Authoritarianism for 30 years – a search so futile that is has been compared to the search for the Loch Ness Monster – we re-wrote the RWA questionnaire in 10 minutes and found consistent evidence of LWA. This leads me to wonder if our field is really trying that hard to find left-wing authoritarianism? It took us ten minutes.
Nevertheless I’m not trying to claim that we are right and everyone else is wrong. I am arguing, however, that the larger case for conservative simplicity should be re-examined. I want our science to consider the legitimate possibility that the vast array of evidence suggesting that conservatives are simple-minded may turn out to actually not mean that conservatives are simple-minded. And if that’s true, it is worth asking: What the blazes does it mean?
I believe the research doesn’t tell us much about the participants’ ideology, but it does tell us a lot about the researchers’ ideology.
Scientists are human. We are not immune to the biases that affect the rest of the human race. When a dominant school of thought pervades an entire field, sometimes evidence might be skewed to fit that paradigm. The questions that might overturn that paradigm might not be asked, because few people think of them – and those that do think of them are not allowed a voice. And indeed, the history of psychological science is replete with biases against various groups in favor of the dominant paradigm, including biases against women (see Hyde, 2005), gays and lesbians (see, e.g., Cochran, Peavy, & Cauce, 2007), and ethnic minorities (see, e.g., Franklin, 2007).
I don’t definitively believe a parallel bias against conservatives has led us to the belief that conservatives are simple: But I’m open to the possibility, and I think you should be, too. As the authors on this blog (quite accurately) make clear, there is definitely a liberal bias in this field. And it is clear to me personally that liberals want to believe they are more complex than conservatives; it makes them/us feel good. And when we want something psychologically, we know that we often will twist psychological reality to get it.
Now, I’ve no direct evidence drawing a line from liberal bias to the research showing that conservatives are simple. But I have some small pieces of indirect evidence, and a large piece of overriding intuition.
First, we have some more recent work suggesting that people in general are motivated to believe that political candidates from their own political groups are more complex than they actually are in reality – and that this effect is especially (and quite possibly only) in evidence for liberals. In other words, liberals may be particularly prone to want to believe that the members of their group are more complex than they actually are. This doesn’t draw a line from liberal bias to present reality, but it does offer evidence that my intuition of the liberal motive in this instance has some reasonable empirical foundation.
Second, our lab has done a ton of integrative complexity research over the years with quite a number of coders. And we used that fact to collect data on some of our own past coders’ ideologies and compared that to the ideology of the people they were coding (Conway et al., in progress). The goals of the project were (a) to examine if our own coders were especially liberal as we suspected, and (b) to examine if there was any evidence of matched-ideology preferences, such that those who shared coders’ ideologies were scored as more complex.
As to the first question, the results were even more startling than we expected: All eight coders who responded (including myself) were more liberal than conservative. Indeed, not only was this sample of coders left-leaning, not a single coder was conservative by any measure, and comparing mean scores to a sample of students from the university at large suggests that our lab draws a more liberal set of students than an already left-leaning university population. Our lab is the liberal of the liberal, so to speak.
As to the second question, it is worth noting that there are about a hundred reasons why you would find no evidence of bias here: All coders were technically blind to the person or party of the paragraphs they scored, the lack of truly conservative coders constrained variability (our most “conservative” coders were moderates at best), among other reasons. And yet, in spite of those constraints, we nonetheless found evidence of ideological bias in coding. The most liberal coders scored paragraphs as more complex if they were scoring liberal politicians, as compared to more moderate coders scoring the same materials.
Before we jump too far over these data, I should note that we are talking about a small number of coders and a messy set of data. Further, the effect sizes were incredibly small – certainly not enough to account for past results showing that conservatives are simpler.
And yet they are illustrative of a larger point that applies across multiple domains, and ought to arouse suspicion about what the heck we are all doing. If coders who do not even definitively know the party of the person they are coding show subtle biases in favor of their own preferred candidates, how much more might the researchers – who are generally fully aware of the big picture – show such biases? And it is the researchers who ask all the questions, who interpret all the findings, who decide when and how much to present. And this is no small thing for a phenomenon – the production of linguistic complexity – that we know is highly context-dependent. Perhaps our largely-liberal field is simply not asking the right questions to find conservative complexity? I, for one, have yet to be convinced that we are trying hard enough.