academia, liberal bias and statistics abuse

Liberal Bias: it’s all fun and games until you start abusing data and statistics. Then the gloves come off.

I have a great time making fun of the hoots and hollers on the conservative side about supposed “liberal bias” in the media and other places. I even have started a little satire website at to make fun of it (if you haven’t yet, go check it out). It amazes me that people can claim that it’s “liberal bias” if Google happens to return derogatory news articles about conservative politicians (as if Google never returns derogatory articles on liberal politicians), or when an article in “Scientific American” doesn’t spend an equal amount of time talking about creationism as it spends on evolution when reporting on fossil discoveries.

But for the most part, that’s just normal workaday crazy talk. I just laugh it off and let it go.

It bothers me, however, when people start claiming that they have data that supports the claim of liberal bias. Especially when they are, quite simply, wrong. As a mathematician and a scientist, abuse of data is something that really bothers me.

I recently read an article in Psychology Today, Is there a liberal bias among American professors?, that presented some data and suggested that the data was scientific evidence of liberal bias in academia. It is wrong, and the reasons why it is wrong need to be discussed.

Before I go on, let me make one thing clear: It’s entirely possible that there is liberal bias in academia. Maybe there is, maybe there isn’t. That isn’t what I’m talking about when I say that this article is “wrong.”

My criticism is specifically about the supposed evidence for liberal bias put forth. For the rest of this article, I’m going to spend a great deal of time talking about the specific measurement the article used and the specific way that those measurements can (and cannot) be interpreted.

In the end, I don’t think that there is scientific evidence that there is liberal bias in academia: or at least, there is none that I have seen. But I don’t deny that it’s possible. I’m not saying that this article is definitively wrong in its conclusion about bias.  I am only saying that this article is wrong when it purports that the statistics it put forth are “evidence” of bias.

Are we clear on that?  OK, here we go.

“Is there a liberal bias among American professors?” is a short article that relies completely on one single measurement as a way of answering its titular question. The measure that it uses is the Democrat to Republican ratio (abbreviated D:R) in different groups.

One table then shows that at a large set of universities, the faculty D:R is higher than 1 (meaning that there are more registered Democrats than registered Republicans) and in many cases it is much higher: for example, at Berkley it is 8.7, meaning for each registered Republican on the faculty there are over 8 registered Democrats.

Another table shows a comparison between different disciplines, with unsurprising results like physics has a higher D:R than Finance and Psychology has a higher D:R than Management.

The conclusion that the article draws is that these numbers are evidence for the hypothesis that there is liberal bias in academia.

Is that a valid conclusion?

Remember, what I am asking is not “is there liberal bias in academia?” but something more specific: can we conclude that these statistics about D:R ratio constitute evidence for liberal bias in academia.

People affiliate with a particular party for a number of different reasons. Not all Democrats agree on all issues, just as not all Republicans agree on all issues. Let’s assume, just for the sake of argument, that people will align with whatever party they feel the most connection to, or that they feel represents them the best on the issues that they consider most important.

Now let’s imagine that you are a scientist, and you care about the importance of being honest and sincere about scientific research. No matter what your opinions are on taxes, regulation, foreign affairs, or religion, there is a good chance that you will look at a party that embraces creationism, that regularly derides and dismisses evolution and global warming research, and you will feel uncomfortable with that party. There is a good chance that you will look at GOP primary debates in which not a single candidate on the stage is willing to raise his hand to say that he believes in evolution, and you will think, “How can I possibly be a scientist, and align myself with that party?”

Now, not every scientist will feel that way. For some people, it won’t be important. For example, they might say, “I don’t care what candidates think about evolution, I care about their fiscal policies, and on that I agree with Republicans.” That is a completely valid view, and I am sure there are those who hold it.

But not all. There will be some people for whom being a scientist feels incompatible with the core message of the Republican party. And that, by itself, will lead to there being more Democrats than Republicans in the sciences in academia.

Does it make sense to call that a “bias”?

First, let’s take a step back and look at it completely in the abstract: If many people in a particular group hold the same view, is that necessarily evidence of bias?

No, not at all. You would never say that people in the United States have a bias toward believing that the earth is a sphere. Certainly, a majority of people in the United States believe it to be the truth, but we don’t call that a “bias”–that is not how we use the word.  Similarly, more of the people who use online dating sites are single than married. Is that a bias? No. Certainly, there is a reason for the connection: dating sites perform a function that is useful for single people, and not-so-much (at least, not in socially acceptable ways) for married people.  But that’s a functional relationship, not a “bias”.  To use the term “bias” in that context is to stretch it beyond its normal meaning.

So what would be an example of liberal bias in academia? One example might be if the percentage of people in academia who feel that the proper goal of government is to regulate business is higher than the percentage of people in the general population who feel that way. Why would this indicate liberal bias? Because this is a matter of opinion, about what the government should do, rather than a question of facts. It is something that is truly debatable, and that depends on your own vision of what society should be like. In this case, if faculty members in academia have a materially different view than the members of people in society at large, then that’s a bias.

What would NOT   be an example of liberal bias in academia? If the percentage of people in academia who believe in evolution is higher than the percentage of people in the general population who feel that way. That is not bias, because it is a scientific question. No matter what the religious right might want you to think, there is no debate among scientists, any more than there is debate over the idea that the earth is a sphere or that the sun is a big flaming ball of gas. It isn’t “bias” if there are fewer creationists who teach physics than there are creationists in the general population. It’s simply that educated people are more likely to understand and accept science than uneducated people.  I apologize if that seems harsh or mean, but it’s simply the way it is.  When conservatives suggest that maybe humans and dinosaurs lived at the same time, it isn’t “a debatable issue”; it’s just wrong.

So given that, what is the problem with D:R as a measure of bias?

The problem is that party affiliation represents a single category that is affected by an admixture of a number of factors.  If there are more Democrats than Republicans teaching biology, it could be because biology departments have a bias toward being in favor of taxing the rich, OR it could be because the Republican party has regularly and vocally allied itself with people who scorn and scoff at the theory of evolution. And when you only look at the party affiliation, by itself, you cannot tell the underlying reason for the skew in the measurement.

That is why this is a bad analysis, and an incorrect conclusion. Because D:R is quite simply a bad measure of “bias” in this context. The D:R measure is something that will be influenced both by debatable political topics AND the fact that the Republican party allies itself with groups that deny and denigrate the value of both science and education.  It’s not “bias” if people who teach science are spooked by a party that derides and denigrates science.

In all honesty, people who are interested in teaching science should be spooked by that.

liberal bias