Monday, August 8, 2016

New study finds that men are often their own favorite experts on any given subject - Pure, unadulterated tosh

Oh dear, what an execrable article based on execrable research. It would appear to be yet another example of the progressive clerisy telling themselves the tales they want to believe without having any rigor of thought to actually test their assumptions. Because the belief system of the progressive clerisy is at such variance with the general population, their circulation rates keep falling and there is little trust in them (60% distrust them according to Gallup).

From New study finds that men are often their own favorite experts on any given subject by Christopher Ingraham. The set-up is the implication that gender studies theorists are right, men are simply egotistical brutes, endorsed by Ingrams with his approving description of the research as "fascinating". It hardly needs a particularly skeptical mind to see that this is ideologically motivated research lacking basic research design features. You don't need an especially skeptical mind; just not a credulous one.

Here is the substance of the research as reported by Ingrams.
When an academic writes a research paper, it is common practice to give citations for various facts and assertions. It is not enough, for instance, to simply assert that "the global rise of the hyperdiverse ant genus Pheidole is an evolutionary epic with many subplots." You need to cite biologist Corrie S. Moreau's 2008 paper on "Unraveling the evolutionary history of the hyperdiverse ant genus Pheidole" to make that argument.

In academia, article citations like these are a marker of authority and influence: If your work gets cited by others hundreds of times, that's a good indicator that you're making a mark on your field. Universities often factor in citation counts when making decisions about hiring, tenure and pay.

As it turns out, academics have a handy tool at their disposal for juicing their citation counts: They cite themselves. There's nothing inherently shady about this practice. If you're an expert in a relatively obscure field like ant taxonomy, you're probably going to need to cite your previous work because few people people are doing similar work.

So Molly M. King and her colleagues at Stanford University, the University of Washington and New York University set out to find how often this so-called "self-citation" happens. They did so by examining a massive database of academic work: 1.5 million research papers in JSTOR, a digital library of academic books and papers published between 1779 and 2011.

What they found, first of all, is that self-citation represents a significant chunk of all academic citations. There were 8.2 million citations contained in the 1.5 million papers they studied. Nearly 775,000 of those citations, or about 10 percent of them, were of authors citing their own work.

[snip]

But more strikingly, King and her colleagues found a huge difference in self-citation patterns between men and women. "Over the years between 1779-2011, men cite their own papers 56% more than women do," they found. And in recent decades, men have stepped up their self-citation game relative to women: "In the last two decades of our data, men self-cite 70 percent more than women."
So if there is nothing wrong with self-citation, why is a gender disparity in self-citation an issue? According to Ingrams,
Regardless of the underlying mechanism, the self-citation disparity has a real-world impact on female academics' careers. Academics are more likely to cite papers that are already well-cited, so citing yourself means more citations from others. And more citations means better career-advancement opportunities.

This phenomenon probably contributes to women's continued under-representation on college faculties. Women have earned at least half of all bachelor's degrees in science and engineering fields since the late 1990s, according to the National Science Foundation, but as of 2013, they represent fewer than a quarter of university faculty members in those fields.
Tosh.

And as a miniscule aside, note the math. The size of the self-citation problem is 775,000/8,200,000, or 9.4512195%. Ingrams rounds this up to 10% when in fact it is either 9% or 9.5% depending on which decimal place you are rounding to. It seems like a pedantic point but there is value in precision and even more value in accuracy. Assigning a value of 10% seems like motivated reasoning on the part of the author, arbitrarily making the issue bigger than it is. It is a minor detail but perhaps indicative.

If you go to the original paper, you can see that the issue exercising the researchers is ideological. The opening summary sounds like a classical gender studies indictment of the ever present ethereal patriarchy, much discussed but never seen.

We know from the work of Claudia Goldin and others that the predominance of men in most fields of competitive endeavor is a function of parenthood and family structure and is not attributable to gender discrimination. Not just academia but law, accounting, politics, C-level business executives, best-selling authors, award winning authors, surgeons, etc. In most of these competitive fields (and others), at the top of the profession, women are about 15-30% of the members. That proportion is due primarily to personal decisions around marriage, children, and family structure to the extent that those decisions impinge on the investment of the voluminous and continuous hours of effort necessary to create high levels of expertise and accomplishment.

If more men are achieving tenured positions due to focused, and continuous (no interruptions) hours over long periods of time, it is likely that they are also creating deep knowledge in narrow fields. If you are the established expert in your narrow field, you are likely to cite your previous work more often simply because it is the only work available.

The researchers in this paper appear to not have controlled for the various important variables which are independent of gender but which are also likely to increase degrees of self-citation. Without even considering these variables, much less controlling for them, the research is without value. It cannot tell us anything other than the authors' own self-motivated speculation. Examples of uncontrolled variables that would likely affect the self-citation rate.
* Collaborative work - The researchers did put in place a mechanism for measuring self-citation when there are multiple authors. However, they did not address whether there are gender disparities in collaboration. I believe I have read something to the effect that female researchers tend to participate in co-authored papers at a greater rate than male researchers. If that were the case, then is there a difference in self-citation simply arising from the difference in sole versus multi-author papers? I don't know but without controlling for that possibility, the research is questionable. For example, let's say that women are much more likely to participate in collaborative research than men. Let's also speculate that in collaborative projects, there is greater social pressure to not play the prima donna and therefore not to self-cite, and that among the team members there is also a greater knowledge of the broader field and so other papers can be cited in place of one's own. In this plausible example, the tendency towards self-citation is not a function of gender but of preference for research form (sole versus collaborative).

* Productivity - If a researcher has produced a highly productive 100 papers, there is likely to be more self-citation than if they have produced a single paper. Indeed, there is no possibility of self-citation in the first paper. The more papers one has produced, the more likely it is that there is a reason to self-cite. If male tenured professors are more productive researchers than female researchers (for whatever reasons) then they will have a higher self-citation rate, not because they are male but because they are more productive.

* Years of continuous intense research - It is a simple fact of observed sociology that male professionals in any field are more likely to have a support spouse who handles all the familial and operational aspects of family life while female professionals are more likely to have a peer spouse (division of domestic labor). The consequence of these private family arrangements means that male researchers are more likely to be able to invest voluminous hours over long periods of time in a fashion that female researchers are not, particularly if the female researchers also have children. The implication is that women researchers are more likely to have periods of career discontinuity when they exit the workforce or are not as engaged for a period of time.

What this means, though, is that if the researcher (male or female) produces three papers in a year, there is a higher probability of citing one's own paper if you are working continuously than if you have exited temporarily. If you have been away and come back, you are more likely to find that someone else has written the paper with the necessary information to support your argument and you cite their paper. A male researcher producing research continuously over a ten year period will have a different self-citation rate than a female researcher who might have had two 12 month interruptions for children. Not because they are male but because they have worked continuously.

* Narrowness and specificity of field - The hard sciences (physics, mathematics, chemistry, engineering, etc.) tend to be cumulative in nature over time, building on earlier factual work. Hard sciences are likely to have a much higher self-citation rate because it builds on what came before. The narrower the field, the higher the rate of self-citation. The softer sciences (economics, sociology, psychology, etc.) tend to be more speculative and therefore less dependent on earlier factual work. If there is a strong gender split between the two types of sciences, as there is, then you would expect to see higher rates of self-citation because of field and not gender. Males tend to dominate in the hard sciences and economics whereas females disproportionately dominate sociology and psychology. Therefore, accepting the preceding observations, you would expect males to have the higher self-citation rate.

* Heterogeneity of researcher topics. If a researcher has produced a highly productive 100 papers across 100 independent topics, there is likely to be very little self-citation. If they have produced 100 papers in a single topic area where later papers are additive to earlier ones, there will be a lot of self-citation. If males are more field specific than females (for whatever reason) then, again, you would expect them to have a higher self-citation rate than females, not because they are male but because they are focused.
It appears that once again, ideology is driving blindness to real world causations. The researchers wanted to find a simplistic reason for why female researchers are less represented at the summit of their fields and they came up with the gem that women simply need to cite their own work more often. This type of approach is what gives sociology such a bad name. Simplistic, poorly designed and completely dependent on correlation without demonstrated causation.

It is also counter-productive. If any or all the above variables are operational, then getting women researchers to increase their self-citation rate will not increase their prevalence in their field. By focusing on the wrong issue, the problem goes unsolved or is made worse by application of inappropriate solutions to misunderstood problems. We need for ideology motivated researchers to quit muddying the cognitive waters with tosh of this sort and start demonstrating real rigor in their research.

No comments:

Post a Comment