Word Puzzle: Why is the Crossword Gender Gap Growing?

Posted: Aug. 21, 2014.

You probably don't want a front row seat to the raging debate that consumed the crossword constructors listserv a couple weeks ago. In the words of one contributor:

I joined this list yesterday, and now the thing I want more than anything in the world is to be off of it.

However, the filtered version of the debate is actually quite interesting, at least to an economist. And now that I've scared the noneconomists away, we can even speak freely in our secret code language of awesomeness! Just kidding, I will mostly confine myself to bare English.

The topic of the debate is the large and growing gender imbalance in published puzzles. Like many fields, crossword construction is (and always has been) pretty male-dominated. But why is it getting worse, unlike so many other historically male-dominated fields? Apparently the percentage of puzzles with female bylines has plummeted from 35% (in the old days) to 24% (ten years ago) to 16% (now). A puzzle about puzzles!

In this post, I will give some thoughts on a few proposed theories. At the end I'll offer a new theory which makes some peace with multiple perspectives and does a pretty good job of explaining the facts.

Apologies to anyone whose two-week-old arguments have been mutilated in my memory banks. I won't mention names, in part because the mailing list is not entirely "public" and in part because I do not trust myself to accurately represent any individual's views from memory.

Prelude

To begin, the gender gap is a complex social issue with many underlying causes. Whatever complex network of factors is causing it, there's clearly a large systemic component that spans many fields. Many contributors wrote in to elucidate some of these factors. And while it's easy to find compelling social factors that contribute to the disparity, it's much harder to make a compelling argument for why things are getting worse.

The Hobby Theory

As an example of one such factor, consider that crosswording is a hobby, and suppose there is disproportionate pressure on women to divert their free time towards family and household chores instead of hobbies. That's highly plausible, but itself does not explain why inequality is increasing. Nevertheless, I mention this theory as a representative source of gender imbalance that will be useful later.

To explain the gender imbalance trend from this theory alone, we'd have to cook up a special story about how free time is increasing faster for men vs women, or how as free time increases for both, women's "free" time is systematically diverted into non-hobby tasks. These assertions can easily be confirmed or falsified by looking at time use data, but even if true, I doubt this theory alone could drive the trend.

Of course, there are many other factors in play. At this juncture, it would not be unreasonable to take the view (as some commentators have) that anything could happen on net, and that we should not be surprised by the trend one way or another. But I think we can do better than this. There are bigger forces at work here, as we shall see. Later, I will describe a theory that automatically converts all sources of static gender imbalance into sources of growing gender imbalance, without having to cook up a special story in each case.

The Angry Blogger Theory

Before we go further, I should definitely mention the theory that started the debate, although it strikes me as rather implausible. The idea is that the crossword blogosphere has become more harsh, which makes it less fun to construct crossword puzzles, and women are deterred by the harsh criticism more than men. This theory, to its credit, came with a testable prediction which the theorizer actually tested, namely that a higher fraction of women would express their dissatisfaction with the blogosphere. Unfortunately, the sample size was very low. In particular, the results were that 15 of 19 men were glad to have their puzzles reviewed by a particularly harsh blogger, while 2 out of 4 women were. Which means almost nothing.

For the record, it is true that the crossword blogs have become more harsh. But while it may (or may not) be true that women on average are more dissatisfied with this trend, blog reviews are but one reason of many to publish a puzzle. At best, this seems like it would be a minor contributing factor, not the dominant driver of the trend. Nevertheless, let us see if we can design a feasible test that targets the link between blogging mood and publication rates a bit more directly.

First, although there are ways of measuring the "mood" of the blogs over time with textual analysis, the time series is likely to be extremely noisy. I doubt we will ever have enough power to say much from this angle. Instead, let us simply assume that the mood has been trending downward over time (an uncontroversial assumption if you read crossword blogs), and look for some lever that causes constructors to care more or less about that mood.

Well, before I was published, my primary concern by a long shot was getting published. Being published at all is just miles ahead of anything else in terms of value. After a little while, I got used to clearing the publishing bar, and the actual reception of my puzzles became a much bigger deal than the simple act of getting published again. Thus, it seems reasonable to assume that the mood of the blogosphere has less influence on first-time publication rates than on subsequent publication rates.

Given this, a simple test of the theory is to look at a difference-in-difference for the attrition rates of men vs women, now vs 5 years ago. As a first attempt, we could define the attrition rate as (number of people with no puzzles in year t)/(number of people with their first puzzle in year t-1). Think of it as the number of people who quit after just a few puzzles; if women are quickly deterred by angry bloggers, then their attrition rate is high. The idea is to compute this separately for men and women, this year and 5 years ago — over which time the blogging mood has fallen sharply. If women's attrition rate has risen faster than men's over the last 5 years, then that is good evidence that blogging mood has something to do with it.

To reiterate, there are many reasons why publication rates may differ between men and women, and why they may change over time. But most competing theories do not predict a differential effect for first-time publication rates vs subsequent publication rates. Thus, the above test can differentiate the Angry Blogger theory from other theories. I don't expect the test to support the theory, but I haven't actually performed it.

The Editorial Bias Theory

The possibility of gender bias in the editing process has also been considered. I won't dwell on it because it was not hotly debated, even though most of the current editors of mainstream crosswords are male. To say it again, since it's so easy to forget, you would have to argue that the continuing male editorship is a source of increasing bias, not just a source of bias. I'm not saying static bias isn't important more generally, but on its own, it won't help us resolve the current puzzle.

The Women and Computers Theory

You can probably guess that this one has generated some controversy.

What I haven't mentioned yet is that crossword construction has become much more computerized than it used to be. By and large, this is a good thing, as it has ushered in an era of higher quality puzzles — a golden age if you will. Computers don't yet make puzzles for us (well they can make bad ones, to an extent), but they are invaluable tools that help us to make better crosswords while devoting our energies to the more creative parts of the work.

The theory is that, for whatever reason, women have less affinity for working with computers, so as computerization takes over the puzzlemaking world, the proportion of active women falls. To quote one editor:

Of course, plenty of women are handy with computers, but by tradition the mathematical sort of exercise that crossword constructing is appeals more to men than women.

This is too much of a stretch to me. For one thing, the gender gap seems to be most pronounced among younger crossword constructors, even though technical disciplines are attracting a more balanced pool with every new generation. Furthermore, the gap in crosswording in this age group is apparently worse than the gap in math at the college level, worse than engineering, perhaps even worse than computer science. Yet, compared to math or engineering or programming, crossword construction is barely a technical discipline. This isn't programming, it's using existing software to produce a work of consumable art. If so, we are trying to explain the large gender gap in a slightly technical discipline by appealing to the smaller gender gaps in more technical disciplines. At best it's a partial explanation, and it predicts the opposite trend from what we actually see.

I could have my stylized facts wrong. But in any case, this theory seems very difficult to test, as the rise of puzzlemaking software is deeply entangled with other changes over the last 20 years. I'm going to move on to my own theory, which subsumes several of the above theories and reconciles many of the bits and pieces.

The Crowded Pool Theory

Suppose that Will Shortz changes the submission policy at the New York Times. Henceforth, everyone with puzzlemaking ability will be thrown into a giant swimming pool. Each year, the fastest 100 people up the ladder will be published. What do you think will happen?

Well, it sort of depends how many people are in the pool. If only 100, we're looking at an orderly affair where everyone is published. But with 1000 people, it becomes a mad dash, and whoever fights the hardest to escape gets published.

I think it used to be the case that to make a good crossword puzzle, you had to be really exceptionally good with words. This was (and is) an elite and fairly gender-balanced group, because it’s heavily based on innate ability, which is pretty balanced. Furthermore, the very best wordsmiths are naturally inclined to make puzzles — because, after all, they’re very interested in words — and so submissions would have been relatively evenly distributed, too.

In a manner of speaking, the swimming pool was filled with a relatively equal mix of 100 men and women. Nobody else could join because they couldn't make puzzles.

Nowadays, puzzlemaking is much easier, and much less correlated with innate ability. With computer assistance, it’s just not as hard anymore. The pool of people who can make a puzzle is now much larger than the pool of people who actually do. That creates some space for whatever pervasive factors that cause gender imbalance everywhere to wreak greater havoc in our little field. Now societal factors can strongly influence who gets selected to actually make the puzzles, from that large pool of candidates.

In other words, the pool now has a lot more than 100 people in it. Even the original, highly skilled 100 puzzlemakers can be beat out by newcomers who are willing to fight harder, now that the playing field has been somewhat leveled.

This is a compelling theory that plays nice with other theories. For instance, consider the Hobby theory above. That could explain one reason why men are fighting harder than women, namely that they have more free time to spend as they like. And of course, many other reasons for static gender imbalance have been proposed. Combined with this theory, they are now converted into arguments for an increasing trend. Nifty!

Others have said that gender imbalance is a complicated issue that we can't possibly hope to capture in a simple theory. The Crowded Pool theory does not deny that; in fact, it makes no comment about the underlying causes of the imbalance. It simply takes the static picture as a given and uses it to produce the dynamic picture that we see. So the theory plays nice with these folks too.

Or consider the Women and Computers theory. The current theory says that the growing gender gap is connected to the rise in computerized puzzlemaking, and thus it lends some justification to the people who thought the two trends had to be linked. However, we now take a very different view as to the nature of that link.

That is, computers are an indirect cause of the growing imbalance, but only in the sense that they make constructing easier (for everyone!), not because men have a greater "affinity" for computers. Indeed, any factor that brings puzzlemaking to a larger audience would have the same effect, whether or not it has anything to do with computers.

The Women and Computers theory is difficult to test. But here's a reasonable test of the Crowded Pool theory. The release of the Wordplay documentary resulted in a large discrete jump in the pool of potential puzzlemakers. Puzzlemaking was thrust onto the radar of thousands of people who'd never considered it before. Oh, anyone can just submit a puzzle? Even me? Looks like fun! Can you hear the collective splash as thousands of new bodies jump into the pool? If my theory is correct, we should see a corresponding discrete jump in the proportion of puzzles published by men, even though there was no discrete change in puzzlemaking technology at the time.

My theory also predicts these observable facts:

To be sure, I'm not saying it's easy to be published in the New York Times. Competition is high and heightening. But at the same time, it's becoming possible to a much broader class of people, if they take the necessary steps to get up the ladder. I don’t know all the underlying reasons why men are submitting in vastly greater numbers than women. But one essential prerequisite is surely that they can. And for most of them, that wasn't always the case.