Are variable outcomes due to variable inputs?

This is the consequence of research done to address a claim made in one of the book related forums to which I belong. The claim was that women are being overlooked in terms of awards because of award committee bias (despite the fact that most book award committees are either gender balanced or female positive in the case of children's literature). The claim struck me as implausible so I did some digging.

The original question was: Why are the Caldecott Medal winners dominated by male illustrators and is it different if you include honor mentions? The Caldecott is the top award given annually to illustrators of children's books for the best work. The main prize is the Medal and there are usually 2-5 honor mentions each year as well.

Looking at the data, it is correct that there is a gender disparity, and it does not make a difference whether we focus on Medals or honors. Males have won 69% of the Medals and 67% of the Honors.

So my first question was whether this disparity was something distinctive to Caldecott or whether it might be more general. Interestingly, this pattern is not different from other awards. It also carries over into bestsellers where there is no committee making an award, there is simply general public approbation reflected in book sales.

The percentage of authors over recent decades who are male by prize (and omitting dual authorship):

Pulitzer - 61%
PW Fiction Bestsellers – 61%
PW Nonfiction Bestsellers – 86%
Book of the Month Club Selections – 63%
Critically Acclaimed – 86%
Caldecott – 69%
Newbery – 34%
Coretta Scott King – 57%

The odd one out in this pattern is the Newbery (the most prestigious annual award for authors of children's books) with 66% of winners being female. So Caldecott is not out of line with other awards or other forms of public choice (bestsellers). Looking at the norm among prizes, there are really two different questions.

Question 1: Why is there a pattern among prizes and popularity with the general public such that males are represented in the range of 60-85% among winners and most popular?
Question 2: Why is this pattern different for the Newbery?

Question 1: Why is there a pattern among prizes and popularity with the general public such that males are represented in the range of 60-85% among winners and most popular?

I suspect we are dealing with a social and economic artifact here arising from the way people organize their lives and priorities.

The disparity in performance (with low female representation at the highest levels of achievement) is not limited to book awards but shows up repeatedly in other fields of endeavor such as: Pulitzer Journalism Award (14%), Senior Corporate Executives (16% female), Members of the House of Representatives (16%), US Senators (14%), law firm partners (16%), accounting firm partners (18%), surgeons (19%), Engineering/STEM degrees (21%), full professors (23%), published authors (30%), average of literary prizes (35%), etc. In fields where there would likely be virtually no attention paid to gender, this 15-30% representation appears. For example, among the most frequently linked online articles, only 20% are by women. Who pays attention to the gender of articles to which they are linking? There seems to be something fundamental here.

The issue of female workforce attrition has been extensively studied by major corporations over the past three decades. Companies tend to hire equally in proportion to supply but experience much higher attrition among their female employees than their male employees. This represents a major cost (a lot of investment in human capital walks out the door before the benefits accrue from those investments) and companies have a high incentive to solve that problem. The attrition disparity is traceable to family status rather than gender. Single males and single females demonstrate comparable attrition rates, compensation increases, promotion rates etc.

Simplistically, female attrition in the corporate environments is primarily driven by conflicts between family obligations and work commitments, particularly in terms of hours to be worked and flexibility to address unanticipated issues. The consequence of these conflicts is that the male/female ratio in most fields starts out relatively even and then begins to skew, sometimes slowly in less competitive fields, usually steeply in demanding fields. Within five years, the equal ratio has morphed to 2:1 and by ten years to 3:1. These root causes have been known for thirty years and there are plenty of innovative solutions that have been developed to help address the issue. None-the-less, the pattern remains and the solutions, while good for particular individuals, have not made much of a difference to the overall numbers.

The reason this attrition is relevant is related to the role that experience and practice plays in achieved excellence. The two relevant variables are volume of hours of experience/practice and the duration/continuity of practice. Top achievers in every field are virtually always distinguished by the exceptionally high number of focused hours of experience and the duration of that experience. It is relatively rare for someone to appear at the top of the league without these attributes. In athletics, chess, classical music performance, and other easily quantifiable fields, there is a strong predictive correlation between number of hours spent training/practicing, continuity of training and level of achieved outcome. Talent and passion are a prerequisite but volume and continuity are the variables with the largest impact on outcomes regardless of gender.

If volume and continuity are the primary determinants of outcome (given basic competence), then what we need to determine is the split between genders of those that are able to focus full-time on careers over long durations. We can measure volume of full-time work by gender pretty easily. The split in full-time year round workers is 58% male and 42% female. This is a floor as it constitutes only those who work 35 hours a week or more. Ideally we would want to see the ratios at graduated levels: 40 hours a week, 50, 60, 70, 80. There are virtually no top performers in any field that only spend 35 hours on their craft/career. So even though this is a low threshold, it is the data we have and what the data says is that if volume of effort is correlated with outcome, straight off the bat, you would expect that 58% of top prizes would go to males.

The detailed measurement of continuity is much more difficult. Despite that, it is well researched and documented that there is a marked drop-off in full-time female labor force participation around childbearing years. This is particularly topical at the moment owing to Sheryl Sandberg’s book, Lean In. In that book, she indicates that 43% of women in careers exit the labor force for family reasons with only 40% of those returning full-time at some point in the future. Recognizing that this is entirely a back-of-the-envelope calculation, these two figures indicate that women only represent some 29% of the pool of full-time workers who do not have a planned interruption in their career (0.42 full time workers X 0.43 interrupted careers = 18%; 42 full time workers - 18 interrupted careers = 24% of women who work full-time and do not interrupt careers; 58+24 = 82 number of full-time and continuous workers; 24/82 = 29% of full-time and continuous workforce who are women). All of this assumes that non-planned, non-family related career interruptions occur at the same incidence rate between the sexes.

It might be a coincidence that the percentage of female full-time continuous labor force participation of 29% is within the observed range of 15-30% of top level achievement; it could be, but I suspect not.

The premise is that intense effort as measured by volume of hours and continuity of effort is predictive of significant achievement. The rough data indicate that the proportion of full-time workers without planned career interruptions is approximately 71% male and 29% female. The forecast is that even at this very basic analytical level, not taking into account other variables that are known to affect achievement, that one would expect major achievements in most fields at a ratio of roughly 2:1 male to female. The conclusion is that indeed, observed outcomes across multiple fields of endeavor are in rough proportion to that of full-time continuous workers.

It probably warrants repeating what I have said in the past: 1) this is a good faith effort to work within the constraints of the data that we have, 2) that better quality data or more complete data might change the conclusions, and 3) that averages are not individuals, there are always exceptions.

If this line of logic and evidence holds up, the implication is that if we are to achieve changes in the representation rates at the highest levels of achievement, it is primarily an issue of changing individual life decisions regarding family and careers.


Question 2: Why is this pattern different for the Newbery?

I have no data to answer this one. One might argue, as others do regarding Caldecott, that there is some form of unconscious bias or systemic discrimination, in this case against males versus females. I am highly skeptical of both propositions.

Instead of seeking to explain the outcome as a function of bias and discrimination, my speculation follows from the above discussion. Perhaps the invested hours in motherhood and particularly the direct care and nurturing of children, serves in some fashion as a form of intense practice. That is to say that the number of hours spent with children provide, perhaps, some form of insight into the language and cognitive development and concerns of children which is a beneficial competency in terms of writing for children. That would be consistent with the argument and the data but it seems just a little too pat. This could be a valid supposition but I don’t set great store by it yet simply because the nature of authorship, authorial success, book consequence, etc. are such murky processes.

One testable prediction that would arise if the supposition is true is that the average age of female winners of the Newbery ought to be 5-15 years older than the average of male winners.

A second testable prediction arising from this hypothesis is that the percentage of female winners of the Newbery who are mothers ought to be higher than female winners in other fields of endeavor such as corporate executives, partners in law firms, etc.



Why is any of this important? Whenever I see people exercised about disparate impact it seems almost uniformly the case that they tautologically assume that the cause of the disparity is somewhere in the process (how awards are made, how reporting assignments are distributed, how people are mentored and promoted, etc.) and usually assume either bad faith on the part of the participants (they are biased), or assume that there is "unconscious bias", a deus ex machina if ever there was one. This orientation also leads to what seems to me to be excessive speculation: women lack confidence, women are not aggressive enough negotiators, women don't persevere, etc.

Rarely does there seem to be an examination of context.

If you are a production manager at a factory and your production output has too high a failure rate, you have two primary suspects. The too high failure rate likely has something to do with your process or it has something to do with your inputs. Sure, perhaps you have uniformly excellent input and your process is shredding stuff along the way leading to a high failure rate. Alternatively, perhaps the process is working fine but you have too high a quality variability on your input.

All the disparity arguments seem to focus solely on process and never on input.

Until you know the root causes for the failure rate, you won't be able to adequately address the disparity. If you accept the premise that voluminous and continuous effort is associated with quality outcomes, then you have to look at the disparity of inputs to get a sense of the disparity of outputs.

If only 30% of authors who work full time and continuously are female, then you are likely not to see more than 30% of award winners or bestsellers being female.