McMaster University Increase Female Faculty's Pay

[WestCoastRunner]

Female faculty at McMaster University will be getting a raise after a two-year study showed differences in salary between the sexes at the Hamilton school.

The analysis found that women faculty members earned on average $3,515 less than their male counterparts in 2012 and 2013 – even after adjustments were made based on seniority, tenure, faculty and age.

This discrepancy has also been addressed at UBC University a few years ago. What I would like to see is that this applies to all working sectors across Canada, not just universities.

Edited May 2, 2015 by WestCoastRunner

[Spiderfish]

Female faculty at McMaster University will be getting a raise after a two-year study showed differences in salary between the sexes at the Hamilton school.

Right on. Who says progress is not bring made regarding gender wage disparity!

[Topaz]

Good for them, but they still haven't caught up with the female MPs , who just got a $4000. pay increase.

[-1=e^ipi]

Did they control for number of citations, number of publications or teaching performance? How was seniority determined? Did they just use a simple Mincerian model of experience?

[WestCoastRunner]

Did they control for number of citations, number of publications or teaching performance? How was seniority determined? Did they just use a simple Mincerian model of experience?

Do your own research. You seem to be pretty good at and it seems you have the time.

If you don't agree with their actions tells us why.

[-1=e^ipi]

Do your own research. You seem to be pretty good at and it seems you have the time.

If you don't agree with their actions tells us why.

I know how these studies work. The $3,515 is what is unexplained by the model. This can be obtained using a simple regression, and Oaxaca decomposition or other techniques. If the model does not take into account all factors that can explain differences in income other than discrimination, then some of this unexplained difference, which is concluded to be discrimination, is not actually discrimination.

Given that you listed the other factors in the model as 'seniority, tenure, faculty and age', I suspect that number of citations or papers published was not included (which may be a good indicator as to the amount of effort that different professors put in, which can explain differences in income). Also, based on past experience, often the proxy for experience is in adequate (example: the Mincerian model of experience doesn't take into account number of hours per year worked in the past), and can understate the experience difference between males and females. Asking if such a study has taken into account all relevant parameters is very common.

[cybercoma]

I know how these studies work. The $3,515 is what is unexplained by the model. This can be obtained using a simple regression, and Oaxaca decomposition or other techniques. If the model does not take into account all factors that can explain differences in income other than discrimination, then some of this unexplained difference, which is concluded to be discrimination, is not actually discrimination.

Given that you listed the other factors in the model as 'seniority, tenure, faculty and age', I suspect that number of citations or papers published was not included (which may be a good indicator as to the amount of effort that different professors put in, which can explain differences in income). Also, based on past experience, often the proxy for experience is in adequate (example: the Mincerian model of experience doesn't take into account number of hours per year worked in the past), and can understate the experience difference between males and females. Asking if such a study has taken into account all relevant parameters is very common.

It took about 2 seconds to look up the summary of the study.

http://www.mcmaster.ca/vpacademic/documents/Gender_Pay_Equity_Analysis_Final_Summary_Mar_31_2014.pdf

[cybercoma]

That was the preliminary study that used data only from 2012. Here's McMaster's statement. They expanded the study and judging by their actions, the expanded data was statistically significant. http://macfaculty.ca/wp-content/uploads/2015/04/Joint-Statement-on-Pay-Equity-April-2015.pdf

[-1=e^ipi]

Thank you cybercoma. As expected, citations, papers published and hours per year worked are not taken into account.

I'm not saying that there is no wage discrimination. But I doubt it is overstated due to not taking these factors into account.

[cybercoma]

Thank you cybercoma. As expected, citations, papers published and hours per year worked are not taken into account.

I'm not saying that there is no wage discrimination. But I doubt it is overstated due to not taking these factors into account.

Why would they take into account hours per year worked? That's not how academia works and you know that.

They did look at how many years the professor was in their rank. Citations and papers published are considered under their rank. Promotion is contingent on output, so their rank is an indication of citations and papers published. You don't become a Canada Research Chair by doing nothing.

[cybercoma]

Also, there's an argument to be made about quantity versus quality. Different disciplines have varying publishing rates and exposure. Someone who's studying in a very niche area may not be cited as much as someone who's publishing about popular topics. That doesn't mean the niche knowledge is any less valuable and could end up being more valuable in the long run. Moreover, the number of publications doesn't indicate the quality of the publications or the amount of work that went into them. I've been involved in research projects that took years to complete with few publications. Then there's some projects that don't take any time to complete but may have numerous studies come out of them. Quantifying citations and publications is not a good way to judge a professor's quality, even though that's exactly what they do these days. That's why there's way more academic journals than are necessary and way more studies than there should be. That's why you have people conducting research and breaking up the findings into a dozen publications, when really they could be reported just as effectively in far fewer articles. Until you're tenured, it's a silly game that favours busy menial work, which I feel hurts actual knowledge production and the pursuit of more time-consuming and complex problems.

[-1=e^ipi]

Why would they take into account hours per year worked? That's not how academia works and you know that.

Because it can explain some of the differences in income between two groups of people. There is also precedence for it. Do you want a link to Marie Drolet from statscan explaining the relevance of using inadequate proxies of experience?

They did look at how many years the professor was in their rank. Citations and papers published are considered under their rank. Promotion is contingent on output, so their rank is an indication of citations and papers published. You don't become a Canada Research Chair by doing nothing.

It's an in adequate proxy. Seniority or Tenure may be correlated with number of citations or papers published, but excluding these other factors will result in less variation being explained by the model and therefore a larger 'unexplained' component of the model, which is often concluded to be discrimination. Checking number of citations or papers published isn't even that hard to do...

Also, there's an argument to be made about quantity versus quality. Different disciplines have varying publishing rates and exposure.

Yes and faculty is already being controlled for.

Someone who's studying in a very niche area may not be cited as much as someone who's publishing about popular topics. That doesn't mean the niche knowledge is any less valuable and could end up being more valuable in the long run.

At this point, you seem to be arguing something along the lines of 'because we cannot have a perfect model that explains differences in income, we should make no efforts to improve the model used in the study to get a result that is closer to the truth'. By this logic seniority, tenure, faculty and age don't necessary indicate ability to do productive work, so let's just throw them out the window and do a Justin Trudeau model where you just divide two numbers and claim women early 77% of what men earn.

Moreover, the number of publications doesn't indicate the quality of the publications or the amount of work that went into them.

But number of citations usually does.

Quantifying citations and publications is not a good way to judge a professor's quality, even though that's exactly what they do these days.

Do you have a better proxy to judge a professor's quality? If yes, what is it? If no, then why not just add number of citations to the model to have a better model than what was used by the original study?

[cybercoma]

Because the number of citations will likely be related to the number of publications and the popularity of the subject, not necessarily the quality of the professor. I don't know for sure, but I assume that it probably wouldn't add anything of value to the model because it's either unrelated to their income or it has multicolinearity with other explanatory variables, such as years in rank and whether or not they're a CRC, as well as the amount of time you've been teaching and your age (the longer you've been working the more time you've had to publish and be cited). I would add it in doing the study, but I imagine it would be deleted from model or have little to no effect on your R², especially since they used forward, backward, and stepwise deletion to exclude insignificant variables in the linear regression models.

In any case, it's a minor quibble. Let's say it does have some explanatory power (and you're right, it should be included in the Oaxaca decomposition anyway), how much do you speculate it would explain in the overall scheme of things? Sure you risk overstating the effects of the explanatory variable with Oaxaca decomposition. However, we're talking about $3515 per year. The average salary of professors at McMaster in 2009-10 was $122,856 (source). The raise for female faculty amounts to < 3% increase on the average salary. This is likely overstated as salaries increased since 2009-10 and $3515 is in 2015 dollars, while $122,856 is in 2009-10 dollars. So we have a < 3% increase without number of citations included. How much would including citations in the model affect that final number? Without doing the study myself, I have to imagine that the difference is menial and amounts to splitting hairs. Though you're right. There's still 26% variation unexplained, but then how do you account for personality conflicts and office politics or professors who simply don't care to move up in rank?

Edited May 5, 2015 by cybercoma

[-1=e^ipi]

or it has multicolinearity with other explanatory variables, such as years in rank and whether or not they're a CRC, as well as the amount of time you've been teaching and your age (the longer you've been working the more time you've had to publish and be cited).

Number of citations may be correlated with the other explanatory variables, but that doesn't mean that adding number of citations to the model won't soak up some of the unexplained variation and reduce the unexplained wage difference. Multicollinearity doesn't cause an OLS estimator to be biased nor does it cause the error estimate of the OLS estimator to be biased so it isn't a reason not to include other factors in the model.

In any case, it's a minor quibble. Let's say it does have some explanatory power (and you're right, it should be included in the Oaxaca decomposition anyway), how much do you speculate it would explain in the overall scheme of things? How much would including citations in the model affect that final number?

I don't know. Somewhere between 0% and 100% of the unexplained wage gap probably. But it is possible that the new factors would increase the gender wage gap or turn it negative.

The raise for female faculty amounts to < 3% increase on the average salary.

The raise being small doesn't justify doing the raise. Also, why not lower the salaries of the male professors, rather than raise the female professor wages? In any case, these other factors should have been taken into account in the study. But it was probably politically unfeasible for the decision maker to request that the study include these new factors; as that would cause misogynist claims against the decision maker and would risk the decision maker's job.

but then how do you account for personality conflicts and office politics or professors who simply don't care to move up in rank?

I don't know how to take that into account, or if it is feasible to collect some sort of proxy for it. Sometimes you just have to do the best you can with the data available. Number of citations, number of papers published and hours worked per year are relatively easy to obtain, which is why I think the study should have included these factors.