Hacking Education: Part 2. The Curious Case of the Mrs

Filed under: Hacking Education — Lisa @ 11:50 pm June 29, 2011

This is part 2 of the “Hacking Education” series exploring the DonorsChoose dataset. If you missed part 1, check it out here.

Differences between men and women are plentiful, so it’s not surprising for there to be differences between male and female teachers. But are there differences between married and unmarried teachers? More specifically, are there differences between teachers with prefix “Ms” and teachers with prefix “Mrs”?

In the DonorsChoose data set there certainly are. For one, a project posted by a “Mrs” has about 7% fewer chances of being fully funded than those posted by her male or unmarried colleagues. Even after accounting for other factors described in part 1, having a teacher prefix of “Mrs” lowers the odds of project completion by 10% as compared to “Ms”, and 15% as compared to “Mr”. In this post, we illustrate other differences  between “Ms”, “Mrs”, and “Mr”.

In the earlier days of DonorsChoose, there were about 5% more “Ms” than “Mrs”. Now, the “Mrs” have caught on and have overtaken “Ms” in terms of percentage of project posted. It’s not that the teachers are getting married, either: there was one unique teacher prefix per teacher account in the dataset.

Time of arrival on DonorsChoose is not the only difference between different types of teachers. Below chart shows the breakdown in project resource type, grade level, school metro, poverty level, and primary/secondary focus area for each of “Mr”, “Ms”, and “Mrs”.

Unsurprisingly, male teachers are more than 10% likelier to ask for technology than female teachers, while female teachers prefer to ask for supplies. Female teachers (especially “Mrs”s) are also much more likely to teacher lower grades, whereas male teachers tend to teach in high school.

Projects posted by “Mrs”s tend to come from lower poverty schools and schools in rural or suburban areas. The latter is probably due to the culture in rural vs. urban places (and whether women prefer to be called “Ms” or “Mrs”). Female teachers are also more likely to post projects in literacy and language, whereas male teachers are more likely to post projects in math, science, and (surprisingly) music and the arts.

All the differences listed above contribute to why “Mrs”s have lower project completion rate. We’ve already seen that donors are more likely to donate to projects in support of students in higher grades, projects in high-poverty and urban areas, and projects in musics & the arts. But there is no single difference in project type that explains the difference in project completion. The below chart shows the rate of project completion for each project type and teacher prefix. In almost all of these cases, “Mrs” has a lower rate of project completion. This suggests that a combination of factors contribute to Mrs’s lowered rate of project completion.

One likely factor is differences in the project descriptions written by different teachers. The below chart shows how much more or less words of a given LIWC dictionary category a teacher with a particular prefix is likely to use, compared to average.

Some information here is hardly surprising: male teachers are more likely to swear (much more likely than “Mrs”s),express more negative emotions like anger and sadness. On the other hand, “Mrs”s talk more about family, home, and feelings. For some reason, they also use more sexual words (?!?!?!). Ms’s have their unique traits as well — they express more anxiety, but also more positive emotion and talk more about friends.

One other thing to note is the use of pronouns. While male teachers use more first person plural pronouns (“we”, “our”, etc), “Ms”s use more first person singular pronouns (“I”, “my”, etc) and 3rd person pronouns. “Mrs”s falls somewhere in between. I suppose something can be said here about how male and female teachers sees & relates to students.

Technical Notes

Most of the analysis was done on projects posted between 2004 and 2010. I used R and ggplot2 to generate graphs, wordle to generate the word clouds, and python for data manipulation. All the code used for the above analysis is in github. (So if I did something wrong, please let me know.)

 

  • http://isomorphismes.tumblr.com/ isomorphisms

    The differences in “achievement” stuck out to me as, erm, liable to cause a ruckus. Is there a convenient way to colour-code the text by p-value?