Week Four:

Hello again to those people who have managed to slog this far into my project. Given the relative length and amount of work I have put into the last few blog posts, I have made the executive decision to combine what would likely be two or three separate blog posts into one. So, strap in boys and girls, and make yourselves comfortable, because this is going to be a long one.

It seems like every week I have found some major new discovery when it comes to this project, one that upends basically everything I have done up to that point (think 2000 or so words worth of analysis going down the tubes). This week is no exception. In addition to performing my analysis on my data, some elements of which I have discussed in previous blog posts, I have also been working on my literature review. This is why last week I decided to check the USDA models for rurality, and this is why I spent time this week on adding a datapoint which the literature indicated would be important but I didn’t think I would ever be able to find. That datapoint was the level of the education of the parents of kids in public schools.

Supposedly the Census Bureau had this data, but if it did, it sure as heck wasn’t sharing it with the public. Instead I managed to get the Institute for Education Science’s National Center for Education Statistics’ Education Demographic and Geographic Estimates database to tell me this data, and it managed to get this dataset from the Census Bureau. But I digress.

This dataset did what you might expect, and explained some things while also making others a whole lot more complicated.

Parent educational attaiment is significant and very useful in understanding English scores and pass rates, and is not significant for math. With the integration of parent educational attainment, we can explain 71% of the variation in English scores, as well as 62% of English pass rates. I believe this would be a good time to remind you, dear reader, that, in public policy (which my analysis arguably is) an explanation which can explain 30% is considered somewhat impressive.

One interesting trend in both of the English tests is that the percentage of the population with a Bachelor’s degree or higher was a better explanation of test scores and pass rates than the percentage of the population which graduated high school. I don’t entirely know what to make of this little factoid, but I will continue my analysis (feel free to post any thoughts you might have on this issue in the comments).

Additionally, the percentage of parents who have a Bachelor’s degree or higher is significant for math pass rates, but only when we regress on percentage white as opposed to percentage African American. If you are reading this, please write “Charlie is not screaming into an empty void” into the comments for a chance at a special prize. Even when we regress on percentage white, the R-Squared value (the value which is used to determine how well models model reality) only jumps by one point from 0.55 to 0.56. I believe that this might just be some weird artifact of how I’m handling other variables but I am not certain.

And that’s about all the interesting things that have happened this week in regards to my project. I’ve honestly just been focusing on doing some of the smaller regressions that I needed to do in order to rewrite my analysis section for most of this week. While some of the things I found out over the course of doing my analysis section are interesting (indeed, some of them were even mentioned previously in this blog post) I don’t think that readers like you would honestly appreciate having me go into great detail about why the percentage of funding for each school district which comes from the state is just as good if not better for the models I am analyzing than the percentage of school funding coming from the districts themselves.

So, that’s about it. That’s all I have for you this week.