...not that "the right thing" here has any moral implications at all - just finally realizing that I really just need to drop this BLS data, use quarterly LED data with sex and education by industry instead, and add three more years of data that have managed to pile up since I first started working on this harebrained idea. Still, this is going to be a good paper, I think.
Gaaahhh.
I also think my department should trust that I would have done well in classes and just let me work on my own stuff.
I've spent most of this month dealing with this problem over and over again. But it's mostly a problem of my own creation becuase I kind of go about things haphazardly.
ReplyDeleteAlso, I don't even know how you are entertaining the idea of doing research while taking classes.
Well, it ebbs and flows. In the last six months I got the vast majority of the research I got done at the beginning of the semester and during winter break. I'm trying to clear off a few tasks before the semester starts filling up. Then again, I have to admit - I'm not the most disciplined studier in the world.
DeleteI'm hoping next semester (and after that) are very different. My co-author on the NBER chapters recently asked me to write a grant proposal with him to the Sloan Foundation for more S&E labor market work. He already has a grant with them, and he's just proposing more work, so I think it's got a good chance of success. I'm carving out a portion of the project myself - ideally to turn into part of my dissertation eventually. Long story short, next fall and afterwards I may essentially buy out my teaching assistantship with this research money, and not feel guilty about working on research for big chunks of time again. We'll see - but I'm excited about that prospect. American University doesn't really do research assistantships - only TAs. So this could work out very well.
Well, I really needed to update the analysis with more years of data anyway, and since I started it a couple years ago I've familiarized myself with this new data. The current analysis has a lot of noise, and I'm identifying a program impact with a regression discontinuity, so noise is not good. The new data is quarterly rather than annual, and I can break it down by sex and education level. Since I'm working with employment and wage data, all of that should cut down the noise a lot.
ReplyDeleteThe other thing I like about it is that it has new hire statistics in addition to employment. I'm looking at a tax credit that firms get for new hires. The concern on this that a lot of people ignore in analyzing these programs is that new hires may increase, but it crowds out hiring that would have gone on elsewhere in the labor market. So what you want to look at is total employment in the area (so the effect takes all the potential crowding out into account). Since this data has new hires and employment, I can run it looking just at new hires, and then point to how the estimate changes when I look at all employment. That sort of comparison is nice.