What a busy week. I took on two side consulting jobs (both very short-term, and I feel kind of guilty about taking them on -- but this is a recession/depression, right? Time to start saving for when I join the ranks of the unemployed?) So now I'm two posts in the hole -- the dissertation post that should have happened Tuesday, and today's recipe-related post. Thankfully, I have things for both--dissertation first.
I would like to note that I write this instead of watching All My Children, which is my usual procrastination tool -- so this is a victory of grand proportions! I would also like to add a photo, since posts without photos are lame. This is a grainy cell phone shot of Rocky, in his fantastic new sweater (Go Huskies!) knit by his foster mom while we were out of town. You can't see his face in this because he was intensely trying to pull us back to Elizabeth's house when it became clear to him that we were taking him home and she was not coming.
Anyway, I'm narrowing in on the research questions for the survey-related paper (and feeling the specter of all the data work I need to do creeping up behind me). A couple of things are pretty obvious: looking into what people are thinking when they apply for the voucher program and when they actually get the voucher in hand, and to how they perceive their options.
But a couple of things are less obvious -- like, how to look at outcomes themselves. For the cross-MSA paper (my dissertation consists of two papers -- here's a summary) I'm comparing where voucher people end up with where non-voucher people end up: other poor people who don't have a voucher, the average renter, and the locations of subsidized housing units. MSA's are the census data term for metropolitan areas. They are pretty big and include a whole commuting area as opposed to just city boundaries.
But for the survey paper I'm looking at the possible outcomes for voucher people themselves. There are a few possible outcomes that they can have. They can use it to stay in their current apartment ("leasing in place," and cutting their housing expenses), they can move someplace really close, like the same neighborhood (I'm calling this a "status quo move"), and they can move to a new neighborhood ... or city, or state. I'm calling this a "potential mobility move." And then, of course, people can fail to find housing at all with their voucher--called "failure leasing up," generally.
These are the things that can be measured and potentially modeled. First, what is the success rate for participants, and does it vary by group (race, income, family size)? Second, what does mobility look like -- leasing in place vs. status quo moves vs. mobility moves? Again by groups. And finally, for potential mobility moves, how many result in significant changes in neighborhood quality? I'll have both the full population of a few thousand households to look at, as well as the smaller sample of people who fill out surveys and tell me about their plans. Yikes.
This is getting long --so here's another photo: Jake an my feet with snowshoes on them. Snowshoes are intense.
So let's say some do move somewhere "new," i.e., in a different census tract from where they started (my back of the envelope numbers look like about two thirds of people move somewhere, about two thirds of those moves are to new neighborhoods). The question becomes whether there are changes in neighborhood quality--does the potential for mobility actually result in mobility? How do you characterize quality so that you can know?
I am very far from the first person to think about how to quantify neighborhood quality. But it is never easy. Poverty rates are the fall-back choice for most research -- but in Seattle, the poverty rates really don't vary so dramatically from census tract to census tract that you can pick up differences that seem particularly meaningful. I could be wrong on that, and I'll find out soon enough, but other "quality of life" indicators are probably more insightful. So maybe looking at crime rates? School quality (test scores?)? Access to transportation? Park space per capita? I need to get up to speed on local data that could be useful.
Then comes the question of modeling outcomes as opposed to measuring them. What that means is taking, for example, whether or not a household moved to a new neighborhood or not as the outcome of interest (literally a yes/no, 1/0 outcome) and estimating the impact of differences in things like the household's race, income, pre-voucher location, PHA that issued the voucher on the probability that a household moved or not. A household's likelihood of moving or not moving is examined as a function of household characteristics, the housing authority that issued the voucher, and any other things that people theorize as important to people's options and outcomes. You have the outcomes for each household, and you have a range of information about each household, and then you plug them in (with a whole lot of crap in between) to a formula, basically.
I am pretty daunted by this -- all of the econometrics freaks me out and is part of why this all takes me so long. But I think desparation to be done will win out over intimidation by statistics.
Next steps, then, are to keep at it. Start test-running the survey to see if it works. And go back to playing with the data for the first paper. I reached out to a third housing authority to see if they are interested in letting me survey their folks -- I'm afraid that one of my current two won't be issuing many vouchers this year, and it could be interesting to see what a different area might bring. I'm also hoping that conversations with my committee members and a very sharp and nice professor at UW who has offered to give feedback will help stuff gel. This all feels like wading through sand half the time.
Ok! Too long first 2009 dissertation post! Seems fitting, since dissertations are generally too long, I guess. Back to work.