This past month or so I've been pretty busy, hence the radio silence with respect to the blog. I moved to a new city, I got a new apartment, and I'm finally semi-employed.
On this latter point, my new job has a serendipitously coincidental title that relates to this post. I am now a family medicine resident at the University of Toronto. The etymology of medical resident is very old and harks back to medical training programs where they would work their medical learners for such long hours that they literally had to live at the hospital - ergo the term "resident" or "house staff". The first recorded use of these terms is in Edinburgh in 1744 at the Royal Infirmary where a clerk of the house or resident of the house carried out duties that resemble what medical residents do today. The first doctor who held this title got into a fight with the head of nursing at the hospital and was forced to resign as a result. The eternal war between nurses and residents has deep roots.
I am now a resident in Toronto in two separate senses but this was not so about about three months ago. Over the May long weekend I had the good fortune of hanging out with some old friends from grad school for an annual cabin weekend (I've been told that as a new Ontarian I now have to call it a cottage). Naturally, as economists, most of the conversations revolved around topics that would put most cocaine-enthusiast insomniacs to sleep. It was the most fun I've had in a long time.
Several people, including myself, were moving or planning on moving, and so for one night the conversation subconsciously gravitated to housing. Now, at the time, I was going to be a homeless person living in Toronto because I had yet to be organized enough to find an apartment. This procrastination went on until the day I arrived in Toronto when I managed to sign a lease. This was about four days before I started my medical residency. I then bounced the first cheque I gave my new landlord. It has not been an auspicious start to my life here. But at the time the conversation struck a chord because of my future lack of housing.
The topic that we were arguing about at the cabin (cottage) was housing for people who really are in need of housing. Not because they can't organize themselves out of a paper bag but because they truly can't afford it. There had been a recent article in the popular press about subsidized housing in Moose Jaw. The city provided social housing to the homeless and saw the costs of taking care of these people plummet. To add to this, somebody had been finishing up a research project on the homeless in southern Quebec. Much discussion ensued about whether housing really was the number one priority for the homeless.
There is some evidence on the impact of giving housing to the homeless and less wealthy. Recent preliminary data from a Canadian randomized control trial corroborates how improvements in health can come with housing the homeless who have mental illnesses. The Moving to Opportunity trial, which recently published results showed that there were significant reductions in obesity, diabetes and crime in persons who moved to better neighbourhoods in the United States. Interestingly, there was no improvement in the earnings of these movers, which suggests these effects were not mediated by income.
I can't run a randomized control trial except maybe on myself and that wouldn't get past my own ethics board because I don't really want to sleep on the street. But there is a decent amount of data at the neighbourhood level on the OpenData Toronto website on housing and in particular on housing insecurity.
In Toronto there are several programs that are administered by the city that are meant to prevent homelessness. Two of these are rent bank loans and subsidized housing. The former is a program that provides short-term bridge loans that are meant to be limited and eventually paid back. These are given when all other sources of rent help are exhausted and the nature of the program is meant to help get people back to a point where they can pay for their apartments themselves. When the number of applicants for this program go up in a neighbourhood, that neighbourhood has a higher level of short-term housing insecurity.
The latter program, subsidized housing, is meant as a pure transfer program. Rent is geared towards income, which is to say that the city does not expect you to pay a market price to use the apartment. It is designed to be a pure transfer, but because of this, there are long wait lists to get into subsidized housing. People applying for these programs know that in the long run they will need this housing because their earning potential for whatever reason will be low. In neighbourhoods where the wait list is high relative to the number of subsidized units there should be a higher level of long-term housing insecurity.
I'm going to focus on how housing insecurity affects one major health outcome: the teenage pregnancy rate. Of all of the health factors in the OpenData Toronto repository, this is by far the most interesting. I only have neighbourhood-level data for 2008 so this is a cross-sectional relationship, which should give some cause for skepticism. Nevertheless, it really is an interesting relationship.
First, what is the geography of teen pregnancy in Toronto? Another great tip I got on my cabin (cottage) weekend from a guy who works in economic consulting was that, outside of economics, nobody cares about anything other than the map. You can show them as many regression results or tables as you want but none of it is going to get through until you get to the slide with the map on it. So here is a map.
A little orientation: this is a map of Toronto broken into its neighbourhoods. The bottom of this map is lake Ontario - Toronto Island is just south of downtown at the very centre of the map - it looks like a small Japan. To the left of Toronto are the suburbs that extend along the lake (sort of southwest-ish) through Mississauga and eventually down to Hamilton. To the right is the same with Ajax and Oshawa. Above Toronto is what is adorably referred to as "rural Ontario" by Torontonians. Also up there are Markham and Vaughn.
This map shows the rate of pregnancy per 1,000 female teens (aged 15-19) for each neighbourhood. The top three neighbourhoods are Broadview North (77 per 1000), Beechborough-Greenbrook (70 per 1000) and Moss Park (59 per 1000). Broadview North is that dark area right at the centre of the map. Beechborough-Greenbrook is among those darker neighbourhoods at the Northwest corner of Toronto. Moss Park is closer to the downtown area, just north of Toronto island.
Now a map of the first measure: the number of rent bank applications in a neighbourhood. Like I said earlier, this should be a reflection of the short-term housing instability. This is also population standardized so that it's the number of people who applied for the rent bank divided by the population in a neighbourhood.
On this map, darker means a higher number of rent bank applicants. There is a cluster of high rent bank applicant neighbourhoods in the northwest as well as closer to Scarborough on the east side of Toronto. If you control for neighbourhood-level income and the number of health providers (a rough reflection of access to health care advice), a 10% increase in the number of rent bank applicants leads to an increase in the teen pregnancy rate by about 14%. Not huge, but statistically significant. Interestingly, the same effect shows up in the general female fertility rate. Higher prevalence of short-term housing instability is associated with higher levels of pregnancy.
Similar results have been seen with income in other more real studies. That is to say when teenage girls live in families on the lower end of the income spectrum, they tend to have more pregnancies. There is significant debate over why this is because it's not a causal relationship in the true sense of the word. Having less money does not hypnotize teens into having sex with one another. One explanation is that teens from disadvantaged families look out into the world and see little opportunity for any other kind of social advancement or life improvement. They usually have bad jobs, little access to education, etc. Having a kid is relatively easy and kids can be very rewarding. Just ask my parents. So teens in these situations have kids. But low income is a red herring variable that really reflects the lack of opportunity teens have. Housing instability could be the same way, so finding this result in short term housing instability makes some sense.
Similar results have been seen with income in other more real studies. That is to say when teenage girls live in families on the lower end of the income spectrum, they tend to have more pregnancies. There is significant debate over why this is because it's not a causal relationship in the true sense of the word. Having less money does not hypnotize teens into having sex with one another. One explanation is that teens from disadvantaged families look out into the world and see little opportunity for any other kind of social advancement or life improvement. They usually have bad jobs, little access to education, etc. Having a kid is relatively easy and kids can be very rewarding. Just ask my parents. So teens in these situations have kids. But low income is a red herring variable that really reflects the lack of opportunity teens have. Housing instability could be the same way, so finding this result in short term housing instability makes some sense.
Next is the ratio of the social housing wait list to number of social housing units in a neighbourhood. This should be a reflection of a neighbourhood's long-term housing instability.
Again, darker means a higher ratio of social housing applicants to social housing units. And again, maybe there is a cluster in the northwest of the city, where some neighbourhoods have high numbers of people applying to social housing with few units. There are also some scattered neighbourhoods that have this high ratio. In the regression analysis though, the opposite relationship exists between this measure and teenage pregnancy. When the prevalence of long-term housing insecurity rises in a neighbourhood, the teenage pregnancy rate goes down after controlling for income and health care providers. A 10% rise in the prevalence of long-term housing insecurity leads to a decrease in the teen pregnancy rate by about 11%.
So what explains this relationship - why is it that short-term housing insecurity leads to higher teen pregnancy rates but long-term housing insecurity leads to lower teen pregnancy rates? I have no idea. Some speculation follows.
Let's consider that the relationship between long-term housing insecurity and teen pregnancy is wrong. In this scenario, the real relationship should echo the relationship between short-term housing instability and teen pregnancy. Hypothetically then, higher levels of instability should lead to higher levels of teen pregnancy.
First: is the fact that we see the reverse relationship the result of an omitted variable issue? Basically, did I forget to control for some confounder that should be in the regression equation? All regression equations are almost always misspecified in some way and there are certainly some reasons why, at the neighbourhood level, housing insecurity might be correlated with some omitted variable. If this is the case though, housing insecurity should be negatively correlated with whatever this variable is so that the relationship is underestimated.
One of the reasons I can think of (although I really hate this explanation) is a cultural one. In walking around Toronto, you get the feeling that expat communities tend to cluster in neighbourhoods. It may be that certain cultural communities tend to have differential rates of using social services and at the same time, have differential rates of teen pregnancy. As an example (just speculative), imagine that Portuguese or Italian Catholic neighbourhoods have high rates of teen pregnancy because of religious bans on contraception. At the same time, the religious community is very active in helping people find social housing when they need it. We would observe in the regression model that these neighbourhoods would have both high pregnancy rates and low wait times. As housing wait lists go down, teen pregnancy goes up, which in the regression model would be falsely ascribed to the housing insecurity rather than Portuguese culture. For what it's worth, I tried throwing in a bunch of variables controlling for culture or proxies of culture and nothing really changed the result.
Second, what if there's an element of reverse causality in this regression equation? This could come for a couple of reasons. It may be that teens who have kids or who are pregnant are less likely to apply for social housing. They might just move in with their parents. But alternatively it may be a reason related to city planning. Despite Rob Ford's best efforts, the city administration did not conduct city planning by throwing darts at a map of Toronto. They likely had a plan of how to deal with teen pregnancy as well as other markers of social upheaval. If they used the teen pregnancy rate as a way to allocate social housing to neighbourhoods, then places with high teen pregnancy rates would get first dibs at new social housing. This would lead to lower social housing wait lists and we would falsely observe that this caused higher teen pregnancy rates rather than the other way around.
For an arbitrary amount of insecurity, deemed to be "short-term", teens make the decision to have a kid because things are not so bad. But once you go beyond a certain point, the relationship flips and things get progressively worse to the point that this dissuades teens from having pregnancies.
Alternately though, perhaps the function of the long-term social housing help has exactly the opposite effect. Teens who are in families who qualify for housing support have significantly more long-term options because of the fact that they are being supported and so decide to delay pregnancy to take advantage of now being able to go to college, get better jobs, etc. Social housing is a promise that support is coming and will be around for a while whereas the point of the loan is, in the words of George Clooney, "to ferry wounded souls across the river of dread until the point were hope is dimly visible. And then stop the boat, shove them in the water and make them swim." Loans are not a promise of long-term help, but social housing is and perhaps might be taken as a signal of better prospects.
I'll finish this post using the conclusion of almost every medical paper I've ever read. More research is required.
First: is the fact that we see the reverse relationship the result of an omitted variable issue? Basically, did I forget to control for some confounder that should be in the regression equation? All regression equations are almost always misspecified in some way and there are certainly some reasons why, at the neighbourhood level, housing insecurity might be correlated with some omitted variable. If this is the case though, housing insecurity should be negatively correlated with whatever this variable is so that the relationship is underestimated.
One of the reasons I can think of (although I really hate this explanation) is a cultural one. In walking around Toronto, you get the feeling that expat communities tend to cluster in neighbourhoods. It may be that certain cultural communities tend to have differential rates of using social services and at the same time, have differential rates of teen pregnancy. As an example (just speculative), imagine that Portuguese or Italian Catholic neighbourhoods have high rates of teen pregnancy because of religious bans on contraception. At the same time, the religious community is very active in helping people find social housing when they need it. We would observe in the regression model that these neighbourhoods would have both high pregnancy rates and low wait times. As housing wait lists go down, teen pregnancy goes up, which in the regression model would be falsely ascribed to the housing insecurity rather than Portuguese culture. For what it's worth, I tried throwing in a bunch of variables controlling for culture or proxies of culture and nothing really changed the result.
Second, what if there's an element of reverse causality in this regression equation? This could come for a couple of reasons. It may be that teens who have kids or who are pregnant are less likely to apply for social housing. They might just move in with their parents. But alternatively it may be a reason related to city planning. Despite Rob Ford's best efforts, the city administration did not conduct city planning by throwing darts at a map of Toronto. They likely had a plan of how to deal with teen pregnancy as well as other markers of social upheaval. If they used the teen pregnancy rate as a way to allocate social housing to neighbourhoods, then places with high teen pregnancy rates would get first dibs at new social housing. This would lead to lower social housing wait lists and we would falsely observe that this caused higher teen pregnancy rates rather than the other way around.
Finally, what if this is a true relationship in the way that there is just something about long-term housing insecurity that means that teenagers get pregnant less? There are two major reasons I can think of why this would be the case - one is more depressing and the other is more uplifting.
First the depressing option. What we might be observing in this data is that the social situation of these teens is so bad that they can't even contemplate the option of having a kid because they really are that badly off. Teens who have short term housing insecurity might look out into the world and see that their situation is bad, but still ok enough to have a kid. Things might get better down the road and so having a kid is still an option. Teens who have long-term housing insecurity on the other hand are so badly off that having a kid is not even an option given the high investment that doing so requires. The data that I have can't really show that because the measures that I have are prevalence of housing insecurity in a neighbourhood rather than a continuous variable that shows how long the housing insecurity might be (i.e., a family might have housing insecurity for several months because of losing a job vs. a family that has housing insecurity that would be several years because of a family member that has a chronic health issue). You could think of this relationship like this though.
First the depressing option. What we might be observing in this data is that the social situation of these teens is so bad that they can't even contemplate the option of having a kid because they really are that badly off. Teens who have short term housing insecurity might look out into the world and see that their situation is bad, but still ok enough to have a kid. Things might get better down the road and so having a kid is still an option. Teens who have long-term housing insecurity on the other hand are so badly off that having a kid is not even an option given the high investment that doing so requires. The data that I have can't really show that because the measures that I have are prevalence of housing insecurity in a neighbourhood rather than a continuous variable that shows how long the housing insecurity might be (i.e., a family might have housing insecurity for several months because of losing a job vs. a family that has housing insecurity that would be several years because of a family member that has a chronic health issue). You could think of this relationship like this though.
For an arbitrary amount of insecurity, deemed to be "short-term", teens make the decision to have a kid because things are not so bad. But once you go beyond a certain point, the relationship flips and things get progressively worse to the point that this dissuades teens from having pregnancies.
Alternately though, perhaps the function of the long-term social housing help has exactly the opposite effect. Teens who are in families who qualify for housing support have significantly more long-term options because of the fact that they are being supported and so decide to delay pregnancy to take advantage of now being able to go to college, get better jobs, etc. Social housing is a promise that support is coming and will be around for a while whereas the point of the loan is, in the words of George Clooney, "to ferry wounded souls across the river of dread until the point were hope is dimly visible. And then stop the boat, shove them in the water and make them swim." Loans are not a promise of long-term help, but social housing is and perhaps might be taken as a signal of better prospects.
I'll finish this post using the conclusion of almost every medical paper I've ever read. More research is required.
