Advocates for more missing middle housing routinely claim that decreasing the barriers to building new houses will lead to (slightly) lower rents. The underlying logic is that more supply, all else being equal, will lead to a new, and lower equilibrium price.
Opponents of missing middle housing will point to cities and towns with increasing rents and say that adding more housing is pointless as the general price trend is upwards even in the places where missing middle housing policies are in effect.
This is a case of looking at the factual instead of the counterfactual.
The previous sentence is extremely geeky and a bit obtuse. But it is critically important when thinking about policy. We can only do some things and not others but we have to decide among a wide array of options, some of which won’t actually happen because we decided not to do that particular thing.
We want to know what would have happened in an alternative universe. We are currently unable to access the multiverse. Therefore we can never observe what did not happen. Instead we need to make a strong guess at what happened in that alternative universe with what we can actually see in our universe. There is a massive scientific field of causal inference that deals with this problem. Plenty of folks at Duke and UNC spend every working minute thinking about measuring the forever unobservable counterfactual.
The classic way to approach this problem is to conduct a randomized control trial. Here we randomly assign an intervention, like a vaccine, to one of two groups. Some people consent to receive the vaccine without knowing that they received the vaccine. Other people also consent to be randomly not given the vaccine. The study scientists then observe the relevant outcomes. If there is a meaningful difference between the average outcome of the two groups, we can feel pretty confident that the intervention actually caused the change because all the other factors that could explain the change are randomly and nearly uniformly distributed between the two groups.
We can’t randomly assign people or programs in most real world situations. There are usually massive ethical and pragmatic problems with real world randomization. So what do we do?
We use quasi-experiments where we can create a pseudo randomized counterfactual of what would have happened without the intervention. For my research into health insurance, I often use a method known as difference in difference where we look at groups that receive an intervention versus those that don’t by looking at trends before and after the intervention happens. If the trends between the two groups change after the intervention by a big enough margin, we think the intervention probably caused it. I also use a method known as regression discontinuity where we look at changes in behavior near a boundary. If there is a big behavior change at a boundary then we should think that the difference in policies likely drives that behavior.
A famous health economics study looks at the survival of very low birth weight babies and the effect of medical technology by looking at the 1500 gram boundary. Babies under 1500 grams were automatically admitted to a NICU while babies above 1500 grams were not automatically admitted to the NICU. Once they cleaned the data, the researchers found that the decision to make 1500 grams a boundary led to substantially higher mortality and medical expenditures for babies with birth weights just over 1500 grams relative to babies with birth weights just under 1500 grams. The babies just under 1500 grans received more medical care that led to better health outcomes.
Another study looked at home prices by elementary school assignment zones. Houses across both a street and an assignment boundary had wildly different resale values than houses on the same street so they conclude that school assignment zones are really valuable.
There are a bunch of other methods that are used in quasi-experimental designs. But they all rely on constructing either an explicit or implicit counterfactual and then seeing how reality reacts against that counterfactual. If there is a difference, then researchers say the policy likely had a causal effect.
This was a bit of a detour into what drives many doctoral students to long, painful and overly caffeinated nights for a civics blog. But we need clarity when we’re talking about policy analysis as to what is being claimed. The image below is a toy model of how advocates for more housing think when they make claims that more housing will lead to lower price levels in the future.
I’m assuming a general upwards trend in price over time. I think this is a reasonable assumption in Chapel Hill as the region as a whole is growing in population and wealth while Chapel Hill has a great set of amenities including awesome schools, good greenways and a world class university. People like living in Chapel Hill. Lots of people want to live here but can’t afford to live here.
The Green line is a projected trend line that grows over time. Prices will be higher next year than this year, and higher two years from now then both this year and next year. However if there is more supply, the rate of growth would decrease and over time a gap would noticeably open up between the actual world with the policy regime of more housing built and the counterfactual, unobserved world of no change in policy. This is the argument that is made in support of more missing middle housing.
But at the same time, in either the counterfactual world of more housing or the actual world of no change in policy as shown above, prices are going up. It just differs by how much! So we need to know if we are arguing against the actual, the counterfactual, or current baseline as those are three very different conversations that lead to people talking right past each other.