On mixing covid-19 and census data

Doing this with Canadian data is a tall order. But since people are doing it, we might as well explain some of the ins and outs.

Mixing census data with COVID-19 case and mortality data seems like an obvious thing to do when trying to understand how COVID-19 affects different groups. But it’s only of very limited use. COVID-19 data is only (openly) available on coarse geographies and can only be matched at the ecological level. Deriving individual level relationships from this is extremely ambitious. At best, it can inform decisions on what individual level data should be collected moving forward.

Projections and self-fulfilling prophecies

Housing and population growth are endogenous in high-demand areas. Which gives cities the tools to exclude people, but should they? Deciding how to grow is a values question, not a technocratic one.

(Joint with Nathan Lauster and cross-posted at HomeFreeSociology) When people want to live in your city, how many should you let in? On the one hand, this is a moral question. Do you have an obligation to people who don’t already live here? On the other hand, it’s a moot question. At least in Canada, cities don’t have the power to control migration. BUT WAIT! Cities DO have power over how many new dwellings to allow.

Toward universal TongFen: Change in polling district voting patterns

Expanding tongfen to arbitrary geometries, with an example application to Canadian federal election polling districts.

Geographic data often comes on different geographic breakdowns. A prime example is census data, where the underlying census geographies can change from census year to census year. This makes it difficult to compare census data across censuses. But comparing census data across censuses at fine geographies is important for many applications. There are two main ways how people deal with this problem. 1. Estimate data for one of the two geographies by (usually at some point) relying on area-weighted interpolation.

COVID-19 data in Canada, effective reproduction rates

A follow-up on what data we need.

We have written about the situation of covid-19 data in Canada previously, and the need for good data is becoming more pressing as we are poised to slowly open up some of our restrictions and need to closely monitor how the spread of COVID-19 is responding. A key number to watch is the effective reproduction rate, the average number of people an infected person passes on the virus ($R_0$). Our collective social distancing has led to the effective reproduction rate to drop below 1, so the spread of the virus is receding.