A romantic tale of inadvertent discovery

There’s something happening here; but what it is ain’t exactly clear

– Buffalo Springfield, For What It’s Worth

If you don’t know where you are going, you’ll end up someplace else.

 – Yogi Berra

The Neighbourhood Change Research Partnership has published a presentation on the correlation between Covid-19 infection patterns and households with mutually dependent adults[1]: COVID-19 & Mutually Dependent Populations in the Toronto Census Metropolitan Area.

This study used custom datasets to illustrate how higher household sizes containing mutually dependent adults correlate closely to the levels of Covid-19 infections.

No other correlation to Covid-19 even comes close, except for the combination of overcrowded conditions and mutually dependent households — a correlation we also explored.

If we are right – and we have very good evidence to support our conclusions —  then most other conclusions about the rates of infection are either wrong or misemphasized.

Here is a sampler of popular opinion that blames the victims in respect of both rate of infection and vaccination:

  • The heavily hit communities are ignoring invitations to get vaccinated.
  • There are more anti-vaxxers in hard hit communities.
  • The heavily hit communities are not social distancing, not wearing masks or not wearing them properly, not hand sanitizing etc.
  • They don’t watch the news, listen to the radio or read the papers so they don’t know what’s going on.
  • They don’t know how to listen to warnings because of poor education, conflicting religious beliefs, etc.

And another sampler that defends the victims:

There may be grains of truth in the popular opinion. Yet, what if those opinions are all less relevant than household size, personal space within homes, and the number of adults living together in the same home?

  • Richer people figure out a way to get vaccinated through connections.
  • Lower income people may not have email, phones, broadband, and access to information.
  • They may not speak English, or have poor social connections.
  • They may distrust the system after centuries of colonization, racism, etc.
  • They are poor and don’t have access to transit, vehicles, directions etc. 

That’s what our evidence shows. We gathered the evidence; we correlated it; we charted it, mapped it, described it, and published it.

That’s how studies are supposed to work. You start out with a hypothesis or have a co-relation in mind that evidence might support. You mine the data, do regression analysis, and you either show an evidence-based correlation or you fail to do so.  

Well – none of that happened in this case. Our discovery of important correlations was purely accidental.

Our story

In 2009, I was a fellow at the Metcalf Foundation and the Foundation was interested in working poverty. I came up with what I thought was a reasonable definition of the working poor and over the next years, published three reports on the subject.

My definition of working poverty only counted individuals living on their own or in nuclear families. I could not tell whether people living in mutual dependency were living in a rich family or poor family. And if not living with their family, I could not tell whether they benefited from living with any family.

So I left the mutually dependent working poor out of the definition, knowing that I was missing a component of the working poor, but with no good way to count it.

In the summer and fall of 2019, I worried more and more about this and mentioned it to Richard Maaranen, the mapper from the Neighbourhood Change Research Partnership who became the co-author of the Covid-19 presentation noted above.

He began to produce maps and charts of mutually dependent adults. Then he moved on to mutually dependent adults who also showed characteristics of working poverty in terms of their incomes. Those early maps, charts, and graphs proved fateful, even though I had no intention of using them at the time.

Then along came Covid-19 in March 2020.

By May 2020, I recall reading the Toronto Star and seeing some Covid-19 maps of Toronto for the first time. I thought they were interesting from the point of view of racialized communities and poverty. At that time, I was consumed with the CERB and CERB claw-backs. The maps showed where CERB claw-backs would be most felt by social assistance and other benefit recipients.

In the late summer, I circled back to the mutual dependency maps, but only those relating to working poverty. These did not relate much to the pipeline of maps that became a steady diet of online news and newspapers. There was nothing to see in the correlation of the mutually dependent working poor to Covid-19.

Then in November last year – for no particular reason – I circled back to the suite of maps and charts, graphs and tables relating to mutually dependent adults. It was on the same day as a new set of maps came out in the newspapers.

I stopped and stared at both – I looked at the Covid-19 maps and then the mutually dependent adult maps and had one of those ‘eureka’ moments.

I spilled my coffee and said: ‘what the….?

I can truthfully say that a curious five-year-old playing a game of matching one map to another would have come quickly to the same conclusion. The match was not just visually compelling – it seemed to be exact.

For a week or so I sat on all of it and stewed.

I asked myself: “What if it doesn’t mean anything and is just another compelling correlation?”

But it kept bothering me and I got others at the Neighbourhood Change Research Partnership interested.

We wrote deck after deck of slides – we looked at the literature – we went back and forth. In honesty, we came to the conclusion that someone else must have done this work.

But we kept looking at it. We came to the conclusion that, since no one else had our custom data set, no one else could create the correlations we saw.

So we started covering our tracks. We went back to Statistics Canada and found minor anomalies in our descriptions and definitions. We corrected them.

Then the Toronto Star, which had long been interested in our preliminary work, asked again if we had anything. This time we did. We published it and it’s now ‘out there’.

We think it will be a genuine contribution to the Covid-19 recovery plan and to the literature.

What does this tell us about the conduct of inquiry?

What it tells me is that neither research nor science always moves from problem to evidence to solution.

Sometimes it moves from evidence to solution to problem.

For us, the Covid-19 problem was not on our agenda. We found a diagnosis and some solutions before anyone knew there was a problem.

I am tempted to think of Alexander Fleming’s discovery of penicillin: the pesky mold on the agar dish that kept killing his samples.

For me, it was a somewhat irrelevant data set that appeared to have neither a use nor a problem to solve.

Sometimes it works like that. A perplexing, seemingly useless data set casts important light on a problem for our age.

The important lessons are to use the point of convergence as points of light. Shout it aloud and see if anyone thinks it to be of value.

Spray the findings out so everyone can see them. Be unkempt. Reach out to strangers. Get your picture taken. And wait.

[1] Working-age adults (age 18-64 and not students) living in a multiple family household which is any mix of two or more couples without children, lone parents, or two-parent families; Working-age adults living with one or more adults as non-couple, non-parent roommates. They can be related such as two adult siblings or unrelated such as three friends living together or a mixture or related/unrelated adults.