Data journalism was becoming a thing about the time I started blogging a little over a decade ago. One format for that once-novel data-driven journalism was presenting a single powerful graphic accompanied by text explaining the graphic in depth. Jump ahead 10 years and we see data hasn’t sparked the interest some of us hoped. Instead of building a narrative to explain data, it’s far more common to invent numbers to explain a narrative. I will use a single slide to illustrate, from a recent presentation of the “Science Table: COVID-19 Advisory for Ontario.“
I tracked down this information after reading a ridiculous sentence in a report from the mainstream media:
“If strong measures stay in place, the daily total could drop to below 1,000 at the beginning of June and a high of 2,000 by mid-July.”
Why would cases start to rise after June begins “if strong measures stay in place.” The linked report includes an embedded video which shows only the graph section of the above slide, but it’s text of the slide that explains the bizarre forecast.
A nod to the power of credentials: note the graphic the last forecast from the same body was immediately demonstrably poor, with subsequent new daily positive COVID-19 cases falling below the best-case scenario, and yet questions such as this from John Michael McGrath of TVO persist:
is the current policy status quo compatible with the assumptions of the best case scenario?
are answered with a very confident:
The assumptions behind this forecast are what drives the curve. The model predicts a, “Stay-at-home order 6 weeks starting Apr 8,” but that length is a function of regulation and not only can be extended, in my opinion it’s unlikely that it won’t be in remaining hotspots given that it would terminate the lockdown order immediately before the May two-four long weekend.
The Table’s other prediction for the model is, “Vaccinating 100,000/day” – a level Ontario has been exceeding, on average, for over a week. Recently Moderna as well as Johnson and Johnson vaccines have arrived and as those are added to the current outlook the expectation should be for an average of approximately 140,000 doses arriving per day during May, and increasing beyond that in June.
Aside from those two stated baseless “predictions” – which are very poor assumptions to feed into a model – the slide notes 4 “best case assumptions” – which strike me as mostly lovely positions: “Effective sick pay, Short list of essential workplaces, Lower mobility, Continued focus on vaccinating high risk communities.” These obviously make some intuitive sense, but they’re presented unconvincingly for people expecting some numeric support in a model. The “R naught” is currently below 1, and both the vaccination rate and share of the population who’ve been infected are rising, so just maintaining the status quo would be enough to reduce cases even in the absence of warmer weather which allows for more time spent outdoors and opening windows when indoors.
One problem I see in wanting a “short list of essential workplaces” is proud people believe they’re essential – no matter how lousy their confidently delivered forecasts are – while humble people don’t realize they are.
It seems likely these projections of the “Science Advisory and Modelling Consensus Tables” will be even further above the actual outcomes than their previous report’s forecasts – which their own graphic shows to have been overly negative. There’s a good reason for all this: the hospitals in the greater Toronto area (GTA) are not only struggling to cope with the now waning influx of COVID patients, but pushing their sick population out to hospitals beyond their hotspot. Maybe a plan for where the big problem is would be a more manageable, and meaningful, endeavor.