An intriguing chart has been circulating the Twittersphere in recent days, after being tweeted by King’s College economics professor Jonathan Portes and retweeted by the likes of Labour peer Andrew Adonis.
The chart, which Portes tweeted under the question “What should we reopen first?”, comes from a paper published on Saturday. It’s written by Seth Benzell, a postdoctoral associate at the MIT Initiative on the Digital Economy, and two others from MIT’s Sloan School of Management.
The chart attempts to categorise 30 different US businesses and amenities to figure out how “important” and how “dangerous” each one is to reopen:
The authors of the paper say they are providing an “empirical contribution regarding which types of locations pose the best and worst risk-reward trade-offs”, and that they “hope this analysis will help policymakers decide how to reopen their economies”.
But we don’t think it will. Or at least, we don’t think it should.
There are quite a few problems with their rankings, as you might have already thought as you peered at the chart. How is that casinos are deemed more important than office-supply stores? Fast-food joints more important than childcare? Nightclubs less dangerous than parks?!
Ranking “danger”
It seems Benzell et al. have a funny way of determining the “danger” of a given location. They explain the data they used for that calculation comes from Safegraph, a firm that, according to the paper, “tracks the movement patterns of tens of millions of Americans at the monthly level”. The way the danger level is then calculated is as follows:
We measure the amount of physical contact (and danger) of a location through four main measures: number of visits, number of unique visitors, and person-hours of visits above two density thresholds. These last measures are inspired by the CDC’s ‘six-foot’ social distancing rule and the rule of 1 customer per 20 square metres implemented in parts of Germany.
But there are several things they fail to take into account.
First, although the average time spent in, say, a nightclub, is relatively low given that it is only open for maybe eight hours at a time and often only on a few days a week, during the time that it is open, people are very closely packed together.
So yes, even though theoretically a sneeze can travel up to 8 metres and cough droplets up to 2, and so working out how many hours more than one person is in a six-foot or 20-square-metre area is of some use, it is surely far easier for mucous and saliva to be spread when people are packed together liked tinned pilchards.
Therefore it might be more useful to work out which places are likely to encourage such packed-togetherness. Nightclubs are one of them; amusement parks — though they have vast spaces with very few people in them — are another (just think of those hours-long queues).
Second, safety is not just about proximity; it’s also about our likelihood of keeping our hands and, um, bodily fluids off one another. Not only does the human density in a club or a bar make touching almost inevitable; it intrinsic to these locations that people do. And once you add alcohol or drugs to the mix, well, we know what humans are like.
(Furthermore, after a visit to one of these establishments, it is not uncommon to end up waking the next day in a different home to that which one lives in, meaning a big increase, effectively, in “person hours”, and bringing risk to the rest of that household. This effect seems to us particularly likely when such businesses are reopened after a long spell of quarantine. Ahem.)
We also need to think of such things when considering when to start allowing “mass gatherings” again. Imperial College might have said that “stopping mass gatherings is predicted to have relatively little impact because the contact-time at such events is relatively small compared to the time spent at home”. Nonetheless there are still banned in many countries. And when we allow them again, we should probably not treat them as homogeneous.
Classical concerts, cinema visits and the theatre, at which you tend to not have much contact with others, are very different in terms of human contact from, say, football matches:
Anyone who has celebrated a goal knows it’s pretty hard to resist the urge to hug those nearby. Similarly rock concerts, which often feature moshing, drinking and drug-taking, should probably be higher up the “danger” index than, say, the Royal Opera House (though one could also make the counterargument that, demographically speaking, rock concerts are safer).
Third, it’s not just about proximity. Although it is believed that coronavirus is mainly spread through the air, we also know that it can survive on certain surfaces, like plastic and metal, for days. Gyms have many surfaces that are touched and spluttered over by a lot of people throughout the day, and yet the academics have deemed them only a little more dangerous than parks, which have virtually none.
Ranking “importance”
We’re also not keen on the methodology the authors use to determine “importance”. They do this by using three measures of “economic importance” — annual payroll, receipts, and employment --- and coming up with an average. They then use one “consumer importance” measure, and give that an equal weighting to the average of the other three.
For the consumer importance measure, the academics use a “nationally representative survey of 1,099 US residents” conducted between April 13 and 15, in which participants were asked “which location, among two options, they would prefer to be open”.
But this doesn’t take into account the acuteness of the preference. Indeed for some people it’s not just about which of two locations they might prefer; it’s about which locations they feel they need. A place of worship, for example, is somewhere many feel a duty to attend, and yet these are deemed less important than shoe shops.
And although the data used in calculating “economic importance” makes sense on some level, what’s important to our economic is also the vital infrastructure that allows us to meet our basic needs — something these categories do not account for.
Childcare, for example, is necessary not just at an individual level but also at a societal and economic one. For instance, it enables key workers to keep doing their jobs. The fact restaurants are listed as economically more important than childcare is therefore absurd.
To be fair to the authors of the report, they do acknowledge some of the limitations of their methodology. As well as pointing out some of the issues we outline, they also highlight their safety calculations do not take into account heterogeneity in the types of visitors, nor the fact that if one type of business is closed, people might substitute that for another one. But we feel that when the limitations are so great, an analysis often stops being useful.
It is very important that we have a range of multidisciplinary analysis that looks into the complex issues we are grappling with, and that we don’t rely too much on a small group of epidemiologists. But this doesn’t seem like it adds that much to the discussion; a better ranking could have been done without having access to any of this data.
Instead, it looks to us like a classic case of the old computer science problem: garbage in, garbage out (GIGO). If the data going into the model is not up to scratch, the conclusions won’t be either (this is one of our perpetual complaints with the idea that blockchain can make things better, and indeed is an issue one needs to be particularly aware of when examining the various bits of modelling currently being done around Covid-19).
It might be useful to have a ranking system in order to determine which public places should be opened up first. But this ain’t quite it.
"like this" - Google News
April 22, 2020 at 11:04AM
https://ift.tt/2VsPNjO
We need to figure out how we will reopen the economy. But not like this. - Financial Times
"like this" - Google News
https://ift.tt/2MWhj4t
Shoes Man Tutorial
Pos News Update
Meme Update
Korean Entertainment News
Japan News Update
No comments:
Post a Comment