Victor Chernozhukov writes:
Two months ago your blog featured Philip Lemoine’s critique “Lockdowns, econometrics and the art of putting lipstick on a pig” of our paper “Causal Impact of Masks, Policies, Behavior on Early Covid-19 Pandemic in the US” (ArXiv, June 2020, published in Journal of Econometrics). The paper found mitigating effects of masks and personal behavior and could not rule out significant mitigation effects of various “shut down” policies. Over the last two months, we studied Lemoine’s critique, and we prepared a detailed response.
Although Lemoine’s critique appears ideologically driven and overly emotional, some of the key points are excellent and worth addressing. In particular, the sensitivity of our estimation results for (i) including “masks in public spaces” and (ii) updating the data seems important critiques and, therefore, we decided to analyze the updated data ourselves.
After analyzing the updated data, we find evidence that reinforces the conclusions reached in the original study. In particular, focusing on the first three points to keep this note short:
(1) Lemoine showed that replacing “masks for employees” (business mask mandates) by “masks in public spaces” (public mask mandates) changes the effect estimate from negative to slightly positive. This critique is an obvious mistake because dropping the “masks for employees” variable introduces a confounder bias in estimating the effect of “masks in public spaces.” When we include both “masks for employees only” and “masks in public spaces” in the regression, the estimated coefficients of both variables are substantially negative in the original data. Lemoine’s argument seems to be an obvious but honest mistake.
(2) The second main point of Lemoine’s critique is non-robustness of results with respect to the data update. However, Lemoine has not validated the new data. We find that the timing of the first mask mandate for Hawaii (and another state) is mis-coded in the updated data. After correcting this data mistake, the estimated coefficients of “masks for employees only” and “masks in public spaces” continue to be substantially negative. This critique is also an honest (though not obvious) mistake.
(3) Lemoine analyzed the updated data that kept the original sample period from March 7 to June 3, 2020. The negative effects of masks on case growth continue to hold when we extend the endpoint of the sample period to July 1, August 1, or September 1 (before the start of school season). With the extended data, the estimated coefficients of “masks in public spaces” range from −0.097 to −0.124 with standard errors of 0.02 ∼ 0.03 in Tables 5-7, and are roughly twice as large as those of “masks for employees only.” A preprint version of our paper was available in ArXiv in late May of 2020 and was submitted for publication shortly after, which is why we did not analyze either the updated data or the extended data in our original paper.
Response to other points raised and supporting details on (1)-(3) are given in the ArXiv paper.
It’s great that outsiders such as Lemoine can contribute to the discussion, and I think it’s great that Chernozhukov et al. replied. I’ll leave the details to all of you.
https://www.worldometers.info/coronavirus/usa/hawaii/
Hawaii has mask mandates, vaccine mandates, travel mandates, and high vaccination rate.
Yet they just had the highest number of cases ever that stopped exactly when the temperature started cooling off.
so something that is constant within a state cannot explain within-state variation. fascinating insight.
Return to reality from statistical fantasyland and re-read the post.
If you do everything the doctor says and then get sicker than ever, maybe it is time for a new doctor.
Are the vaccinated in Hawaii getting sicker than ever, or are you aggregating at too large a scale and misrepresenting the implications?
https://www.khon2.com/coronavirus/covid-vaccine-breakthrough-cases-in-hawaii-spiked-since-july/
Oh interesting, 3% of hospitalizations are breakthrough cases, the vaccine is so ineffective!
That is hospitalizations, not cases. I’m still waiting for someone to account for different testing rates and reason for the test between vaccinated and not. Every paper mentions this in the discussion but no one is doing anything about it.
There were a few where regular testing was supposed to have been done amongst healthcare workers. But those papers are strange too since they aggregate data from last winter all the way into the summer and/or have odd exclusion policies. E.g., why would you ignore cases the first week after vaccination when that was reported to cause clinical levels of lymphocytopenia (immunosuppression) in about 25% of the pfizer subjects? Then a bunch of “real world” studies reported ~40% avg increase in infections over unvaccinated during that first week. That is going to “cull” the most highly exposed people from the vaccinated group.
The total cases is probably a decent lower bound if we assume there are another 2-4x cases in the transmission chain going untested.
“If you do everything the doctor says and then get sicker than ever, maybe it is time for a new doctor.”
I guess “sicker than ever” means preventing hospitalizations is irrelevant? Wouldn’t you say in a state where the vast majority are vaccinated the fact that 3% of hospitalized patients are vaccinated is strong evidence that the vaccinated are not “sicker than ever”? Well, you probably wouldn’t, but a person not using motivated reasoning would.
The masks are supposed to be about stopping cases. And even though it was known before they came out that the vaccines were not going to stop infection/transmission* they are still trying to implement these weird “vaccine passports” that only spread the virus more as happened in Hawaii.
* the lack of mucosal immunity was known since the 1960s for I.M. vaccines, and was seen in the animal trials so Pfizer/Moderna purposefully did not do any regular testing during the RCT
Studies in the UK, where the data are better and they do random testing and they haven’t discouraged testing among the vaccinated (as has happened here with the CDC) are likely more informative:
https://www.medrxiv.org/content/10.1101/2021.09.28.21264260v2.full-text
Anoneuoid –
> Yet they just had the highest number of cases ever that stopped exactly when the temperature started cooling off.
It seems to me that your view is that COVID case rates are lockstep in relationship with temperatures (even on relatively short time scales), as can be determined by looking at lines on our screens at Worldometers and reverse engineering – and pretty much irrespective of other factors that others think are also relevant at the individual state level.
So I”m wondering if you could explain some other state-wide trends I noticed looking at Worldometers. :
Why did cases increase dramatically from mid-July to early September in Oregon, only to then drop dramatically since then?
Why did cases increase pretty dramatically from late-June to mid-August in NY, increase at a much more gradual rate from mid-August to mid-September, and then actually decrease gradually from mid-September to mid-October (cases are basically at the same rate now as they were 9 weeks ago)?
Why did cases stay pretty flat in Arizona from mid August on (there was a slight increase from mid-August to mid-September)?
Why have cases been flat to decreasing for the last month in Idaho?
Why did cases peak in mid- to late-August in Cali, only to drop gradually and steadily over the last two months?
Or maybe I’m missing something about your argument?
I will note that Washington has a similar pattern as Oregon, even as Idaho does not.
This article talks about seasonal influences (on a much larger geographical scale) of temperature (plus humidity) from within a more sophisticated framework (that doesn’t just hand-wave away confounding variables like social factors) – and perhaps has some relevance to the factors that might help distinguish different patterns among states with similar temperature patterns. But I tend to doubt they’d argue their approach would apply for regional differences on a scale as small as at state-wide differences.
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021GH000413
It does seem pretty obvious to me that the more people are inside, breathing recirculated air, the more covid transmission you will have. However, there certainly are other drivers. For example if you hold the Sturgis motorcycle rally in mid 2020 you’ll set off a huge firestorm.
It would be nice to see an analysis of COVID at say census metropolitan area level together with weather. I do think it would tell us something useful.
Daniel –
Sure. It does seem compelling to me also that there’s a “seasonal” effect – that would likely be a combination of behavioral and environmental factors (temperature, humidity, maybe UV and air-drying capaxity, indoors versus outdoors, etc.).
But it doesn’t seem to me that behavioral differences would be a factor only in association with temperature but not through interactions with any number of potential mediators/moderators.
In particular, you would model this by changing the R0 seasonally. As an extreme example, if the entire population currently has mucosal immunity then it doesn’t matter if R0 rises from 0.5 to 100. You still won’t see a spike.
And average temperature is a weak proxy for breathing recirculated air. It would be better to use total energy devoted to HVAC, total hours below ~60 F or above ~80 F, or even just min and max temperatures of the month.
I’ll look again after the JHU weekly data updates later today, but the correlation of average monthly temperature with change in cases since the first week of of Sept is something like 0.7.
There was no huge firestorm in SD after the Sturgis rally. Why do people keep repeating that tired falsehood with no evidence?
Maybe because it is true.
https://www.forbes.com/sites/jemimamcevoy/2021/09/02/covid-surges-nearly-700-in-south-dakota-after-sturgis-motorcycle-rally-an-even-higher-rate-than-last-year/?sh=532b8ec174c6
“when the temperature started cooling off.”
The temp in Hawaii has a larger difference between night and day than from Summer to Winter. There is a reason most homes have no insulation, no heating and no AC.
“high vaccination rate”
Hawaii has a 59% vaccination rate, ranking 20th in US per CDC.
“vaccine mandates”
There is no state vaccine mandate. State and county workers do have testing mandates but most are 1x week and they can be waived if you are vaccinated. Religious or med waivers are freely given. Honolulu Gov. doesn’t allow testing option for it’s employees, vaccine only.
” travel mandates”
What limited travel mandates they had are long gone. Now you can show unverifiable “proof” of vaccine or go into 10 day unverified “quarantine.”
Pumping 20k people a day into two small islands may have contributed to the dramatic increase. The decline in cases also corresponded to the seasonal decline of trans-pacific travel.
> Hawaii has a 59% vaccination rate, ranking 20th in US per CDC.
And I’ll offer a wild guess vaccination rates are lowest and rate of covid cases are highest (not to mention lowest use of AC and highest use of ceiling fans circulating fresh air) in the same demographic: Native Hawaiians.
Pacific Islander (not Hawaiian), Filipino and Japanese bear the heaviest burden for fatal cases in Hawaii. Vaccine & race data is not really collected/reported accurately. I don’t think race really plays a role in AC use, but I could be wrong.
As for this…
> I don’t think race really plays a role in AC use, but I could be wrong
I would guess that NHPI are lower SES and relatedly more likely to live in older housing stock – which would be correlated with less reliance on AC (and more on ceiling fans)…
You really thik not?
Hawaii is not the mainland. You are clearly talking about something you know nothing about. AC use is driven by living at lower elevation. If you like hot dry weather you live in Kailua-Kona, if you like rainy cool weather you live in Hilo, if you like fireplaces you live in Volcano. Nothing to do with SES.
If you’re going to link to tv stats you might want to follow them to see what’s really there. The State DOH has race data for 49% of vaccines administered. Add this to what is actually counted as Native Hawaiian and you clearly can’t draw any conclusions. What data the State does have, hospitalizations and deaths, shows a clear delineation between Hawaiian and Pac Islanders. They are not the same group. I really expected a better class argument on this site. I will be ignoring you from this time on.
BHB –
Last part first:
> I really expected a better class argument on this site. I will be ignoring you from this time on.
It seems you’re no longer reading, but often times people say that on blogs when it really isn’t the case…in which case I would suggest that you not judge an entire blog based on one commenter that you feel is ignorant. But that’s up to you, obviously.
> Hawaii is not the mainland. You are clearly talking about something you know nothing about. AC use is driven by living at lower elevation.
Obviously the differences between the regions would be a factor. But I see no reason to see why that aspect would be mutually exclusive with SES and cultural differences. And you haven’t actually spelled out a reason why it would be (you’ve merely argued by assertion).
Are you saying that there’s no connection between SES and housing stock? Or between SES and people’s ability to afford AC given the high cost of electricity there? Or between SES and people’s ability to even install AC. Are you saying there’s no linkage between SES and likelihood of being a renter – in which case people are very much dependent on whether a landlord has installed AC.
> If you like hot dry weather you live in Kailua-Kona, if you like rainy cool weather you live in Hilo, if you like fireplaces you live in Volcano. Nothing to do with SES.
First, not everyone lives where they live simply because of what kinds of weather they like. Second, according to your logic no one would even use air conditioning. Those who like hot wether would live where it’s hot and not use it. Those who like cooler weather would live where it’s cool and have no need to use it.
Third, again – there doesn’t need to be mutual exclusivity. Among those who live in hotter areas, it only seem logical to me that people’s economic status would be a factor in ac use.
> If you’re going to link to tv stats
TV stats?
> you might want to follow them to see what’s really there. The State DOH has race data for 49% of vaccines administered.
I stated as much.
> Add this to what is actually counted as Native Hawaiian and you clearly can’t draw any conclusions. ?
Did you not see the article I linked that gave stats on lower vaccination rates among NHPI?
> What data the State does have, hospitalizations and deaths, shows a clear delineation between Hawaiian and Pac Islanders. They are not the same group.
Yes, I already acknowledge my mistake in not making a distinction.
https://statmodeling.stat.columbia.edu/2021/10/16/causal-impact-of-masks-policies-behavior-on-early-covid-19-pandemic-in-the-u-s-chernozhukov-et-al-respond-to-lemoines-critique/#comment-2026306
But on top of that, what meaning are you attaching to the delineation between NHs and PIs?
As compared to an overall 49% vaccination rate at the time of writing…
-snip-
The DOH also uses a race metric to keep track of the vaccination roll out. The Native Hawaiian Pacific Islander community was disproportionately hit by the virus, but only 18% of NHPI community are fully vaccinated.
https://www.khon2.com/coronavirus/vaccination-efforts-increasing-within-communities-with-lower-turnout/
Can someone explain (i) to me. For example, what’s the DAG that says it a confound that needs adjusting? I’m guessing it really is obvious, but it isn’t immediately so for me.
Does everyone understand that Hawaii is in the tropics, which doesn’t have seasons? And, temperatures vary more by elevation and location than time of year.
Not sure what you mean. This is just from a quick search but is probably decent for our purposes: https://weather-and-climate.com/average-monthly-Rainfall-Temperature-Sunshine-fahrenheit,Honolulu,Hawaii
Most people prefer a temperature of roughly 65-70F and otherwise will turn on the HVAC (if they have it available). Hawaiians will be spending more time breathing AC-cooled air during the summer than the winter according to that data.
The important thing is people breathing recycled air indoors, for which average outdoor temperature is a weak proxy.
Anoneuoid –
> Hawaiians will be spending more time breathing AC-cooled air during the summer than the winter according to that data.
This is just sloppy.
Have you ever been to Hawaii?
Do you have ANY idea if the prevalence of AC there.?
Keep in mind that with reduction a in tourism the occupancy of higher end places (those lowly to have AC) is probably significantly down.
One would think you might investigate that just a tad before attributing causality to a very complex problem to your assumptions about temperatures people prefer and the breathing of recycled air.
I’ve stayed in places in Hawaii that were more or less completely open air, let alone air conditioned.
Finally (then I’ll drop it), if drop in average state-wide temperature were the driving force in determining COVID rates, seems to me you’d expect to see a much larger drop in COVID rates from mid-August to mid-October in Arizona as compared to Hawaii.
Yet the drop-off was far greater in Hawaii.
I haven’t seen anyone argue that temperatures (or “seasonality”) isn’t a factor. But your comments about temperature as explanatory, without even remotely addressing other predictors, does come across to me as an agenda rather than a simple interest in pure science as you have described.
One last thing (because unlike Joshua I can’t drop it). If you look at a graph of covid cases for Honolulu County (Hawaii’s most populous), cases peak in on Sept. 2, 2021, and then begin to decline. The state imposes a vaccine passport plan on Sept 15 and imposes other restrictions. After that cases fell even more precipitously. The difference in average temperatures between Sept and Oct is about one degree and the average high temp is in the high 80s in both Sept and Oct, and if you understand the tropics, you know that the low temperature is at night (always) and the high at noon (always). The pattern of rise and fall has nothing to do with use of airconditioning which would not be affected by a 1 degree drop in temp. that left the high still in the high 80s. Rather Hawaii fits the pattern we have now seen everywhere. Restrictions are loosened. Cases go up. People realize there is an outbreak and return to their cautious behaviour. Government imposed restriction follow but lag, but some combination of self-imposed caution and gov regulations lead to a decline in the spread.
According to this, vaccinated people were allowed in without quarantine starting July 8th:
https://www.gohawaii.com/travel-requirements
Since I.M. vaccination does not give you mucosal immunity (maybe some small amount of neutralizing humoral antibodies extravasate into the mucosa for a month or so when concentration is very high though), it is easy to see why the lack of quarantine would introduce more cases to a mostly naive population. I don’t disagree with that part.
But then once cases started to plateau they introduced restrictions again, coincidentally with the peak temperature. It could be in Hawaii’s case the change in AC usage from Aug-Oct is not too important. However, their peak temperature is correlated with that in much of the rest of the US.
Yes, the restrictions follow the cases and usually cases are already decelerating when more restrictions are implemented. I am saying it looks like the cases follow the HVAC usage. Why else would there be such a nice correlation with temperature?
I think it’s great that they responded; this really raises my opinion of the authors of these kinds of studies .
Some quibbles on their points:
a) just adjusting Hawaii is questionable; if we only remove variables that are miscoded in one direction, this introduces bias.
I’m not sure what the best way to deal with this is. On one hand, no dataset is perfect. On the other hand, garbage in => garbage out and there is no way to correct for systematic bias. Perhaps computing some kind of “robustness against bad data points” measure would be the way to go. The fact that a misspecified Hawaii ruins their result suggests it is not very robust in this sense.
b) Their argument for smoothing is kind of weak. They write “Using unsmoothed data can only makes sense if we artificially try to lower the precision of the statistical inference”. Whatever is making the data rough is not being accounted for, which implies that if we just pretend the data was smooth to begin with then that artificially *raises* the precision of the statistical inference. Both “laws created” and “corona cases” correlate with week-days, can we rule out confounding effects being in play here and just assume the “real” data is smooth?
(c) If Lemoine says his placebo tests found spurious results, he should share his methodology. The placebo test the authors did seems kind of arbitrary, if Lemoine’s placebo tests are more arbitrary that suggest Lemoine is “p-hacking” his placebo tests, else I’d think the authors might be.
Anoneuoid said, “Most people prefer a temperature of roughly 65-70F and otherwise will turn on the HVAC (if they have it available). Hawaiians will be spending more time breathing AC-cooled air during the summer than the winter according to that data.”
The average temperture in Hawaii in the “winter” is 78%. Even on your graph you linked to, the temperture only falls below 70% during the January. And, that is just for Honololu.
Hawaii reopened for tourism, cases came back and then the state and local officials reimposed more restrictions. Using Hawaii to argue for the seasonality of Covid is comical. It is completely different than a location with a temperate climate.
I assume you mean degrees Fahrenheit instead of “%”, but perhaps we need better data. I looked at total electricity usage and see it typically peaks in Aug-Oct and drops ~30% from the peak by the winter:
https://www.eia.gov/opendata/embed/iframe.php?series_id=ELEC.SALES.HI-ALL.M;ELEC.SALES.HI-RES.M;ELEC.SALES.HI-COM.M;ELEC.SALES.HI-IND.M&analysis=none
It looks like this for me:
https://i.ibb.co/bPnTQst/hienergy.png
Do you know what that would be besides a drop in AC usage? I’ve acclimated to sleeping at 80F before even though my usual preferred sleeping temperature is more like 60F. So it wouldn’t surprise me if the comfortable range in Hawaii is 10-20F higher than the average.
Hawaii also fits in with the rest of the states: https://i.ibb.co/NZLNYpT/tempvscovid.png
If it was just Hawaii I wouldn’t pay it much attention.
Anoneuoid says “I looked at total electricity usage and see it typically peaks in Aug-Oct and drops ~30% from the peak by the winter.”
You realize that it is still October. The cases decline in Sept. while total electricity usuage was still on the increase.
I was being inclusive when I said Aug-Oct. The peak is typically in Aug and a couple spot-checks indicates it usually drops by 2-5% from Aug to Oct. The data for this year is not up there yet.
How many hours of breathing in recycled air does that correspond to, I don’t know. It seems too low to explain much of the drop.
Hawaii is in the middle of the Pacific, and heavily dependent on oil to generate electricity. It costs Hawaii alot more for oil than in the upper 49. (It has no oil fields and must ship everything in.) Electricity prices go up in August in Hawaii because crude oil prices go up in August because worldwide demand goes up in August and it has to pay a premium for shipping it in. The retail cost of things on an island in the middle of the Pacific has more to do with the cost than the demand, which you still have no evidence changed. Again, there is no reason to think the decline in cases that happened in Sept had anything to do with an up tick in the use of airconditioning. You are taking very noisy data and trying to tell a story with it, but there are so many intervening variables that it is a pointless task.
I should have said, “The retail cost of things on an island in the middle of the Pacific has more to do with the supply than the demand.”
The link shows electricity usage in millions of kWh. Are you saying people in Hawaii use more electricity in the summer because it is more expensive? I would expect the opposite.
You can also looks at the monthly cost of electricity in Hawaii, which has no obvious seasonality like the usage:
https://www.eia.gov/opendata/embed/iframe.php?series_id=ELEC.PRICE.HI-ALL.M;ELEC.PRICE.HI-RES.M;ELEC.PRICE.HI-COM.M;ELEC.PRICE.HI-IND.M&analysis=none
Sorry, but I’m not sure you are even looking at the evidence provided to you.
I really appreciate that they took the time to write a detailed reply and I wish more people reacted in the same way to criticism. That being said, I find their reply unconvincing. I don’t have time to write a full reply at the moment, because I’m busy with other stuff and will be for a while, but in case you are interested I just wrote a thread on Twitter where I made a few points in response to it. I didn’t try to be exhaustive, so please don’t assume that just because I didn’t mention one of the points they made in their response I agree with it, because that is not the case. I think most of their points were already addressed in my post, and sometimes they even ascribe to me claims I never made, so I just wanted to point that out with a few examples. Eventually, I’ll probably write a proper reply to their response, but like I said I don’t have time at the moment.
Phiippe –
From your Twitter thread:
> Another weird criticism they make is this point. I didn’t just *claim* to have done a placebo test, I *did* do a placebo test and I did easily find spurious effects. They make it sound like there can be some doubt on the matter even though I published my code!
I’m not entirely unsympathetic to you pointing out the potential bad faith connotation of “claim.” The use of “claim” instead of something like “argued” is a pet peeve of mine.
On the other hand, they did include the following:
> His blog explains his placebo test as follows:
Where they then quite you. They also included the following:
His placebo test is based on the simulated data based on his SIR model.1
Both of these statements would be consistent with them assuming that you did in fact conduct a placebo test.
But further, you may well have just misread what they were trying to convey:
Lemoine claims that he did a placebo test and easily found spurious effects.
As suggested by my emphasis, the use of “claim” may well have been referring to, well, your claim: That you easily found spurious results.
They may well have not intended to convey a connotation that you didn’t conduct the placebo test, but your contention that a placebo test would result in spurious results, “easily.” And even there it may well be that they don’t doubt that you “easily found” spurious results – but that your process in doing so was in error.
Ironically, your interpretation that they were contending you were acting in bad faith may be rooted in a bad faith reading on your part.
Even if I accept your interpretation, this still makes it sound like I didn’t easily find spurious effects, but I did and this is very easy to verify since I published my code and anyone can check that I didn’t do anything funny to find spurious effects that would otherwise have been difficult to find. There can be no doubt in the matter, the only interesting question is the relevance of my placebo test for their results, but if that’s what they disagree with me about there was no need to use the word “claim” here. Anyway, this point about their use of the word “claim” really wasn’t important, it’s just something I wanted to note in passing. So if you want to tie yourself in knots to interpret this passage in the most charitable way possible, which is funny given that you systematically go out of your way to interpret me in the most uncharitable way possible, be my guest. As for me, since they ascribed to me a ridiculous claim that only someone who doesn’t understand basic statistics would make and that I obviously didn’t make, I’m less inclined to charity.
Philippe –
> Even if I accept your interpretation,…
Accept or not accept as you wish. I’m just telling you what I think might be a plausible reading – and that I think you missed that plausible reading in a way you might not have had you been reading more charitably.
> this still makes it sound like I didn’t easily find spurious effects,..
Again, I addressed this in my comment above – which for some reason you haven’t ackniwlesged – but I think it’s more likely their syntax reveals that they don’t agree with your results, not that they question whether you did a test or that when you did so you found the results as you described.
> but I did and this is very easy to verify since I published my code and anyone can check that I didn’t do anything funny to find spurious effects…
Funny? Well, anyway, sure – it would have been more useful has they run your code and then described their results with analysis.
> There can be no doubt in the matter, the only interesting question is the relevance of my placebo test for their results, but if that’s what they disagree with me about there was no need to use the word “claim” here.
Again, I address that above. The word “claim” does have a negative connotation. But I don’t think it has to read as you have asserted. I think they’re questioning the validity of your findings rather than whether you conducted a test or whether you found the results you described when you did it..
> Anyway, this point about their use of the word “claim” really wasn’t important,…
Agreed. So maybe you shouldn’t have written about it?
> So, if you want to tie yourself in knots to interpret this passage in the most charitable way possible…
Tie myself in knots? You assumed a particular interpretation of somewhat ambiguous syntax. I’m telling you what I tbi k is a more plausible reading. Not sure why that me tying myself into knots.
> which is funny given that you systematically go out of your way to interpret me in the most uncharitable way possible,..
Tu quoque?
I’m not sure why quoting you, and questioning you when you say that you didn’t say what I quoted you as saying, is uncharitable. I even explicitly acknowledged that maybe there might be some aspect that I wasn’t understanding – but rather than continue to engage you deemed me as a waste of your time.
> As for me, since they ascribed to me a ridiculous claim that only someone who doesn’t understand basic statistics would make and that I obviously didn’t make, I’m less inclined to charity.
Well, it’s good that here you acknowledge that you’re not inclined to read what they say charitably. The corollary of that is that you’re likely to misintrept their meaning when it’s ambiguous. That was my point. I think it’s better to extend cognitive empathy and work hard at understanding what might be someone’s intended meaning.
As for your view that they’ve ascribed to you a rudementary error – this isn’t really about you. But anyone can make make even rudementary mistakes. Sometimes not in a technical sense but more in the sense of fit for purpose. You, in fact, have ascribed to people with clear expertise and skill, rudementary mistakes.
Again, I suggest a more charitable reading might increase the value of the exchange.
Joshua:
Without discussing any of the details of this particular dispute-within-a-dispute, I wanted to pick up on this statement of yours:
I know what you mean, but sometimes I wonder. A few things can be going on here. First, people can abuse the norm of charity by straight-up lying or carefully avoiding uncomfortable truths. We’ve seen this with various perpetrators of junk science who we’ve discussed in this space over the years. Second, recall Clarke’s Law that sufficiently incompetent research is indistinguishable from fraud. Third, sometimes there’s no real charitable interpretation, and when attempts to be charitable can be read as condescension. For example, what’s the charitable interpretation of that Chicago economist dude who misread my writings on election forecasts, attributing to me the opposite interpretation to what I actually said? My most charitable interpretation here is that the guy was busy, he was writing a partisan hit piece for an extremist website and he didn’t bother to fact check his own article because it wasn’t really a high priority—maybe he was writing it as a favor for one of his political friends, I have no idea.
My point here is not that we “need to fight fire with fire” but rather that sometimes a charitable interpretation is hard to come by, and in those cases maybe it’s better to be more open and just express our annoyance, in a way that’s open to correction if necessary. In the above example, Philippe could not come up with a charitable interpretation and he said so, then you were able to explain to him that such an interpretation was possible. This could be a better way to progress than for Philippe to have tried to impute in the first place a charitable explanation that he did not have at hand.
At this point, you might ask why don’t I correct the Chicago guy directly and ask what he meant when he was writing that stuff about me—or you could step back and ask him why he didn’t contact me directly in the first place, which would’ve spared him all sorts of trouble? In my case, I’ll just say that I’ve generally found such interactions unpleasant: the kind of people who are willing to publish such ridiculous statements don’t usually seem to be the kind of people who respond well to criticism. In his case, I don’t know; my best guess is that he wanted to write something partisan and incendiary, and checking with me would’ve carried the risk that he would’ve felt the obligation to moderate his piece. I’m guessing he’s more careful in his scholarly work.
Andrew –
Thanks for taking this up.
> First, people can abuse the norm of charity by straight-up lying or carefully avoiding uncomfortable truths. We’ve seen this with various perpetrators of junk science who we’ve discussed in this space over the years. Second, recall Clarke’s Law that sufficiently incompetent research is indistinguishable from fraud. Third, sometimes there’s no real charitable interpretation, and when attempts to be charitable can be read as condescension. For example, what’s the charitable interpretation of that Chicago economist dude who misread my writings on election forecasts, attributing to me the opposite interpretation to what I actually said? My most charitable interpretation here is that the guy was busy, he was writing a partisan hit piece for an extremist website and he didn’t bother to fact check his own article because it wasn’t really a high priority—maybe he was writing it as a favor for one of his political friends, I have no idea.
Sure. You don’t have to conclude that a more charitable reading is the more plausible interpretation.
For me this goes back to constructing a good argument. If you don’t charitably interrogate the “naysayer” or opposing views, you can’t really formulate solid rebuttals to them. All you really end up doing is trading off arguments against strawmen. In the end, you may be right in your argument either way. But as an observer, it’s easier for me to gain a foothold on an exchange if all of the participants argue against what their interlocutor actually said, or clarifies what they’ve said before making assumptions.
Maybe “charitable reading” isn’t the right term here, but it’s kind of a vernacular for what I”m talking about. What I really mean is that if you don’t commit to understanding what someone plausibly intended to be arguing, the value of a response would be limited
A popular framing that’s kind of similar is the “scout vs. solider” mentality (see Julia Galef) when engaging in a process of exchanging view. It isn’t so much that there’s never any value in a soldier mentality. But I think if you don’t have the information a scout might gather, your work as a soldier is likely diminished.
> My point here is not that we “need to fight fire with fire” but rather that sometimes a charitable interpretation is hard to come by, and in those cases maybe it’s better to be more open and just express our annoyance, in a way that’s open to correction if necessary. In the above example, Philippe could not come up with a charitable interpretation and he said so, then you were able to explain to him that such an interpretation was possible. This could be a better way to progress than for Philippe to have tried to impute in the first place a charitable explanation that he did not have at hand.
I don’t entirely disagree. 3rd parties can be a useful part of this process. But I also think that seeing those hard to see interpretations is, to some degree, a matter of commitment and a developed skill. It’s a part of being willing to interrogate ourselves for our own biases. I think there’s (unfortunately) a hard limit on our ability to get past our own biases and motivated reasoning (I think much of the relevant literature suggests it’s just not possible) – but I still think there’s a skill than can be developed there and I certainly think that an openness to the process can be cultivated.
> At this point, you might ask why don’t I correct the Chicago guy directly and ask what he meant when he was writing that stuff about me—or you could step back and ask him why he didn’t contact me directly in the first place, which would’ve spared him all sorts of trouble? In my case, I’ll just say that I’ve generally found such interactions unpleasant: the kind of people who are willing to publish such ridiculous statements don’t usually seem to be the kind of people who respond well to criticism.
I very much agree with that. But I do also think that there’s a skill to be developed in how one can approach these encounters, so I don’t see that process as a static state, or one that’s hopeless.
> In his case, I don’t know; my best guess is that he wanted to write something partisan and incendiary, and checking with me would’ve carried the risk that he would’ve felt the obligation to moderate his piece.
That was my interpretation. And part of the reason why I interpreted him that way was every much because he displayed a lack of “charity” in interpreting your work. Again, as an observer, the ability for someone to explore the full range of plausible readings is a key indicator for me when I”m trying to evaluate their arguments. Sometimes, in highly technical discussions, since I can’t judge the arguments on their technical merits, that’s pretty much what I have to go on.
> I’m guessing he’s more careful in his scholarly work.
Well, that IS a charitable reading. Yes, it’s entirely possible that a propensity for a shallow analysis only happens in areas where he’s particularly ideologically triggered. In my experience, however, there tends to be a general pattern of association.
Joshua:
It’s always good to see Julia mentioned, since she was my student many years ago.
I think the charitable/uncharitable thing is a different dimension than scout/soldier. When I mock Gremlin Man or Pizzagate or the Disgraced Primatologist or whatever, I’m still acting as a “scout”—I’m trying to understand. I’m not trying to win a battle. This sometimes comes up in comments: someone will say that my criticisms will be more effective if I don’t do X, or that argument X weakens my case and I should stick with the clearer cases of Y and Z. And I’ll reply that I’m not trying to make a case! One reason I do the derisive nicknames is to be more entertaining, which I think is a value in itself.
But I’m not saying that what I do is the right approach for others. Heck, I don’t even claim is the right approach for me. It’s just what I’ve been doing. To get back to Philippe: again, without taking any position one way or another on his statements, it could well be that he’s currently playing an effective role in the ecosystem by writing with a hair trigger. And it might also be that he’ll read this comment thread and, become more charitable in his readings, and be even more effective in the future! How’s that for empty happy talk?
Maybe a shorter response that’s more decipherable…when I say this:
>. I think it’s better to extend cognitive empathy and work hard at understanding what might be someone’s intended meaning.
The point being that you don’t have to agree with their view, or even think it’s viable or moral or ethical or even of value. But I think generally you’re better off if you work at understanding an issue from their perspective. Maybe that’s where the term “charitable” is misleading. And the “empathy” part of “cognitive empathy” is also maybe misleading.
Maybe the “scout versus solider” framing is less misleading in that sense. I’m not sure it’s directly parallel but it’s probably close enough for Jazz.
BHB –
I was careless with my terms. I was thinking of those of Pacific Island heritage. A quick Google seems to confirm my INUITION, as it teaches me new term: Native Hawaii Pacific Islander (NHPI).
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7226312/
… intuition…
INUITION is something that teaches you how to refer to indigenous peoples at higher latitudes.
Yes, you are right. The link is to millions of kWh. That was sloppy. But, I did look at the link, and it still seems obvious that the decline in covid cases which begins around Sept 2 and declines precipitously cannot be linked to changes in temperature, because that doesn’t decline below the mid 80s until December. The low temps this week will still be above 72 degrees. The decline is totally driven by more cautious behaviours and public health interventions, which are taken more seriously in Hawaii and are much more effective because the numbers are lower and actual quarantines and case tracking can be imposed.
comment was meant as a reply to Anoneuoid
Perhaps, I don’t really see a ~5-10% reduction in AC usage (assuming about half the electricity usage in Hawaii is devoted to AC) accounting for such a huge drop in cases. Maybe it was larger this year, which we won’t know until next year since there is a 3 month delay in that data.
The only reason I think there is an AC effect there is that the same drop was seen along the entire gulf coast, including in Florida and Texas where interventions are not taken seriously.
A far simpler critique of the original paper is the absurdity of claiming these methods can produce evidence of a causal effect in the first place. Getting lost in the weeds of Hawaii’s climate misses the forest for the (palm) trees.