good article, I saw from an online tracker thingie that I was in the first 5% in line in Oregon to get the vaccine, because I am over 65 and have diabetes, obesity, etc. I thought no, not fair. I don't have to work and have total control over my risk level. I don't even have to leave my house at all if I don't want to, and really haven't for over six months now. Yet all these young grocery clerks, teachers, people who are forced by their jobs to be around of a lot of people are way behind me. It makes no sense to me.
So I said that on twitter and got attacked for being arrogant and who am I to speak for old people?
So, ok. I guess at this point I am just going to "just trust the experts". Too much fighting over Covid, let's all just try to get along. I am just going to go when they tell me to and that's that.
“ With this simple race-neutral framework, with no individual moral judgement, we would still take race into account when allocating the vaccine.”
Why is that necessarily true? Wouldn’t we want to measure and regress on variables that have a potentially causal relationship to death?
It seems unlikely that being Black is causal in a biological sense to death from COVID. You note some of the confounding variables in this article. Wouldn’t we want to measure things like comorbidities, distance to healthcare facilities, etc that would have the explanatory power?
Genuinely curious! Understand it’s a larger social science question.
what will happen if they went ahead with a race based algorithm is that people would start saying it was cultural, that Black people were doing more house parties and more risky behavior.
Yep. That would be the "moral judgement" I talked about. Of course, in this case the moral judgement would be pretty made-up, ad-hoc, and based on racial politics.
right, having read all this, I think the only fair way is to base it on risk, but if they can somehow include self imposed risk, or better put perhaps unavoidable risk, then those like myself who don't have to engage in risky behavior to meet basic needs go further back.
Except tonight in the local paper it says we got vaccines in for health care workers here and 40% of them don't want to take them. So I figured, ok f that I am calling in to see if I can get one.
Even if we don't understand the causality, all that matters here is that the correlation is stable with respect to the policy -- in other words, that allocating vaccines a certain way doesn't *change* the relative mortality rates. And it seems like the risk of that is fairly low here...
Identifying the actual underlying explanatory variables - e.g. certain medical conditions - and using them instead of a proxy category that may correlate with them - e.g. "Black" - will necessarily lead to a more efficient allocation of the vaccines. This is simply because you will be able to identify the most at risk individuals across the whole population, not just a certain demographic. An exception would be if the factors causing Black people to be more at risk are not understood or not observed, but that case is not being made here I don't think.
Health care workers don't just a risk of infection they also have a risk of spreading to the most at-risk people, i.e. the sick and hospitalised. Is that part of the calculation? This is, eg, part of the rationale of the health service I work for giving free flu shots to workers.
This is an outstanding article on the use of statistics in making real world public policy decisions. Noah properly points out the public trust issue that surrounds all public policy decision making in this country just now. Trust for decision makers is extremely low. And, I do not think it matters from which side of the political spectrum public decision making is viewed.
I loved statistics when I got my MPA. I particularly liked regressions and time series analysis.
But, multivariate regression analysis involving a multitude of variables is anything but a transparent process. The actual calculation is not something you would want to print line for line and hand out to voters or for that matter put up on a blackboard for all to see and consider. Although, that would be amusing to see as a part of a debate. Additionally, what to consider in the algorithum as pointed out is fraught with "social justice" types of judgment calls before one can even hit the return key to begin the calculation. Add to that the distinct possibility that statisticians may even be less trusted than politicians when it comes right down to a final reckoning by the electorate. (Lies in order of their perfidity are: 1) lies, 2) damn lies, and 3) statistics.) And, we don't even get to vote for our statisticians.
This is probably why a keep it simple stupid approach is used by policy makers who want to be reelected, especially at a congressional and local levels. Unlike private companies who tend to be judged on a profit and loss statement by stockholders, politicians are judged by their individual electorates, a very large part of whom only have a desire to look out for what they think are their own best interests, as influenced by activists and media groups of various sorts, the experts be damned. The sense of many is that if you took all of the various experts on a subject and laid them on the ground, head to foot, in a straight line, they would still point in different directions.
My experience suggests to me that when public decision making gets too complicated, few will believe in it and trust will erode further.
You are welcome. I always enjoy your articles. Your race neutral, only lives matter argument was very interesting indeed, although I am not at all sure that it will fly in the present political environment. More's the shame,
What's your take on Congress and gov't workers getting it before everyone else? That really pissed me off. I'm 5 minutes from Pfizer and don't have it yet and some of these MAGA people are getting it.
The big argument for vaccinating health care workers first is that it will cut the chances of a complete health care system breakdown in the face of rising overall risk rates. Even now, hospitals have been prioritizing workers, clinical and non-clinical, who have contact with COVID patients or patients with unknown COVID status. Some doctors are getting the vaccine up front. Others, like a friend of mine, will wait a bit.
I forwarded this email to a friend, with the note "This guy once wrote a long essay about the advantages rabbits have over cats as house pets, thus earning him a permanent spot on the list of people I disagree with but read carefully."
Great summary of the divide between the quantitative and political arguments of using risk prediction in health policy. I have had to think about this issue a lot in the past as a health economist working on risk-stratified screening and disease prevention. The politically explosive element in the case was the idea that racial equity should be considered on top of absolute benefit. But I think you are correct to think that even ‘race-neutral’ policy that uses race as a predictor of risk would be fuel on the political fire.
I think about the issue of public trust slightly differently. Trust is what is required for people to accept any kind of statistical decision making. Even when the process is well understood such as a simple age threshold there can be problems. An obvious question would be “why was race not considered if we know it is an important risk factor?”. Do the policy makers not care about 55 year old Black lives as much as 65 year old White lives (assuming they have the same risk)? The public have to trust that a simple model makes the most sense and actually strikes a reasonable balance, rather than just being convenient to particular interests. That is why it is fundamental for the decision making process to be trustworthy, i.e. transparent and clearly independent from politics. The biggest failing of the ACIP recommendations was to fail to be transparent about how they got to the recommendation to prioritise essential workers before the over 65s from the evidence they presented. They highlighted their assessment that the ‘ethics’ component was favourable for essential workers but failed to explicitly state or justify their judgement that this outweighed the greater number of deaths averted for over 65s. This played into current suspicions of everything being linked to politics and undermined belief in their independence.
One last thought, if they had simply proposed a risk equation with age, race and gender that minimises deaths then this might have played out very differently. I don’t know whether or not that would be politically acceptable right now given the issues you have highlighted. But I do think that the process would at least seem more trustworthy and might build trust in public health policy making. People might not understand multiple regression but they can understand “we believe they are at higher risk” and “we think this saves the most lives”.
The multivariate analysis does not strike me as too much of a risk for spreading distrust as long as the factor being optimized is transparent and well supported among the public. The biggest limiting factor is probably the ability of the system to actually implement anything too complex.
There are, of course, proxies for race, such as geography like highest community spread that could be used, and would likely be less inflammatory.
good article, I saw from an online tracker thingie that I was in the first 5% in line in Oregon to get the vaccine, because I am over 65 and have diabetes, obesity, etc. I thought no, not fair. I don't have to work and have total control over my risk level. I don't even have to leave my house at all if I don't want to, and really haven't for over six months now. Yet all these young grocery clerks, teachers, people who are forced by their jobs to be around of a lot of people are way behind me. It makes no sense to me.
So I said that on twitter and got attacked for being arrogant and who am I to speak for old people?
So, ok. I guess at this point I am just going to "just trust the experts". Too much fighting over Covid, let's all just try to get along. I am just going to go when they tell me to and that's that.
Trust the experts!? What kind of a crazy fool strategy is that! :)
lol
“ With this simple race-neutral framework, with no individual moral judgement, we would still take race into account when allocating the vaccine.”
Why is that necessarily true? Wouldn’t we want to measure and regress on variables that have a potentially causal relationship to death?
It seems unlikely that being Black is causal in a biological sense to death from COVID. You note some of the confounding variables in this article. Wouldn’t we want to measure things like comorbidities, distance to healthcare facilities, etc that would have the explanatory power?
Genuinely curious! Understand it’s a larger social science question.
what will happen if they went ahead with a race based algorithm is that people would start saying it was cultural, that Black people were doing more house parties and more risky behavior.
Yep. That would be the "moral judgement" I talked about. Of course, in this case the moral judgement would be pretty made-up, ad-hoc, and based on racial politics.
right, having read all this, I think the only fair way is to base it on risk, but if they can somehow include self imposed risk, or better put perhaps unavoidable risk, then those like myself who don't have to engage in risky behavior to meet basic needs go further back.
Except tonight in the local paper it says we got vaccines in for health care workers here and 40% of them don't want to take them. So I figured, ok f that I am calling in to see if I can get one.
Even if we don't understand the causality, all that matters here is that the correlation is stable with respect to the policy -- in other words, that allocating vaccines a certain way doesn't *change* the relative mortality rates. And it seems like the risk of that is fairly low here...
Identifying the actual underlying explanatory variables - e.g. certain medical conditions - and using them instead of a proxy category that may correlate with them - e.g. "Black" - will necessarily lead to a more efficient allocation of the vaccines. This is simply because you will be able to identify the most at risk individuals across the whole population, not just a certain demographic. An exception would be if the factors causing Black people to be more at risk are not understood or not observed, but that case is not being made here I don't think.
Thanks for the reply!
Health care workers don't just a risk of infection they also have a risk of spreading to the most at-risk people, i.e. the sick and hospitalised. Is that part of the calculation? This is, eg, part of the rationale of the health service I work for giving free flu shots to workers.
Absolutely. No one disagrees that we should vaccinate health care workers first, for this reason and others.
This is an outstanding article on the use of statistics in making real world public policy decisions. Noah properly points out the public trust issue that surrounds all public policy decision making in this country just now. Trust for decision makers is extremely low. And, I do not think it matters from which side of the political spectrum public decision making is viewed.
I loved statistics when I got my MPA. I particularly liked regressions and time series analysis.
But, multivariate regression analysis involving a multitude of variables is anything but a transparent process. The actual calculation is not something you would want to print line for line and hand out to voters or for that matter put up on a blackboard for all to see and consider. Although, that would be amusing to see as a part of a debate. Additionally, what to consider in the algorithum as pointed out is fraught with "social justice" types of judgment calls before one can even hit the return key to begin the calculation. Add to that the distinct possibility that statisticians may even be less trusted than politicians when it comes right down to a final reckoning by the electorate. (Lies in order of their perfidity are: 1) lies, 2) damn lies, and 3) statistics.) And, we don't even get to vote for our statisticians.
This is probably why a keep it simple stupid approach is used by policy makers who want to be reelected, especially at a congressional and local levels. Unlike private companies who tend to be judged on a profit and loss statement by stockholders, politicians are judged by their individual electorates, a very large part of whom only have a desire to look out for what they think are their own best interests, as influenced by activists and media groups of various sorts, the experts be damned. The sense of many is that if you took all of the various experts on a subject and laid them on the ground, head to foot, in a straight line, they would still point in different directions.
My experience suggests to me that when public decision making gets too complicated, few will believe in it and trust will erode further.
Thanks, John!
You are welcome. I always enjoy your articles. Your race neutral, only lives matter argument was very interesting indeed, although I am not at all sure that it will fly in the present political environment. More's the shame,
reasonable.
What's your take on Congress and gov't workers getting it before everyone else? That really pissed me off. I'm 5 minutes from Pfizer and don't have it yet and some of these MAGA people are getting it.
9 months of telling us not to be afraid of it, and they cut in line for the vaccine.
The big argument for vaccinating health care workers first is that it will cut the chances of a complete health care system breakdown in the face of rising overall risk rates. Even now, hospitals have been prioritizing workers, clinical and non-clinical, who have contact with COVID patients or patients with unknown COVID status. Some doctors are getting the vaccine up front. Others, like a friend of mine, will wait a bit.
I forwarded this email to a friend, with the note "This guy once wrote a long essay about the advantages rabbits have over cats as house pets, thus earning him a permanent spot on the list of people I disagree with but read carefully."
Great summary of the divide between the quantitative and political arguments of using risk prediction in health policy. I have had to think about this issue a lot in the past as a health economist working on risk-stratified screening and disease prevention. The politically explosive element in the case was the idea that racial equity should be considered on top of absolute benefit. But I think you are correct to think that even ‘race-neutral’ policy that uses race as a predictor of risk would be fuel on the political fire.
I think about the issue of public trust slightly differently. Trust is what is required for people to accept any kind of statistical decision making. Even when the process is well understood such as a simple age threshold there can be problems. An obvious question would be “why was race not considered if we know it is an important risk factor?”. Do the policy makers not care about 55 year old Black lives as much as 65 year old White lives (assuming they have the same risk)? The public have to trust that a simple model makes the most sense and actually strikes a reasonable balance, rather than just being convenient to particular interests. That is why it is fundamental for the decision making process to be trustworthy, i.e. transparent and clearly independent from politics. The biggest failing of the ACIP recommendations was to fail to be transparent about how they got to the recommendation to prioritise essential workers before the over 65s from the evidence they presented. They highlighted their assessment that the ‘ethics’ component was favourable for essential workers but failed to explicitly state or justify their judgement that this outweighed the greater number of deaths averted for over 65s. This played into current suspicions of everything being linked to politics and undermined belief in their independence.
One last thought, if they had simply proposed a risk equation with age, race and gender that minimises deaths then this might have played out very differently. I don’t know whether or not that would be politically acceptable right now given the issues you have highlighted. But I do think that the process would at least seem more trustworthy and might build trust in public health policy making. People might not understand multiple regression but they can understand “we believe they are at higher risk” and “we think this saves the most lives”.
The multivariate analysis does not strike me as too much of a risk for spreading distrust as long as the factor being optimized is transparent and well supported among the public. The biggest limiting factor is probably the ability of the system to actually implement anything too complex.
There are, of course, proxies for race, such as geography like highest community spread that could be used, and would likely be less inflammatory.