One of the (many) disillusioning aspects of the Covid response of the New Zealand government (politicians and public service) has been the apparent total absence of any use of cost-benefit analysis techniques to help inform thinking about policy responses. No cost-benefit analysis on any aspect of the policy response has ever been published (or hinted at), on the couple of occasions I’ve OIAed any such analysis (just to be sure) agencies have been quick to deny any such analysis exists, and when one independent agency (the Productivity Commission) did do a little exercise along these lines at one point last year it was shunned as almost “unclean”. And if there had been any slight excuse early last year about “no time” – not convincing even then – officials have had 22 months now to get toolkits in place. But they (and their political masters) seem to prefer seat-of-the-pants thinking, all with minimal transparency. (On that latter note, it is now three months since the current lockdowns began and not one piece of official advice, not one Cabinet paper has yet been released, despite the enormous economic and social costs of the choices the government has made.)
The point about cost-benefit analysis is not that using those techniques, or that way of thinking about the issue, will generate “the” right answer. On many of these things there is no “the” right answer. The merit lies in a combination of (a) forcing people to write down their assumptions, including which variables (even hard to estimate ones) should be relevant to a particular decision, and (b) then enabling users to get a sense of how much difference a different set of assumptions might make to the bottom line. Using the techniques facilitates disciplined thinking and transparency, the latter itself supporting scrutiny (especially important when such costly and often intrusive/restrictive decisions are involved.
Consultant economist and former Victoria University academic Martin Lally has been one of those attempting to apply a cost-benefit approach to thinking about Covid policy responses. I wrote here about one of his pieces from last year. I haven’t always agreed with his conclusions, but (as above) that isn’t the point. The value in such exercises is to prompt people to think harder about which assumptions they might disagree with, why, whether all the right variables are being taken into account, and what differences different assumptions might make.
On Saturday Lally sent out a short piece he had done, attempting to sketch how one might apply a cost-benefit type of approach to thinking about vaccine coercion (or – ugly Americanism – “mandates”). He has given me permission to reproduce it here, which I will then follow with my own thinking. It is a sketch outline, towards the sort of fuller cost-benefit analysis one might hope government agencies – with access to much more resource – would routinely be doing and revising.
Lally notes that he – like me – is fully vaccinated “without coercion”.
Vaccine mandates for the general population are proving to be extremely controversial. Opponents point to the right to choose. Proponents point to the costs that the unvaccinated impose upon the vaccinated, in particular the increased risk of death to some vaccinated people (because the vaccine is not perfect and the more so as the time since the vaccination increases) and the increased load on the health system from unvaccinated people leading to some (vaccinated) people receiving an inferior level of care for non-covid conditions than they otherwise would.
This is yet another example of the trade-offs we face in life, individually or socially, and is therefore capable of being illuminated (and possibly resolved) by cost-benefit analysis.
To illustrate this, suppose that 400,000 New Zealanders will not be vaccinated unless coerced (10% of those above the age of 12). This corresponds to the 10% of the over 12s who have not yet had a first dose, and therefore could reasonably be viewed as a lower bound on those for whom coercion will be required to achieve their vaccination. Standard CBA for health issues involves discounts to QALYs [quality-adjusted life years] for imperfect health status. For example, a person suffering from type 2 diabetes warrants a discount of about 20% per year of their remaining life. The same principle applies to coercion, i.e., it reduces the quality of life of the coerced person. These 400,000 objectors are likely to be of about average age and in good health, which implies about 40 years of remaining life. Let W denote the annual discount on their quality of life arising from being coerced into vaccinating. The QALY loss from the coercion is then 400,000 [people]*40 [individual years] *W = 16,000,000*W.
Now consider the costs that the unvaccinated impose on the rest, of the types mentioned above. Let D be the estimated deaths from the existence of unvaccinated people, if coercion is not adopted compared to adoption of coercion. The deaths here are of people likely to have low residual life expectancies and health problems that would lower their quality of life even if they didn’t die due to the existence of the unvaccinated. Suppose the average residual life expectancy is ten years (generous as covid victims [fatalities] have an average of about five years), and the discount for health problems during this ten year period is 20%.
If coercion is adopted, the 400,000 people alive today who will suffer the coercion will experience a QALY loss of 16,000,000*W whilst the vaccinated avoid a QALY loss of D*10*0.8. So, coercion is warranted if and only if D*10*0.8 exceeds 16,000,000*W.
For example, suppose W is 5%, i.e., coercion is equivalent to a quality of life discount of 5% per year. The parameter D would then have to exceed 100,000 for coercion to be justified, i.e., there would have to be at least 100,000 additional deaths amongst those alive today and vaccinated resulting from catching covid from an unvaccinated person or from inferior hospital care resulting from hospital overload due to unvaccinated people requiring covid treatment. This is not plausible. Alternatively, if W is 1% (coercion is equivalent to a quality of life discount of 1% per year), then D would have to exceed 20,000 for coercion to be justified, i.e., there would have to be at least 20,000 additional premature deaths amongst those alive today and vaccinated resulting from catching covid from an unvaccinated person or from inferior hospital care resulting from hospital overload due to unvaccinated people requiring covid treatment. This too is not plausible.
It is implausible that W is less than 1%, and it is implausible that D would be more than 20,000. It follows that coercing people into being vaccinated does not seem to be a good policy choice.
If you think I am wrong, I invite you to supply a CBA consistent with your view. Simply saying that unvaccinated people inflict damage on the rest of us is not enough.
It is a reasonable challenge.
In my case, it isn’t that I think his policy conclusion is wrong. I don’t think either vaccine coercion or the associated (coming) pass laws can be justified by the scale of the threat Covid poses. But I think Lally’s initial exercise – while illuminating – may overstate the case, at least on cost-benefit grounds (there are some – few – things no price should be placed on, or which we should be very reluctant to do so in the face all but the gravest threats).
Perhaps my greatest unease is around time horizons. The approach seems to treat Covid as something that within a few years will settle down to be either non-threatening or chronic/endemic or something society will choose not to do anything much about. Why do I say that? Because otherwise we have no basis for reaching any judgements (plausible or not) for many lives coercion might save – over, say, 50 or 100 years – but also because if it was to be treated as a permanent issue one would have to include effects not just for the current population cohort but for those yet to be born.
It seems a reasonable approach to me for now – and this is a place where real options analysis is relevant, taking account of irreversibilities – but if so then how credible is it to suppose (assume) that those subject to coercion will experience the same reduction in their quality of life for their entire remaining life (40 years on Lally’s assumption)?
As Lally notes when we exchanged (brief) notes on this, there are two classes of people subject to coercion, those who give in to it, and those who don’t. Lally’s approach does not seem to incorporate the effect of or on the latter group at all. I’ll come back to them.
But what of those who do give in to the coercion and get vaccinated? It seems quite plausible that, at least initially, many will be quite resentful and experience the reduced quality of life Lally mentions (a few may be relieved as coercion gets them out of a corner they’d boxed themselves into). Quite how intense that loss is may depend a bit on what motivated each individual resister. But if Covid settles into being a chronic thing that no one pays that much attention to in a few years hence, how plausible is that those coerced now will be still experiencing a significant (same annual) loss of wellbeing 30 years from now? If we, arbitrarily, allowed for this loss of quality of life for just the next five years then that cost would be reduced to only 1/8th of the scenario Lally uses.
What about what the coercion saves? Lally’s initial approach looks at lives saved this way: “additional premature deaths amongst those alive today and vaccinated resulting from catching covid from an unvaccinated person or from inferior hospital care resulting from hospital overload due to unvaccinated people requiring covid treatment”.
Again, this points to an implicit assumption about the issue being medium-term but not long-term. The potential impact on the health system – and thus on mortality risks for non-Covid conditions – is a real one at present, but surely can’t be with (say) a 10 year horizon, since healthcare capability can be added in that sort of timeframe. But although Lally highlights the potential deaths avoided, he does not factor into this simple model the losses from either severe Covid cases among the vaccinated, or the loss of quality of life to those whose treatment for other conditions is delayed. On our current understanding, he is no doubt right to play down the mortality risks from Covid to the vaccinated (probably quite few in number, and most likely to be people with remaining life expectancy well less than 10 years), but I have less of a sense of how large those other numbers might be.
As I noted, Lally’s approach does not take account of those on whom coercion will not work. That number might be small, at least after a few months, but what if it isn’t? And even if most of the resisters eventually given in (a) they are probably the ones who will face the greatest and longest-lasting loss of wellbeing (people who resisted from conviction rather than just hesitation), and (b) we know that people who lose their jobs in recessions can experience lifetime losses of income, a result that could well translate to some of this group.
And although repressive enough pass laws can probably reduce the risk of these resisters (a) getting Covid, and (b) passing it on to other (vaccinated) people, that is going to be a reduced risk not an eliminated one. A full cost-benefit analysis would need to include an assumption as to how many lives the compulsion exerted over this group might save. I’m not in a position to offer a number myself.
The other factor that would need to be taken into account in a fuller cost-benefit assessment is the cost-benefit of alternative options. For example, what if instead of vaccine “mandates” and pass laws, the government mandated the use of rapid antigen tests in places where particularly vulnerable people were present (eg rest homes, hospitals), or – at times when there was much Covid in a community – at the entrance to any large indoor event? Antigen tests are, after all, focused on identifying infectiousness, presumably the main (health) concern. That testing would have costs – there are no cost-free options – but relative to the vaccine coercion options some real savings (re issues discussed above). The Ministry of Health still appears to have some mysterious aversion to antigen tests, but there is no sign their distaste has ever been properly costed.
I don’t purport to know the appropriate parameter values for each of these variables. But it is the sort of exercise – done more fully – that officials should be presenting to ministers, and making available to outsiders for information and scrutiny.