Philosophers often distinguish several kinds of possibility: physical, logical, metaphysical. One can think of a kind of possibility as a restriction on the set of possible worlds that the speaker is prepared to consider – the physically possible, for instance, is what is found in some possible worlds that observe the existing laws of physics (a space that appears to be radically expanded by the indeterministic interpretation of quantum mechanics). I am not a realist about possible worlds, but they provide a handy tool to organise one’s ideas.
Conversation often leaves the species of possibility vague. If I can’t come to dinner, it is not aways necessary for me to say exactly why, and it would be oddly pushy for my host to demand a complete specification of the scope of possible worlds I have taken into account in reaching my judgement. Vagueness can be benign.
But not always. Here I want to investigate the boundary between two important but somewhat vague kinds of possibility: technical and political. I want to suggest that a muddle about this boundary has recently contributed to 60,000 or so deaths – and counting – in my country – and has the potential to cripple some important forms of political action.
People seem slowly to be realising just how badly Britain, and particularly England, has done with COVID-19. There were two imperatives: to prevent deaths from the virus, and to avoid wrecking the economy in the process. Some have argued for strongly prioritising one of these imperatives at the expense of the other, and both views can be defended. What can’t be defended is failing on both counts. To judge from press reports, countries as varied as South Korea, New Zealand, China, Uruguay, Ghana and Senegal have been able to do much better than us on both counts South Korea (with a population over 50 million) has so far had less than 300 COVID-19 attributed deaths; Britain, at least 40,000 – 65,000 if the measure used is excess deaths. And, I am told, South Korea didn’t need a lockdown; they have even kept nightclubs open.
Nor can our performance be excused on the grounds that we had too little information or other preparation to act early enough. British intelligence should have realised the seriousness of the situation at the latest when China closed down Wuhan on the 23rd of January; we knew, or should have known, as much as any other country, except possibly China itself. Maybe we had less materials for testing, more already infected people than South Korea, or a population less familiar with epidemics, by mid-January; but it’s very hard to believe that explains a 100-fold difference in deaths. South Korea has simply been better governed during this crisis, and we need to think why.
The possibility of a second wave doesn’t make much difference to this conclusion. It might make the lockdown retrospectively futile, though I doubt this; it cannot justify the combination of economic risk and health calamity. No massaging of the numbers or the initial conditions in the two countries, or any likely trajectory from here on, will change things much; we are living through one of the most obvious and worst failures of public action since 1945.
A simple timeline here. Doctors in Wuhan began seeing an odd pneumonia in late 2019. WHO was informed of the outbreak on the 31st of December 2019 and of the coronavirus causing it on 9 January 2020. A genome was published on the 10th of January. The Chinese government initially played the outbreak down, but then dramatically shut down Wuhan on the 23rd of January. Even at that stage, human-human transmission had not been oficially confirmed, but scientists were already very concerned. The WHO declared an emergency on the 30th of January and a pandemic on the 16th of March. The British government encouraged handwashing, and required people with symptoms, and people coming from some areas of China and Italy, to isolate, but imposed no major mandatory restrictions on the population a a whole until mid-March. Suspected cases were initially tested, but this was abandoned in early March. SAGE, the Scientific Advisory Group for Emergencies, dramatically changed its advice on the 16th of March. and Britain entered a lockdown on the 23rd of March from which it is now, in mid-June, tentatively emerging. British deaths from the new coronavirus are among the highest, per capita, in the world.
One story people like to tell here is: this government hates experts: they didn’t listen to the science. I think this is dramatically wrong. The politicians did fail to engage with the scientific advice they were getting in the right way. But such engagement does not simply mean listening to and accepting the advice; it involves scrutiny and interrogation of assumptions, discussion rather than deference. If anything, there was too much deference and too little discussion.
Pandemic policy in Britain was elaborated in documents such as the pandemic flu plan. This document (PDF link to apparently the latest version from 2011), contains the following striking sentences:
“Modern mass global transit also affords opportunities for the virus to be rapidly spread across the world, even before it has been identified. The short incubation period of influenza means that within a relatively short period of time a significant number of cases will appear across the globe. It is likely to take at least four to six months after a novel virus has been identified and isolated for an effective pandemic influenza vaccine to become available from manufacturers..
This means that it almost certainly will not be possible to contain or eradicate a new virus in its country of origin or on arrival in the UK. The expectation must be that the virus will inevitably spread and that any local measures taken to disrupt or reduce the spread are likely to have very limited or partial success at a national level and cannot be relied on as a way to ‘buy time’.” (para 2.11-12)
The strategy proceeds to distinguish various stages of response. In the second, termed ‘ assessment’, people are to be tracked and traced. But that’s not going to last a long time.
“These two stages – Detection and Assessment – together form the initial response. This may be relatively short and the phases may be combined depending on the speed with which the virus spreads, or the severity with which individuals and communities are affected. It will not be possible to halt the spread of a new pandemic influenza virus, and it would be a waste of public health resources and capacity to attempt to do so.” (para 3.13)
The death rate projected from the pandemic under consideration is “up to 210,000 – 315,000 additional deaths, possibly over as little as a 15 week period” (box on page 17).
In addition, the strategy is festooned with expressions of scepticism about all manner of possible preventive interventions. The tone of the whole document is a somewhat Olympian fatalism about the deaths of large numbers of people.
Now this very quick abandonment of containment mirrors very closely what the British government did – and what the successful governments didn’t do – in response to Coronavirus. No doubt there were many other influences on the government’s strategy; but the continuity of thinking between the strategy and the actions taken is too strong to be a coincidence.
Policy towards the coronavirus is steered by a committee of ministers, COBRA, which is advised by a technical scientific committee, SAGE. There was some criticism of the presence of government advisers including the Prime Minister’s adviser Dominic Cummings on this committee. Recenty, the minutes of SAGE up till May have been published. A reading of these minutes confirms what the pandemic flu strategy suggests. The pandemic flu strategy was used as a basis for planning in the initial stages (28 January, confirmed 11 February). The assumption that containment was impossible appears never to be challenged and perhaps never really discussed in the early stages. The number of prospective deaths was not known, but the analogy with flu and the apparently high case fatality rate made it easy to calculate on the back of an envelope that if the disease was not contained six-figure death tolls were a serious possibility. (I remember becoming aware of this from a newspaper story during February (26 February) and thinking: if someone were planning to let off a nuclear weapon that was going to kill that many people, we would treat it as an emergency). The main focus of discussion appears to have been not to save lives but to reduce the burden on the NHS, which was to be consulted to establish what shape of curve woud be least inconvenient. Only on the 13th of March, as far as I can see, did SAGE even start to discuss the possibility of more than temporary containment.
One response to these similarities is: we planned for the wrong illness. Pandemic flu,. it might be argued, can’t be contained, but it turns out COVID-19 can. I don’t buy this. In purely technical terms, flu appears to be easier to contain than coronavirus. All the estimates I have seen suggest higher numbers for the initial replication rate for coronavirus infection, in a given cultural and socio-economic context, than for flu. Admittedly, seasonality complicates the reponse to flu and the virus may evolve faster, but I have found no clear demonstration that it is medically harder to stop a flu epidemic than a COVID one.
The problem was not the wrong illness. It was that the pandemic flu plan’s Olympian pessimism about the possibility of stopping a flu epidemic was never a purely technical judgment in the first place; it was in part a cultural and political judgment, and turns out to have been a bad one.
‘It will not be possible to halt the spread of a new pandemic influenza virus’, presented in an official strategy, sounds like a technical scientific judgment, though the rather insistent added comment “and it would be a waste of public health resources and capacity to attempt to do so” gives a clue that economists may have been somewhere near the discussion.
What do we mean when we say that something is a technical judgement ? Roughly this: it is a statement that claims technical epistemic status; that is, it can only be competently assessed first-hand by people who are expert in the relevant technical discipline. Others have to defer to ther experts. And in order to merit deference, the judgement has to command widespread (not necessarily unanimous) consensus among the relevant experts.
When we say something is technically impossible, as with other species of possibility, we imply a constraint on the range of possible worlds under consideration. Roughly, what seems to be envisaged is the non-existence of possible trajectories from the outbreak to its containment in any possible worlds that observe the laws of science as these are understood in the relevant technical disciplines, given the existing state of technology. To assess this first-hand requires knowledge of these disciplines.
What I want to suggest is that the statements I have quoted from the pandemic flu strategy were never technical statements in this sense. Serious reflection would have shown that a combination of social distancing, controls on mobility and test-contact-trace on all symptomatic cases would have some chance in purely technical terms of containing a pandemic of either flu or cortonavirus. The most generous interpretation I can give the statement is something like this: “a pandemic could only be stopped by a draconian and permanent reduction in mobility, social contact and economic activity so large that no government would contemplate it”. (The word ‘draconian’ is actually used on the 13th of February when SAGE discusses limiting transport, implying value judgements that are obviously political). If this is the fairest interpretation, the technicians, or the civil servants, who drafted the 2011 document were second-guessing the politicians and the population rather than giving the whole range of physically possible options; and SAGE effectively went along with it.
Moreover, the assunmption cannot possibly have been adequately tested given its centrality to the plan. In general, SAGE was extremely careful to find evidence for all its major assertions; indeed some potentially important interventions which had strong theoretical basis, like masks, were not promoted because the evidence was weak. But this fundamental asssumption, on which SAGE was working, could not be exhaustively tested because no modern country had implemented anything like the lockdowns that were to spread across the worlds. It’s only now that the lockdowns are being released that we are beginning to test it; and the signs are that it was false.
One way to represent this interpetation in formal terms is this; the costs of containment would outweigh any plausible estimate of the benefits. You then get an interesting paradox: an increase in transmissibility increases both sides of the inequality and it is therefore possible for a more transmissable disease to be more ‘possible’ to stop than a less transmissable one, because the benefits as well as the costs are increased.
Another reaction might be that the 2011 prediction had become true by 2020 because the government had failed to invest in testing facilities, PPE and contact tracing networks. Even if this were true it can hardly have been what the planners had in mind; it would have become an ironically correct prediction of political failure, not a statement of the situation if the government acted as planned. And in any case it is quite likely that a timely reduction of internal and external mobility, social distancing, and trace and contact and isolation on all symptomatic cases- applied as soon as possible after the Wuhan shutdown – would have contained the epidemic even in the absence of testing. We’ll never know, because we didn’t try.
That the flu pandemic strategy and the assumptions that SAGE worked on involved second-guessing the politicians need not itself have been a problem. But the meaning and justification of their assumption was not transparent and the decisions they were assuming turned out not to be the right ones. What SAGE could and should have done, in retrospect, is to communicate both the gravity of the situtation and the actions that might give some chance of containing the epidemic, even if they thought the probability of success was small or the actions were unlikely to be politically acceptable.
One factor is this. People’s ex ante preferences about a disaster often change when the disaster approaches. Cancer patients facing a bad prognosis sometimes accept treatment with little chance of success when they would ex ante have said it is not worthwhile. When the choice set contracts, the ranking of remaining choices can change; this has been familiar in economics since the discovery of Allais’ paradox. Maybe the sheer otherness of a new coronavirus, compared with the familiarity of flu, made people more responsive. The prospect of hundreds of thousands of deaths, combined with a slim chance of stopping the whole thing by timely action, might have concentrated even Boris Johnson’s mind. The politicians needed to understand they had a choice to make.
What we saw instead was a nervous pass-the-parcel between scientists and politicians; the politicians trying to pre-empt debate by insisting they were following the science, the scientists disclaiming any political pretensions. Both sides carry some responsibility; the scientists for not presenting a wide enough range of options: and the politicians for abdicating their responsibility.
Coming to policy decisions usually involves intellectual, not merely technical work – that is, it requires someone to relate the findings of different disciplines rather than working within one. This is even true of the work of SAGE, which involves achieving some sort of consensus between preactitioners of very different disciplines; and some disciplines, such as epidemology, where the practitioners themselves have diverse disciplinary backgrounds. Even more challenging is the task of of relating economic and health effects; even though a cost-benefit analysis might be illuminating (some heroic attempts can be found online) such analyses are rarely conclusive and are often best understood as stimuli to further thought, partly because they involve inherently controversial views about values. Such discussions happened far too late in this case. (SAGE discussed the focus on lives saved or years saved on the 30th of April, long after its most decisive judgements had been made, and (reasonably) never claimed to be able to predict the economic effects of lockdown which would have to be included in a cost-benefit analysis).
A second ambiguity in the statement ‘It will not be possible to halt the spread of a new pandemic influenza virus’ is: which agent is being referred to ? Is this a restriction on the capacities of a policymaker acting in isolation, or on a population that acts as a whole ? Thinking again in terms of possible worlds: the first interpretation restrict the scope to possible worlds in which people other than the policymaker show predictable behaviour and the policymaker alone is free to choose their actions to minimise the epidemic; in the second, we allow this freedom to the population as a whole.
SAGE certainly takes behavioural science seriously. But behavioural science, like neoclassical economics, tends to take the basic motivation of people’s behaviour as essentially given. The idea of experiement rests on this; the behavioural scientist acts, by setting up the experiment, and then hopes that he can identify predictable patterns of behaviour from his subjects’ response. This may be useful for prediction, but if it is the mindset underlying policy it may tend to underestimate the possibility of collective action – of getting the population to share the responsibility for action. Though the SAGE minutes do talk (25 February) about fostering collectivism, there are a number of cases where policy could have involved the population more.
First, Neil Ferguson’s model of the pandemic, which was used to justify lockdown, assumes compliuance ranging from 75% to 50% with social isolation methods measures. Later reports suggests a 50% compliance was actualy achieved but I suspect these data refer to a different concept (perhaps total rather than partial compliance). These numbers might be a little unambitious; it would have been good to do sensitivity analyses with different values of the parameter; an urgent question might then have been how to get this proportion up. (How quickly such analyses can be produced depends on the complexity of the model; there’s something to be said for keeping the model simple enough to allow quick runs with alternative assumptions).
Secondly, data published on the incidence of COVID has been sparse. For most of the lockdown, it has been easy to find news of new cases at the national and I think at the English, Scottish, Welsh and Northern Irish levels, but very hard beneath that (it may be available somewhere but it evaded the Google searches I attempted). Hence it has been impossible for most people to know just how intense the epidemic was in their particular area. This has very recently changed. I suspect it is deliberate; the government did not want people making their own judgements about lockdown.
Thirdly, local government has been substantially excluded from decision-taking and poorly informed for the decisions it has had to take. The local environmental health officers, whose local knowledge might have been central to a ‘trace and contact’ approach, seem to have been ignored. The development of the tracing app has been put in the hands of someone known as the former head of an ISP notorious for bad customer service and security breaches. And the apparent inability of Johnson and Cummings to understand the public fury about Cummings’ journey to Durham and his day trip to the rather Freudian destination of Barney Castle is revealing; even if Cummings has more substantial reasons for the Durham journey than has has wanted to say in public, these don’t mitigate the harm to the lockdown.
Now one shouldn’t have expected a Conservative government to be altogether good at understanding popular collective action, which is, after all, much more intimidating to the conservative cause than any virus. But I have found myelf using the word ‘Olympian’ more than once in this piece. Both the Prime Minister and his main adviser have a classical background. It is possible to read the classiccs from top-down, as providing the canonic ideological justification of hierarchy and imperialism, in which those in power are able to identify with the Homeric deities (I am reminded that the house prefects in ‘Another country’ are known as ‘gods’, though I don’t know whether any school actually used this piece of argot). But, at the very start of the Iliad, Homer tells us that when Agamemnon misbhaves the consequence is a plague in which ‘the people died’ , “’ολέκοντο δε’ λαοί” (Iliad 1.10 with typographical constraints). The consequence of Agamemnon’s response is the first strike in Greek literature. Agamemnon, admittedly, can’t match the shamelessness of our leaders; he does, eventually, sort of, apologise. It is possible to read the classics from bottom up.
The semantics of possibility are complicated further by uncertainty. Return to the two grand formulations of pessimism in the flu pandemic strategy: “It almost certainly will not be possible to contain or eradicate a new virus in its country of origin or on arrival in the UK. …..It will not be possible to halt the spread of a new pandemic influenza virus, and it would be a waste of public health resources and capacity to attempt to do so”. Note how the later, and less technical formulation, dispenses with the tedious qualifiers ‘almost certainly’ and ‘expectation’. Such a shift within the same document is a sign of some sort of internal strain.
Interpreting ‘it will almost certainly not be possible’ in terms of possible worlds is actually quite tricky. Try this: what it means is that within the set of possible worlds where a flu pandemic has got started, it is likely that there is no possible world which leads as a result of public action to its containment; however, it also concedes (as the later more brisk formulation does not) that one can’t be absolutely sure ex ante; such a trajectory might exist in some possible world.
What the pandemic flu strategy recommended, and SAGE followed, was to act as if the possibiilty of containment could be dismissed. The possibility is barely discussed in the SAGE minutes until the 13th of March where they say; “SAGE was unanimous that measures sseeking to completely suppress spread of COVID-19 will cause a second peak. SAGE advises that it is a near certainty that countries such as China, where heavy suppression is underway, will experience a second peak once measures are relaxed”. In effect this judgement reiterates the stance SAGE had taken throughout January and February and was about to abandon; the explicit formulation, and the exceptional mention of unanimity, is a sign that SAGE was by now aware of external criticism. But the judgement is sloppy. South Korea, as noted above, achieved containment without lockdown; and it matters a great deal how large the ‘second peak’ is assumed to be. It may well be near-impossible to prevent some subsequent outbreaks, but the signs are that such outbreaks may be locally containable without a return to massive lockdown, if good systems of test, trace and contact are in place. (The disowned idea of herd immunity also seems to be lurking in the verb ’cause’).
We have already seen that the judgement was wrong and less purely technical than it looked. But the further problem was that it involved a very odd and defeatist treatment of risk. What the planners should have recognised is the enormous benefit to achieving containment by a successful trace and contact approach in the early stages; so large that it might be – and in the event was – worth trying to achieve such an outcome even if the odds were (particularly in the absence of stocks of testing materials) stacked against its success.
The idea of ‘pessimism of the intellect, optimism of the will’ is variously attributed to Rolland and Gramsci. Gramsci’s formulation relates to his long incarceration and is worth quoting both for its dignity and its Hegelian trope; “it seems to me that in such conditions, prolonged for years, with such psychological experiences, one should have achieved the maximum degreee of Stoic serenity, and have acquired such a deep conviction that man has in himself the source of his own moral forces, that everything depends on him, on his energy, on his will, on the iron coherence of the ends he set himself and the means that he develops to realise them – on never despairing again, on never fallling back into those vulgar and banal states of mind called optimism and pessimism. My state of mind synthesises these two feelings and overcomes them; I am pessimist with the intelligence, but optimist by the will. I think, in every circumstance, about the worse hypothesis, in order to put all the resources of will in movement, and to be up to overcoming the obstacle. I have never made illusions for myself and I have no delusions. I have always particularly armed myself with a limitless patience – not passive, inert, but animated by perseverance”. (Lettere del Carcere, letter of 19 December 1929 to Carlo, probably his brother; my, rather literal, translation).
What SAGE embodied was the pessimism of the intellect untempered by optimism of the will. SAGE spent much time trying to identify a reasonable worst case; it never seriously asked what best case, even if unlikely, might be achievable. What looked like sober realism was lethally costly. And this applies not only to the grand decision not to attempt containment but a whole host of micro-inactions, since such things as face masks and and stopping large gatherings were assumed to be useless unless hard evidence was available that they would work – evidence which, in view of the speed of the epidemic and the number of other factors involved, was always going to be very difficult to produce. (Drugs, of course, are another matter, being more dangerous and easier to test; SAGE was clearly trying to get medical trials up and running as fast as possible). The risk of doing too much, or going too early, was for some reason much more salient in SAGE’s deliberations than the risk of doing too little too late.
Giving advice is difficult and a skill that tends to be learned, or not learned, on the job. The errors I am identifying in SAGE’s conduct are errors that I have seen and may well have made when I worked as an economic consultant; what is unusual is not the errors, but the drastic visibility of their consequences. SAGE’s members are distinguished and industrious people and the references to pastoral support in the SAGE minutes suggest they have been under intense strain. What went wrong is something structural.
SAGE exists, according to its website, to ‘ensure that coordinated, timely scientific and/or technical advice is made availabe to decision makers to suppoirt the cross-govenrment decisions in COBR”. There are two problems lurking here. First, the need for ‘coordination’ can easily encourage groupthink. The fatalism that characterised SAGE’s advice was not universally shared even in Britain, where John Ashton’s angry public denunciations of the the neglect of basic public health practices were one example among many; and it did not reflect the views of WHO or of the East Asian policymakers who knew a bit about coronaviruses. Similarly, Neill Ferguson’s modelling may be very good and has been unfairly attacked, but it isn’t the only view on offer. SAGE is in principle meant to say whether their advice reflects a general technical consensus; but expecting time-stressed advisors facing political resistance to articulate this may be unrealistic. Secondly, ‘advice’ can take very different forms. It might mean recommendations for action given objectives explicitly prescribed by the politicians; it might mean recommendations for action based on SAGE’s own view of the objectives or SAGE’s best guess of the politicians’ priorities; or it might mean the provision of a set of options, with estimates of the likely consequence of each one. Given that many politicians’ priorities may be internally inconsistent, unpredictable, or implicit rather than explciit, the scope for confusion is considerable. And the different approaches would have different implications for the involvement of policymakers or other advisers in SAGE’s deliberations.
Perhaps the greater failure lies on the side of the politicians. The COVID pandmic is inter-sectoral; it needed leadership from the a strong Prime Minister or a strong deputy. Such a person was needed to mediate between the MOH and the Treasury, and also between the technical advice and the formulation of policy. In relation to SAGE, they might have done three things: ask how many people are likely to die (something that had already been very approximately guessed before Neil Fertguson’s full model): ask how the horrifying prospect of a six figure total could be avoided; and – the hardest thing – refused to accept that containment was impossible without seeing the underlying reasoning; reasoning which I have suggested was partly political and might not have survived scrutiny. That takes a politician with brains (which Boris Johnson has), intellectuality and application. Judging from his absence from COBRA duiring the crucial period, Johnson wasn’t interested.
Back to philosophy. Two major lessons emerge. First, the work of relating different disciplines and using technical evidence to formulate strategy requires intellectuals, not merely technicians. Our problem in Britain is that we understand, or think we understand, ‘experts’ such as scholars or technicians – and are willing enough to defer to their expertise – but we have no well-developed concept of the intellectual, who works between discplines and does not claim expertise. I shall have more to say about this in further essays on this blog. Many expert disciplines have much internal dissent; and even when there appears to be internal consensus, it can be wrong, and its status and scope needs illumination by the light cast from outside by intellectuals.
Secondly, statements about possibility are easily misunderstood and often poorly backed by evidence. But they are too important simply to be avoided. Practical ethics arguably gains by categorically ruling out certain actions as unthinkable. Many textbook problens in physics are solved by imposing constraints such as the conservation of energy and momentum on the movement of particles; the development of special relativity depended on Einstein’s gift for seeing which restrictions might most productively be imposed on the the behaviour of bodies moving at very high speeds. It is not that statements of impossibility can be avoided: it is that one needs to be clear about what one is doing in making, or using, them. What is their epistemic status ? – matters of logical proof, like Gödel’s and Arrow’s theorems – basic organising postulates, like the conservation of energy, or contestable and refutable hypotheses within a discipine ? What is the scope of possible worlds the statement quantifies over ? Who is the implied agent ? And: what is the practical cost, in any given situation, of simply assuming their truth rather than trying to prove their falsehood ?
I think that this applies also to political projects. Take a project designed to promote worker ownership. There are several reasons why such a project might be impossible. Its assumptions might violate accounting identities; in this case there is a basic incoherence. It might involve flows of commodities or changes in stocks that can only be sustained for a time; here too there is a problem (see Sraffa for an approach to economics that tries to use just this constraitn but leave behaviour free). It might involve changes in the motivation of the relevant agents. Neoclassical economics tends to encourage the idea that this can’t be done; but a more general and productive use of economic theory might be to make explicit the changes in motivation that would have to happen to make the project succeed, and see them as part of the project. (Somwhere, the great development eonomist Albert Hirschman suggested that the most interesting achievements of development projects often consist of removing the constraints that were impeding their intended objectives). Or it might be consistent with current individual behaviour, but obstructed by politics. It is not that anything is possible; it is that there are different kinds of possibility and you need to be clear about which you mean; and in many cases you will never really know unless you try.