I was going to move on to another topic, but last night University of Waikato economics academic John Gibson sent me the links to a couple of other papers I hadn’t seen, and I thought it might be worth writing about them. Gibson is one of New Zealand leading empirical research economists and during the lockdown I wrote about one of his efforts to think through, and put numbers on, the costs and benefits of the lockdown, linking lost GDP to possible reductions in life expectancy.
The first of the links Gibson sent through probably won’t appeal to most readers. In this paper, two Motu research economists, Arthur Grimes and Benjmain Davies, set out to formalise how one would apply what is known as “real options analysis” to the choices the government made in late March. Real options analysis was an addition to the economics literature in the 1990s. In many ways, it was one of those blindingly obvious ways of looking at things that was probably second nature – often unconsciously so – to many people making all sorts of decisions in life, but which hadn’t been part of the formal economics toolkit until then. As Grimes and Davies describe it
A standard result from real options theory (Dixit & Pindyck, 1994; Guthrie, 2009) is that delaying decisions to act can be valuable when (i) decision timing is flexible, (ii) some outcomes are partially or fully irreversible once action is taken, (iii) uncertainty exists about the evolution of an exogenous process that impacts the outcomes of interest, and (iv) the decision-maker can learn about the evolution of the exogenous process over time. Delaying action preserves the option to make a future decision without locking in irreversible costs prior to new information arriving.
It is often applied to private sector investment decisions, but can just as much be applied (and probably should be more often) to some government investment or regulatory etc decisions (or other private choices – one might think, for example, of a proposal of marriage).
But as Davies and Grimes note, in the choices the government faced in late March, there was uncertainty and potentially irreversible losses whichever direction the government took. In their words
Conditions (i)–(iv) were all met at the outset of COVID-19. However, the two-sided uncertainty at that time made it unclear which option should be preserved: the option to protect economic output initially (by avoiding lockdown) or the option to preserve the chance to eliminate COVID-19 (by entering lockdown).
This is, of course, something of an oversimplification, since there were degrees of possible regulatory responses (and “elimination” itself was not yet the government’s stated goal at the time), and many – but probably not all – of the economic losses would have happened anyway, as individuals and firms responded to perceived risks – but the point of the short paper is to illustrate the framework, not to offer empirical answers. The authors chose the March lockdown decision to illustrate the framework, but they could just as well – or so it seems to me – have applied it to, for example, the decision to close the border to travellers from the PRC in early February, or to the decisions the government took last week regarding the latest Covid outbreak.
It would seem, also, to be a useful framework in which to think about the way ahead from here – not in any mechanical sense, but as a way of helping to organise thinking.
The second paper, by Gibson himself, is likely to be of more general interest and – since it does reach a specific conclusion – controversial. I’m a little surprised it doesn’t seem to have been covered elsewhere already. An earlier version of Gibson’s paper is available here as a University of Waikato Working Paper, and although I will be quoting from a more recent version that Gibson sent me, the abstracts of the two versions are word-for-word identical.
On this occasion the conclusion is well-captured in the title of the paper, “Government Mandated Lockdowns Do Not Reduce Covid-19 Deaths: Implications for Evaluating the Stringent New Zealand Response”. Capture your interest? It certainly did mine.
It is an empirical paper using US county-level data, and thus taking advantage of the fact that regulatory powers on these matters typically do not rest at federal level.
Gibson begins by noting that epidemiologists’ simulation models are simply not fit for purpose when it comes to evaluating likely deaths (and, thus, deaths saved from interventions). Writing of the apparent influence such models had – whether for support or illumination – in New Zealand, Gibson writes
It is unfortunate that epidemiological simulations had such impact. The Susceptible,
Infected, Recovered (SIR) epidemiological model, and variants with Exposed and Dead (SEIRD), have infectious people mixing (homogeneously) with others; each person has equal chances to meet any other, regardless of their health status. Yet in reality, people engage in preventative behaviour to reduce risk of exposure; allow for this, and some public actions designed to reduce disease spread may do more harm (Toxvaerd, 2019). These models also have too many degrees of freedom, so are poorly identified from short-run data on cases. For example, Korolev (2020) shows long-run forecasts of U.S. COVID-19 deaths from observationally equivalent SEIRD models ranged from about 30,000 to over a million.
Forecast deaths depend on arbitrary choices by researchers, and data at the time cannot show which forecast is right as so many models are observationally equivalent in the short-run. Elsewhere, Swedish researchers using the Imperial College approach forecast (in mid-April) 80,000 Covid-19 deaths by mid-May (Gardner et al, 2020). In fact, just 3500 died by May 15, with the forecast more than 20-times too high. A final example is the Otago forecasts, which had assumed no case tracing and isolation; using the same simulation model, Harrison (2020) set tracing and isolation success at 50% and forecast deaths fell by 96%.
Harrison(2020) is Ian Harrison’s paper that I have previously written about here.
Gibson’s approach is different
My research design exploits variation among U.S. counties, over one-fifth of which just had social distancing rather than lockdown. Political drivers of lockdown provide identification. If the Prime Ministerial claim, that sans lockdown tens of thousands of New Zealanders would die, is correct then one would expect to see more deaths in places without a lockdown. This may explain global fascination with Sweden, as a country without lockdown. However a within-country research design has two benefits; less variation in measuring Covid-19 deaths than for between-country comparisons, and it better suits the highly clustered nature of Covid-19. For example, Lombardy’s Covid-19 death rate was 1500 per million versus 300 per million elsewhere in Italy. The New York death rate (by May 15) was 1410 per million but just 190 per million in the other 49 states. Taking China’s data at face value, Hubei’s death rate was 76 per million versus 0.12 per million elsewhere. With such clustering, analyses using national averages may mislead.
In practical terms, his regression model is as follows
The regressions use 22 control variables, including county population and density, the elder share, the share in nursing homes, nine other demographic and economic characteristics and a set of regional fixed effects. These controls explain about two-thirds of variation in log deaths (as of mid-May). Even with these controls, the errors for the log death equations may correlate with treatment status, if selection into the treatment group (77% of counties) is due to unobservables. Political drivers of lockdown are plausible instruments; counties without lockdown are all in states with Republican governors and if a gubernatorial election is set for November 2020 (11 are) lockdown was more likely. Conditional on the state-level factors, the extent a county became more partisan between the 2012 and 2016 Presidential elections, relative to the state-level change, affects odds of lockdown. It is hard to think of other paths for these variables to affect Covid-19 deaths than via political calculations about lockdown.
There is a fair amount of technical detail in the paper. Many of the expected things do turn out to have mattered. Thus
…almost two-thirds of variation is explained by early May. The models show deaths are higher if the elderly or those in nursing homes are more of the population; patterns noted in popular discussion of Covid-19. Deaths are higher if whites are a lower share and blacks a higher share of the population, as noted by Millett et al (2020). Counties with higher inequality and more people without health insurance experience more deaths. Fewer deaths occur if the smoking rate is higher, similar to what is found in the U.K. for 17 million NHS patients, where Williamson et al (2020) find current smokers less likely than others to die (as hospital in-patients) with confirmed COVID-19
But this is Gibson’s summary of his results
So the firmest conclusion is that over more than two months after New Zealand’s March 23 lockdown decision, there was no evidence of more Covid-19 deaths in places without lockdowns.
Moreover, he suggests that this was apparent from data that would have been available to New Zealand policymakers when they made lockdown decisions from March to May (and, presumably, of course for this month’s decisions).
Some readers may be inclined to instantly dismiss Gibson as some sort of off-the-planet person simply out to get the government. I have no idea of his personal politics – and, as I’ve noted, he is a highly regarded New Zealand economists who seems to go where the data lead – but in any case as he notes it isn’t as if he is the only one to find similar results
This ineffectiveness has several causes: real-time activity indicators suggest threat of Covid-19, rather than lockdown per se, drives behaviour (Chetty et al, 2020). Just one-tenth of the 60% fall in consumer mobility in the U.S. was from legal restrictions, with the rest from people voluntarily staying home to avoid infection (Goolsbee and Syverson, 2020).
I don’t suppose anyone has Raj Chetty pegged as (say) a Trump supporter, and as for Goolsbee, that is Austan Goolsbee, former chairman of the Council of Economic Advisers in the Obama administration. In addition to his paper, there is an accessible interview with Goolsbee here. This bit captures the point
Adi Kumar: You and Chad Syverson recently published a paper with the National Bureau of Economic Research called, Fear, lockdown, and diversion: Comparing drivers of pandemic economic decline 2020. There you attribute most of the drop in business activity in the United States to people’s own decisions to stay at home, rather than government-imposed restrictions. Can you explain your hypothesis and its implications for policy makers grappling with strategies to reopen the economy?
Austan Goolsbee: We looked at phone records that tracked the locations of 2.3 million businesses around the country. These were mostly retail and services, the kinds of places people physically visit. When we plotted business activity against lockdown timelines through the pandemic, we found that consumer behavior was not aligning with lockdown orders. The visits had trailed off before these were imposed.
We began asking whether government orders drive behavior or not. It’s the classic “identification problem” in economist language—was it causation or just correlation? The disease triggered fear and led people to stop going outside. Then authorities passed laws requiring that they stay at home. So it’s important to figure out how much of what happened next—the sharp fall-off in consumer activity—came from individual choice and how much from public policy.
Our basic idea is to compare places where policies are different on either side of a state border. In Illinois we had shutdown orders, but across the border in Iowa they didn’t. Several metro areas span that border and we have 110 different industries. Take barber shops as an example. If the policies were driving the activity, then we should have seen people still getting haircuts in Iowa but not in Illinois. But that didn’t happen. In the same week, everyone stopped getting their hair cut by similar amounts. That kind of evidence leads to the conclusion that the 60 percent drop in consumer activity from pre-COVID-19 times to the depths of the pandemic was more about individuals’ own decision to stay home. We found that only about 7 percent of the fall-off was due to the policy. Everything else we attribute to other factors, mostly fear.
Those results aren’t about deaths directly, but about mobility and economic activity, but of course the logic of the case for lockdowns is that it is those reductions – forced interpersonal distancing – that reduces future case and death numbers.
We saw this in New Zealand itself before official restrictions were put in place. I guess everyone has their own story: mine was of a trip to Auckland on 19 March, before there were any domestic restrictions in place. Flights were already being cancelled, Wellington airport mid-morning was largely deserted, and my taxi drivers in Auckland told of the hours they had spent waiting for a single fare.
So what are the implications? This is from Gibson’s abstract
Instead, I use empirical data, based on variation amongst United States counties, over one-fifth of which just had social distancing rather than lockdown. Political drivers of lockdown provide identification. Lockdowns do not reduce Covid-19 deaths. This pattern is visible on each date that key lockdown decisions were made in New Zealand. The ineffectiveness of lockdowns implies New Zealand suffered large economic costs for little benefit in terms of lives saved.
He is, at least implicitly, arguing that we’d have had just as much distancing, in aggregate, without lockdowns (in this case he refers to so-called Level 3 and Level 4 restrictions), as with them. (As he and others – including Grimes and Davies – have noted, official advice later released reveals that as late as 20 March official advice to the government had been to stay at the new “Level 2” for 30 days.)
From the final page of his paper
In terms of implications for the future, these results add to the evidence that lockdowns are ineffective. This was also the prior view in public health; for example Inglesby et al (2006: 371) noted: “It is difficult to identify circumstances in the past half century when large-scale quarantine has been effectively used in the control of any disease.” So when the next pandemic occurs, the Covid-19 lockdowns should not be considered a success that should be replicated. …..If decision-making from March and April is reviewed, any claim that lockdown was necessary to save lives can be treated with strong scepticism. It is especially concerning that there were data available, on the dates of those key decisions, to show that lockdowns are ineffective at reducing Covid-19 deaths.
How plausible is all this? Perhaps experts in the specific data Gibson uses, or specialist econometricians, can pick some holes and raise some specific doubts. But as Gibson notes his isn’t the only paper pointing in this sort of direction (and he tells he is finishing another paper suggesting that for a major emerging country “the mobility declines predated the local lockdown orders by two weeks”. One should probably never revise one’s view too much based on a single set of results, but they can’t be discounted either.
Note also that these results are about “lockdowns”, and are not a direct commentary on the role and value of things like enforced isolation of those found to have the infection or, at least as I read it, on border closures and associated managed isolation policies. If so, perhaps it is plausible to suppose that private choices – firms, households, rest homes, community, sporting and religious groups etc – would have brought about sufficient distancing in New Zealand to have resulted in eventual (domestic) elimination, perhaps at no more deaths than actually occurred in New Zealand. Perhaps.
If that were so, of course, it would pose questions about the value of the partial lockdown Auckland is currently experiencing.
As Gibson notes, the response of some people is likely to be along the lines of saying that even if his results are (a) robust and (b) applicable to New Zealand, why does it matter? We (and those US counties that saw deaths fall) got there anyway.
A non-economist might say “what difference does it make?” If people would reduce
interactions anyway, due to perceived Covid-19 risks, having government force them to stay home would seem costless. Yet as economists know, a government diktat approach runs into the central planning problem; no central planner has all the information (collectively) held by parties involved in voluntary exchange (Hayek, 1945). For example, absent lockdown, if a butcher felt they could operate safely and if customers felt they could safely shop at this butchery, voluntary and beneficial exchange could occur. Instead, under the central planning approach applied in New Zealand, butchers were shut but supermarkets selling meat were not. Potentially, much economic surplus (for both consumers and producers) was lost.
And that would seem to be just a small part of it (Gibson was tightly word-count constrained). We currently have massive overlays of officials deciding who/what is or isn’t a permitted exception to the internal border restrictions – the sort of thing that should have been sorted out months ago, given the lockdown policy was always potentially regional – and associated delays, rather than private firms and households making their own choices about what risks to run, or not. Or we had the official gross over-reach that prohibited you from going for a quiet solo swim at a calm suburban beach in mid-autumn, that makes rules on where, when and what you could hunt – none of which had anything to do with public health. Or that prohibited a priest attending in person to a dying – or bereaved – parishoner, or that prohibited funerals altogether for a time. Or that banned people from making choices to have small distanced outdoor services – having advice to hand about risks – to celebrate Easter, because presumably such things were too hard for officials, unimportant to our ministers etc. Or – in my case, and perhaps trivially – the lockdown rules prohibited me taking my son out for driving lessons, again with no public health implications for anyone.
All that of course, assumes that Gibson’s results are robust. But it all goes to the more general point that proper marginal cost-benefit analyses should have been being done by officials and ministers – should now be being done – and aren’t. It has been known from the start that private distancing choices would make a material difference, but those rational private choices have too rarely been seen factored into New Zealand official decisionmaking.
There is, however, one area in which I think Gibson overstates his case, perhaps quite materially, and that is on the economic consequences of the lockdown choices. He notes that
Treasury assume that output at Level 4 was reduced by 40%, at Level 3 by 25%, and
at Level 2 by 10-15% (Treasury, 2020). So even with a V-shaped shock and recovery rather than a U or L shape, 33 days of Level 4 and 19 of Level 3 (that ended May 13) would reduce output by ten billion dollars (ca. 3.3% of GDP) compared to staying in Level 2 throughout.
In terms of the (recent) past, the ineffectiveness of lockdowns implies that New Zealand suffered large output losses, of ten billion dollars or more, for no likely benefit in terms of lives saved as a result of the decision to move almost immediately from Level 2 to Level 4.
But this is almost certainly wrong, and in fact inconsistent with many of the other sorts of results re mobility etc that Gibson cites. We simply do not know how large a share of those (guessed) Level 3 and Level 4 losses would have occurred anyway, as people and firms wound back their own activity. I know that I had already decided that our children would not have been going to school the next day, if the government had not pre-emptively closed the schools. Not many people would have been at restaurants, cafes, movie theatres or perhaps even churches in late March and early April, no matter what the government had decided (perhaps especially if they’d still been waving around scary death predictions). Quite possibly much of the construction sector would have stayed working throughout – even officials wanted to keep that open in Level 4 but ministers refused – but a large chunk of the lost output would have happened anyway, at least for several weeks. From my exchange with him last night, I get the impression Gibson is more optimistic about the economic difference than I would be, but the 3.3 per cent of GDP number must be seen as an overstatement of the economic cost of the lockdown itself.
As I’ve said repeatedly in this series of posts, I’m not championing any particular policy approach from here (although I have been inclined to the view – in March and now – that the government itself has been inclined to over-react, using sledgehammers (at little or no cost to themselves and officials, in fact possibly feeding saviour narratives) when something more nuanced could credibly have done the job). I’m not even fully convinced by the Gibson story but – in particular coming from someone of his stature – it deserves to be taken seriously, tested and critiqued rather than – as some will be tempted to, for a variety of different reasons – dismissed out of hand.