We argued in Lack of US market & macro volatility both reassuring and troubling (17 January) that “the market’s willingness to look through domestic political and geopolitical events suggests that only a significant exogenous or endogenous shock currently beyond markets’ radar screens (an “unknown unknown”) is likely to really move the needle”.
That unknown unknown, a “black swan” event, has turned out to be a global viral pandemic on a scale not seen since the Spanish influenza pandemic of 1918-1919.
The coronavirus outbreak is now three months old but governments, central banks, corporates and households still face a critical known unknown, in our view, namely the total number people who 1) had the coronavirus, acquired immunity and are no longer contagious and 2) currently carry the coronavirus and are thus potentially infectious.
This includes people who have not been clinically tested – more than 99.9% of the world’s population. We estimate that only 3.3 million people (4 out of every 10,000) have been tested for coronavirus, although testing data are patchy and often released with a lag. The main reason so few people have been tested is the still limited capacity to rapidly and reliably test a very large number of people.
In econometric terms that is a very small sample from which to extrapolate country-wide trends. One implication is that the actual mortality rate may be far smaller than reported.
The high number of tests-per-capita conducted in countries such as South Korea has been posited as an explanation for their relatively low number of coronavirus-related deaths. However, other factors have likely been at play, including the timing of clinical tests, demographics, national health systems’ capacity to treat infected patients and the timing and efficacy of self-isolation and self-distancing policies, including country “lockdowns”.
For now what policy-makers know they don’t know will likely continue to influence country-specific containment plans, as well as domestic measures to support economic growth while ensuring the functioning of financial markets.
Back in mid-January, when only a handful of coronavirus cases had been confirmed in China, we argued in Lack of US market & macro volatility both reassuring and troubling that “the market’s willingness to look through domestic political and geopolitical events suggests that only a significant exogenous or endogenous shock currently beyond markets’ radar screens (an “unknown unknown”) is likely to really move the needle”. That unknown unknown, commonly referred to as a “black swan” event, has turned out to be a global viral pandemic on a scale not seen since the Spanish influenza pandemic of 1918-1919.
The coronavirus outbreak is now three months old but governments, central banks, corporates and households still face a critical known unknown, in our view, namely the total number (with the emphasis on total) of people who have so far contracted the coronavirus.
True total number of coronavirus cases still the critical but elusive variable
The rate of increase in the number of daily recorded cases – people who have been clinically tested and found to be carrying the virus – and deaths has accelerated in the past month, particularly in Europe and more recently the United States. As of 26th March, about 473,500 people had been clinically tested positive for the virus, of which over 21,000 have died and 136,000 recovered, leaving about 337,000 active cases (see Figure 1).
However, we would argue that the two critical, “known unknown” variables are the total number of people who:
1. Had the coronavirus, acquired immunity and are no longer contagious (highlighted in blue in Figure 2); and
2. Currently carry the coronavirus and are thus potentially infectious.
The latter will of course influence the former over time and includes:
- Active cases – people who have been clinically tested as positive and not recovered or died (data are publicly available and timely);
- People who tested positive and recovered (and thus accounted for as a “closed case”), but would test positive if tested again; Data from quarantine centres in Wuhan show that the possibility of recovered patients testing positive again was 5-10%, according to the state-run Global Times;
- People who initially tested negative but would now test positive (the first rounds of testing were after all conducted nearly three months ago);
- People who have not been tested (and thus not been included in the official count of “total cases”) but would test positive today.
The number of people in the fourth category clearly has the potential of being by far the largest because more than 99.9% of the world’s population has not yet been tested. Specifically we estimate that only four out of every 10,000 people on earth has been clinically tested for coronavirus as of date. In econometric terms, that is a very small sample from which to extrapolate country-wide trends.
One immediate implication is that the actual mortality rate may be far smaller than reported. Based on the number of deaths (21,344) and reported cases (473,543), the average mortality rate at a global level over the past three months has been about 4.5%. However, the numerator (reported cases) is likely to be far smaller than the total actual number of people who have so far contracted the coronavirus. Therefore the ratio of deaths to total cases is likely to be considerably smaller than 4.5%.
Only 4 in 10,000 of the world’s population has so far been tested for coronavirus
We estimate, using publicly available national data and figures from international institutions, that as of mid-March about 2.2 million people had been tested. This number has since increased about 50% to 3.3 million thanks in part to an increase in the number of testing kits manufactured and distributed and advances made to shorten the lag between testing and a diagnostic. This still equates to only 0.04% of the world’s population of 7.7 billion or four in 10,000 people.
Testing data are admittedly patchy and often released with a lag. For starters not every country reports their testing numbers using the same metrics, with for example some countries reporting the number of specimens tested, not the number of people (multiple specimens can be tested from each person)[1]. Some countries are only reporting partial data (Chinese authorities have only released testing data for the Guangdong province, not the country as a whole) or data at a state level which then need to be aggregated with the risk of double-counting (e.g. United States, Canada).
We identify four reasons why at a global level (and in most countries) the sample of people who have been clinically tested is very small and arguably statistically insignificant. First, the case for testing people at very low risk (e.g. young, healthy people living in areas with very low population densities) is not compelling. Second, people with only minor symptoms may not get themselves tested given the constraints which health systems already face. Third, according to recent reports some people may be carrying the virus (and thus be contagious) but remain asymptomatic and thus see no need to get clinically tested.
However, the main reason so few people have been tested is the still limited capacity to rapidly and reliably test a very large number of people, although at a country level the numbers admittedly vary greatly. South Korea for example had reportedly conducted over 315,000 tests as of 20th March or 0.6% of its total population – one of the highest ratios among other large economies (see Figure 3). It has since conducted about 18,000 tests a day, taking the total number tested to over 425,000, which is equivalent to over 0.8% of South Korea’s population or 8,000 tests per million people. It is a similar picture in Norway.
Number of tests conducted only part of the story…and solution
The high number of tests-per-capita conducted in South Korea has been posited as a possible explanation for the relatively low number of coronavirus-related deaths in the country to the extent that extensive testing has enabled authorities to isolate infected patients, trace those they were in contact with and contain the rate of infection (see Figures 4 & 5). This theory may also explain the low per-capita number of deaths in Norway relative to most other West European countries.
However, the correlation between the number of per-capita-tests and number of per-capita-deaths is weak (see Figure 6). There are some clear outliers, including at one extreme Italy and at the other countries like New Zealand which have to date recorded no coronavirus-related deaths. Moreover, Germany, Canada, Australia, South Korea, Norway and the UAE have all recorded low per-capita-deaths but with great variations in their per-capita-tests. Moreover, Italy has to date reportedly conducted more tests per-capita than Canada but recorded a materially higher number of per-capita deaths. It is a similar picture if we compare Spain and France.
This clearly suggests that other variables have contributed to countries’ relative number of deaths-per-capita. The timing of clinical tests – i.e. how many tests were conducted when the epidemic was first reported – has arguably been such a variable. Testing many people early on in the viral epidemic means a significant number of people can potentially be isolated and the rate of transmission capped. Conversely, testing even a very large number of people months into a pandemic – when a large percentage of the population may already be infected – is far less efficient in terms of limiting the rate of infection and potential number of deaths. A case of closing the stable door when the horse has bolted.
Other variables beyond the scale and timing of clinical tests have also in all likelihood had a significant impact on the virus’ transmission rate and the number of deaths. These include i) demographics (including average age of the population), ii) general health of the population, iii) national health systems’ capacity to treat infected patients and iv) the timing and efficacy of self-isolation and self-distancing policies (including wearing of masks, hygiene, avoiding crowded places etc…) and more lockdown measures.
Known unknown bearing down on government and central bank policy-making
The bottom line is that the heterogeneity of countries’ demographics and health systems allied to the small number of clinical tests conducted, particularly early on in the outbreak, and the varying efficacy of containment measures shroud the total number of people who have been or are currently infected and the virus’ potential rate of transmission and deadliness.
In the face of this potentially game-changing “known-unknown” a growing number of governments have taken what could arguably be termed the path of least regret, namely instituting some form of lockdown of their populations. We estimate that 17 countries currently have in place nationwide policies of self-containment, albeit with varying degrees of intensity and enforcement of compliance, with Italy having led the way in Europe[1]. These include Belgium, Bolivia, Colombia, Czech Republic, Denmark, Germany, France, India, Italy, Jordan, New Zealand, Norway, Portugal, Slovenia, South Africa, Spain, Tunisia and the United Kingdom and this list looks set to grow further in the near-term, in our view.
Whether and when these lockdowns will materially slow the rate of new infections and deaths and thus be relaxed or conversely tightened up remains an open question. In the meantime, what policy-makers know they don’t know will likely continue to influence country-specific containment plans, as well as domestic measures to support economic growth while ensuring the functioning of financial markets
[1] For example, Japan had in the month to 19th March reportedly conducted about 32,125 tests but because some people were tested multiple times it had actually only tested 16,484 individuals.
[2] There is no formal definition of what constitutes a lockdown in the context of the coronavirus.
