The daily release of countries’ COVID-19 case counts consistently makes front page headlines despite conveying little information. By contrast, a relatively underutilized measure provides more relevant cross country comparisons: the slope of the epidemic S curve.
Here’s the problem with case counts. First, they are driven by countries’ populations: the US has more cases than Italy but once adjusted for population, US cases are 25% of Italy’s. Second, the figures are highly dependent on the testing capabilities of countries. The more countries run tests, the more cases they find. The percentage of positive test results would be more informative. But that’s also problematic. With most countries experiencing test kit shortages, they have opted to test the most exposed segments of their populations, which creates a strong upward bias in the results.
Thankfully, the COVID tracking project provides data on US states’ confirmed cases, tests numbers and test results, providing some sense of the measurement biases involved in looking at the number of cases alone.
This article is only available to Macro Hive subscribers. Sign-up to receive world-class macro analysis with a daily curated newsletter, podcast, original content from award-winning researchers, cross market strategy, equity insights, trade ideas, crypto flow frameworks, academic paper summaries, explanation and analysis of market-moving events, community investor chat room, and more.
The daily release of countries’ COVID-19 case counts consistently makes front page headlines despite conveying little information. By contrast, a relatively underutilized measure provides more relevant cross country comparisons: the slope of the epidemic S curve.
Here’s the problem with case counts. First, they are driven by countries’ populations: the US has more cases than Italy but once adjusted for population, US cases are 25% of Italy’s. Second, the figures are highly dependent on the testing capabilities of countries. The more countries run tests, the more cases they find. The percentage of positive test results would be more informative. But that’s also problematic. With most countries experiencing test kit shortages, they have opted to test the most exposed segments of their populations, which creates a strong upward bias in the results.
Thankfully, the COVID tracking project provides data on US states’ confirmed cases, tests numbers and test results, providing some sense of the measurement biases involved in looking at the number of cases alone.
Table 1: US States Tests and Cases as of March 30th, 2020
Source: COVID Tracking Project, Macro Hive
First, while New York, New Jersey and Massachusetts are in the top five for both absolute and population-weighted cases, California is the outlier (Table 1). It has the fourth-largest number of cases, but relative to its population of 40 million falls in the second quintile of states.
And, of course, there are huge difference within states: New York city for instance has become an epidemic hotbed, while counties in upstate New York have no confirmed cases. The CDC has identified 15 states with widespread community transmission, with the rest having only defined areas of community transmission.
Chart 1: Cases and Test Results
Source: COVID Tracking Project, Macro Hive
Second, the sample selection bias mentioned above is very large (Chart 1). All of the states have positive tests results higher than population-weighted cases at least by a factor of 100. That said, states that have run more tests tend to have a smaller sample selection bias. As states are able to run more tests, they select more representative samples.
These differences in states’ sampling methodologies highlight the difficulty of comparing case numbers cross-state, even when adjusted for population. A less biased metric could be the slope of the epidemic S curve, assuming that states keep their sampling methodologies roughly constant. We can capture the slope by computing the number of days required for the cumulated case number to double.
That provides a more nuanced picture of the US epidemic. First, it shows that New York is starting to flatten its curve but that the epidemic is progressing fast in neighboiring Connecticut and New Jersey. Second, among worst affected states, Louisiana has made more progress than Massachusets. In addition, even the least affected states are starting to flatten their curve, which suggests that, overall the social distancing measures recommened by the CDC are starting to have an impact.
Table 2: Cases as of March 30th, 2020
Source: John Hopkins, Macro Hive
Applied at the country level, the numbers of days to double also provides a better picture of relative containment progress (Table 2). It shows that while Switzerland has the second highest number of population-adjusted cases, it has also made substantial progress in flattening its curve. Conversely, while the US and UK have low population-adjusted case numbers, these are likely underestimates: the low number of days to double the case numbers (i.e. the steep slope of the S curve) suggests testing is till still catching up to the actual epidemic (due to data issues, I am discussing neither China nor Japan).
Chart 2: Asia: Days to Double Case Numbers
Source: John Hopkins, Macro Hive
In addition, the number of days to double facilitates a more accurate comparison between the efficacy of different containment policies. For instance, while Korea was a late-starter in implementing containment measures, its strategy of massive testing and contact tracing has proved the most successful (Chart 2). It has had more limited economic costs than, for instance, Italy’s lockdown, though this has come at the expense of a substantial loss of individual privacy.
Chart 3: Europe: Days to Double Case Numbers
Source: John Hopkins, Macro Hive
Sweden’s days to double have recently fallen, though this reflects the country’s strategy rather than a setback (Chart 3). Sweden has opted to use part of its testing capacity for “sentinel testing” (i.e. testing of its general population to measure the general progression of the epidemic). As a result, the country has been able to better anticipate the demands on its health care system and to modulate social distancing measures based on its health care capacity. For example, with new cases picking up, on Friday Sweden tightened its limit on public gathering to 50 from previously 500.
This strategy avoids the economic costs of Italy’s and the privacy costs of Korea’s. On the other hand, it is more reliant on a specific epidemic model. The risk is that if the model underestimates the epidemic’s progression, Sweden could be forced to enter lockdown to relieve the pressure on its health system. Nevertheless the experiences of countries such as Korea or Sweden suggest that while economic normalization is likely to be a trial-and-error process, it is well within reach.
Dominique Dwor-Frecaut is a macro strategist based in Southern California. She has worked on EM and DMs at hedge funds, on the sell side, the NY Fed , the IMF and the World Bank. She publishes the blog Macro Sis that discusses the drivers of macro returns.
(The commentary contained in the above article does not constitute an offer or a solicitation, or a recommendation to implement or liquidate an investment or to carry out any other transaction. It should not be used as a basis for any investment decision or other decision. Any investment decision should be based on appropriate professional advice specific to your needs.)