Silver BlogCoronavirus Data and Poll Analysis – yes, there is hope, if we act now

We examine the growth of coronavirus daily cases in most affected countries, and show evidence that social distancing works in reducing the rate of spread. We also analyze KDnuggets Poll results - the scale of change to online and how Data Science work is likely to increase or drop in different regions. Stay Healthy and practice social distancing!

The latest KDnuggets Poll asked KDnuggets readers about the impact of this virus on their travel and work. Before analyzing the poll results, I want to take a look at the coronavirus data.

The novel coronavirus pandemic is among the greatest tests that humanity has ever faced, and our response in the next few weeks will make the difference between many millions dead in the worst case, and much smaller numbers if people collectively practice social distancing and other needed steps to buy time for medical researchers to come up with treatments.

There is a lot of data on coronavirus, and KDnuggets readers, as people with Data Science training, should be the first to understand that this data is incomplete. First, not all people who have the virus are tested. Not all countries or regions implement testing at the same level, or report the results accurately. However, it is still useful to look at the data over time to see trends.

One of the most illuminating charts is how the number of cases grows. The initial growth rate from the first reported case is slow, so we start measuring from day 0, defined as the day when the number of cases is greater than some large threshold.

I took the number of cases from JHU repository and plotted the number of cases for countries with the most cases and the highest growth, excluding China (where the growth has almost stopped). I took day zero as the first day when the number of cases exceeded 90 (not 100, since some countries were very close but below 100).

The horizontal axis shows the time, starting with day 0, and the vertical axis shows the number of cases on log scale. Log scale allows us to see patterns of exponential growth much more clearly. Next to the country name is the first day (Day 0), when number of cases exceeded 90.

Coronavirus Cases Top Countries
Fig. 1: Coronavirus cases for most affected countries, (except China) since day when cases>=90, as of March 21, 2020.

Black line is the exponential trend based on Italy cases. Some observations.
  • We see that South Korea has succeeded in bending the curve after about 2 weeks.
  • Iran is also slowing the growth, although there is some doubt as to whether they report the data accurately.
  • Italy growth rate has slowed - their line is curving down from the black trend line.
  • Germany is tracking Italy almost perfectly, with a delay of about one week.
  • France and Switzerland are a little below Italy's path, with Switzerland about one week behind France.
  • Unfortunately, US (red) and Spain (gold color) are exceeding the Italy growth rate, and are likely to overtake China and then Italy in the number of cases soon.
  • The largest (worst) growth curve is in New York State, where the number of cases as of March 21 has exceeded 15,000.
However, examining the total numbers is not the best way to measure the trend. We can get a clearer picture if we plot the daily growth rate, defined as Cases(DayN+1) / Cases(DayN).

Furthermore, daily changes have a lot of random noise, so we smooth the curve by looking at the 5-day moving average of daily growth rate - see Fig. 2.

Coronavirus Cases Top Countries, Daily growth rate, 5-day moving average
Fig. 2: Coronavirus daily change in cases for most affected countries, (except China) since day when cases>=90, 5-day moving average, as of March 21, 2020.

The good news here is that for most countries the growth rate is slowing down. South Korea has succeeded in slowing the growth rate to only about 1%. Iran has also slowed the growth rate, although there are doubts as to how accurate is their reporting.

The drastic measures in Italy are having an effect - the daily growth dropped from 40% to about 12%, as of March 21. We also see declines in daily growth rate in most other countries. The one exception is the US, where the growth rate has increased, especially in New York City and State. This may be partly because of increased testing, and partly because of the faster spread of the virus in New York, where according to news reports and what I hear from my friends there, many are not practicing social distancing.

So, we have evidence that social distancing measures work in slowing down the virus, and now it is incumbent upon all people in affected countries and region to practice it to avoid the worst case scenarios.

Poll Results and Analysis

OK, back to the poll. The latest KDnuggets Poll, taken over period March 11-20, 2020, asked KDnuggets readers about the impact of the novel coronavirus on their travel and work. 1268 respondents took part.

The first question was
Q1: Has coronavirus impacted your conference/other travel plans (conferences/trips cancelled)?

The average percentage of those who answered Yes was 75%, and we can see that the percentage has increased over 10 days the poll was active - see Fig. 1.

Poll Coronavirus Travel
Fig. 3: Percentage of people who conference/other travel plans were affected by the novel coronavirus.. The blue line is the overall average = 75%.

Next, we asked

Q2: Because of the coronavirus, your work/study is now:

Poll Coronavirus Work Online
Fig. 4: How work online has changed because of coronavirus. 66% work more or only online, including 28% who now work only online.

We also asked about employment type, and the breakdown was
  • Industry/Self employed, 55.8%
  • Student, 16.6%
  • Academia, 11.8%
  • Government/non-profit, 7.6%
  • Other, 4.7%
  • Unemployed, 3.5%
Next we chart the impact of coronavirus on online work based on employment type, for 3 largest groups. Combining "more online" and "only online" we see that on average 66% of respondents are now working more or only online. This ranges from 76% for Academia (who now work 50% more online) to 61% for students (who now study 32% only online).

Poll Coronavirus Work Online by Employment
Fig. 5: How work online has changed because of coronavirus, by employment type.

Finally, we asked about expectations for the amount of work.

Q3: Because of coronavirus, in the next 3 months do you expect:

31% expect less work, and 19% expect more work, with 50% expecting the same amount.

We break this by employment type in Fig. 6. Lets call the difference between percent who expect less work and percent who expect more work "the decline". We see that industry people expect a decline about 11%, and students expect a decline twice as large - 22%. Surprisingly, those in academia actually expect more work. Perhaps some of them will work on coronavirus data.

Poll Coronavirus Work Amount by Employment
Fig. 6: How work online has changed because of coronavirus, by employment type.

Finally, we break down the expected amount of work by region.

The regional breakdown was
  • Europe, 33.5%
  • US/Canada, 30.8%
  • Asia, 22.1%
  • Latin America, 5.4%
  • Africa/Middle East, 5.0%
  • Australia/NZ, 3.2%
Poll Coronavirus Work Amount by Region
Fig. 7: How work online has changed because of coronavirus, by region.

We note that 30% of European and 25% of US/Canada respondents expect less work, while significantly higher fractions expects a decline in Asia, Latin America, and even in Africa/Middle East which has not yet been hit by the virus, but is already feeling the big economic impact.

The 5 countries with the largest response were
  • US, 26.7%
  • India, 12.5%
  • Germany, 6.4%
  • UK, 5.7%
  • Canada, 4.1%
Poll Coronavirus Work Amount by country
Fig. 8: How work online has changed because of coronavirus, by country, top 5 countries.

We note that around a third of India, UK, and Canada-based KDnuggets readers expect less work, compared to only 25% of US-based KDnuggets readers.