Coronavirus - Not Good People Analytics - For Leaders of Companies in the U.S.

Covid-19 arrived in North America on the U.S. East and West coasts, and just within a few weeks, Covid-19 has been discovered in 49 of 50 U.S. States.

From: https://www.cnn.com/2020/03/03/health/us-coronavirus-cases-state-by-state/index.html

On the day I decided to write this article I had just been informed the Covid-19 virus has been confirmed in Springfield, Missouri, where I currently live. Springfield is not a major city. Springfield is geographically dead-center in the U.S.

Covid-19 is highly infectious. You can forget about the idea of containing the virus outside of the U.S. The virus is already here and spreading, meaning it will go everywhere that people go. There are still airplanes, trains, buses, and cars traveling between cities all over North America.

At this point, the job to be done here is to try to use the data collected on what occurred in other places to try to predict what will happen where you are, when, and make better decisions based on this information.

Now that the virus is here, what should you expect and how can you plan?

The real headline is that what is happening in Europe right now will be happening in the U.S., in just days, and you need to know what that will look like so you can make the right decisions now before it is too late.

A person by the name of Al Chen, someone I am connected with indirectly via the ex-googler network, is updating a document with graphs and tables that provide some useful data views, tools, simulations, and resources on the Coronavirus data. I'm using that as my starting point and layering on additional graphs and illustrations where these are helpful.

<Technical Stuff> Al put the data in Coda. Coda is a new cloud-based document application with some interestingly useful features. The benefit of Coda is that it is dynamic, it has text and data and it can be updated more frequently. I will do my best to interpret here, but since this article is static, this interpretation will fall out of date over time, relative to the source. More about Coda at the very bottom of this post. I'll also see if I can do some cool stuff with Coda, and share that as well, but I wrote this here for speed and for the benefit of Linkedin's content matching algorithm.

Scientists are using data like the ones in Al's dashboard/report/document to try to understand and get in front of the virus. Journalists are using this same data to communicate. Leaders are using this same data to plan.

Al has produced graphs and tables to try to figure out some of the nagging questions that you need to answer to predict the behavior of the virus over time. These graphs are also a useful thinking device for how the cases are likely to distribute across our hospital system over time, and what impact that will have on the quality of care given different scenarios. All of this can be done by applying learning from other locations lines, determining which line the location you are interested in is closest to, and then projecting forward the behavior of the line for purposes of planning. You can make these estimates visually or mathematically.

The shape of a virus - different strokes for different folks

The first illustration I am going to show you is not the coronavirus (Covid-19). The illustration below is the Zika virus.

What I want you to see here is that the rate of infection of the same virus can be different in specific locations based on a number of conditions. This previous research on virus spread illustrates both the challenge as well as the path forward in predicting what will happen as a result of a virus. You will learn further into this article that it also has something to say about what you should be doing sooner (now), as opposed to later.

The above graph is from the Spread of the Zika virus in the Americas.

To set up an equation to determine how many cases you will have in a location (continent/country/state/city) at a given time, you need to know the date the virus entered the location, how many people are there, the shape the line follows, and then some other factors that can influence the line. You don't necessarily need to know all the variable influences because you can follow the example of other locations by observing line trajectories.

Of course, there are some important implications from the distribution of the population at each location, the hospital capacity, and the actions you take on the route of the virus progression. I will speak to these differences below. To operate with the correct assumptions you will need to figure out what the characteristics of the virus are (infection rate and death rate), which other locations you are most like, and when the virus reached 100 cases. If you can determine these variables you can get a good sense of what is likely to happen at your location in the next 7 to 21 days.

Al's Covid-19 Coda document is just the facts without much explanation. It is visual and it includes supporting data tables. My article is theory, evidence, explanation and interpretation. I am providing the context for the data you are hearing about and extending from this, hopefully, insight and clarity for decisions.

I grabbed some helpful graphs and illustrations from other sources. I cite those sources wherever I use them and some of the best ones at the bottom of this article. I'm happy to amend anything I have done if you find anything incorrect, or if you have anything better that I should use or if I have not appropriately attributed a source. If so, get me a note.

Twighlight

Below are some current Covid-19 graphs from Al's data. I'll explain first, and then below, you can look at the data for yourself.

I have grabbed three graphs and filtered them for just Italy, South Korea, France, Germany, and the United States. The virus is in other countries, but I was trying to make it easy to read. In Al's live Coda document, you can apply the filters you want.

To start out I have grabbed three graphs: 1.) Days Since The First Case, 2.) Days Since The Location Outbreak (100 confirmed cases) and 3.) Confirmed Cases Over Time. More data follows those.

Combined the three graphs begin to give you a picture of how the virus appears to begin really slowly before suddenly accelerating. They also give you a picture of how the lines can take different trajectories from there. If you are predicting what will happen in a specific location then you need to understand: 1) the possible trajectories, and 2.) what conditions help you determine which trajectory it will likely be. You may be surprised to learn that you only need a small amount of data to make fairly accurate enough forecasts for decisions.

In the graphs below, height is the number of cases and width is time. The colored lines represent different countries, labeled at the bottom of each graph. Notice that these lines in different countries are accelerating in different ways. There are reasons for this. Keep reading, and I'll tell you more about this.

Covid-19 Coda document

In the graphs above, observe how much time it takes for the virus to get going from the first contact. You don't know what is happening at the point you get the first few cases. Most people react to the first few cases with the thought, "This is just a few cases. This virus is meaningless." Well, it is not until 14-21 days that you observe what these few instances mean. By then it is too late to take the actions you should have taken before. The numbers begin to accelerate dramatically, and when you can finally see the acceleration, there is less you can do, and everyone is in a real panic. You can't see what a few cases were trying to tell you until later.

I should have panicked sooner!

The number of the population infected in a location that has recently acquired the virus should accelerate something like the path of those examples that proceeded them, but lag by a certain number of days. Call the number of days the actual knowledge of the viruses local spread falls behind reality (Lag). (Lag) can represent the currently unknown amount of days until the virus will be recognized in your country, state, city or neighborhood as a serious contagion with commensurate emotion. (Lag) can be estimated mathematically with only a few points of data. Your understanding of (Lag) can become increasingly accurate as you accumulate more data points. The more you understand (Lag) the more ahead of things you can be at your location.

The horror of (Lag) is that the virus is spreading, and you can't see how you should have felt before until later. Without a working (Lag) model, you don't know the right way you should feel at a particular moment in time until later. Pay attention to this. (Lag) is something you need to know so that you can properly interpret the limited information you have.

The line for your location will look like the line of one or more of these other countries, but you don't yet know where you are on the chart, and if your curve will be steeper or wider (prolonged over longer periods of time) than the examples.

Note differences in the curve between South Korea and Italy below to see what I mean.

See how Italy is accelerating, and South Korea is declining—more about acceleration and how it can be controlled below.

Keep in mind that the graphs above are only from February 3 to March 13, and the virus didn't make it to some of these countries, including the U.S., until a week or two ago. That is why all these countries' lines are short and ugly. It is like you have taken a picture, but only 1/10 of the person you are photographing is in the frame, and you are trying to use what you have to guess what the rest of them is going to look like.

Metaphorically you could also say you are looking at a little green seedling and asking yourself if this thing growing on the side of your house is just going to be a poison oak bush or a towering oak tree. Once you know that you want to know if it will be fully mature tomorrow, next week, or the week after that. There are strange physics in this place you have entered.

The power of compound interest and other growth functions

Pay attention to the acceleration of cases over time. In particular, watch out for acceleration that looks like an exponential growth function. By this, I mean the instances double, then that doubles, then that doubles... Growth is expressed visually in the vertical gain of the line. Imagine a poisonous spider is jumping up steps. It is an exponential function if every time he jumps, he can double the number of steps he can jump over. This virus spider can get jumping up this curve very quickly.

Right now the total growth of the virus looks like this:

The above graph illustrates an exponential growth rate. That should be concerning.

Here is another way of thinking about how exponential growth works. If you start with a penny and double it every day for 30 days, on the last day, you would have 1,073,741,823 pennies. It seems I have never been good with money, or maybe I am, but I have never had much of it to work with. In any case, if you gave me a penny, and I could find a way to make my pile double every day, then after 30 days, I would have $10,737,418.23.

One penny is about how much money I have right now, but I am entrepreneurial, so I'm optimistic. Send me your people analytics projects, but please, not all at once. Can you see me rubbing my hands together right now? No, that's not for the money! That's just the hand sanitizer.

The virus has already been discovered in 49 of 50 states. I am not sure I can project accurately from these data yet, but an educated guess it will be out breaking (+100 cases) in most major U.S. cities in somewhere between 7 and 21 days. I think I can say this with high certainty. Add more days, and there is increasing confidence.

The bad news good news of 100 confirmed cases per city

First the bad news. At the stage that the number of observed cases in or near a city exceeds 100 people and/or two or more people have died, the virus becomes increasingly difficult to contain because there are so many unobserved cases, spreading exponentially. The only way to slow the virus at that point is to quarantine everyone in or near that city.

The good news. At the stage that the number of observed cases in or near a city exceeds 100 people and/or two more people have died the spread of the virus becomes increasingly easier to forecast. You use an exponential growth function if no containment measures are applied and you use the Hubei growth function if extreme containment measures like they did in Hubei are applied.

We are much better off applying containment efforts before that happens - e.g. in the next 7 days. So there is a fairly high likelihood that the U.S. federal government will see the same data I am looking at and either advise this or order it. If you don't decide to do it yourself the government will have to.

It is worth noting that the number of possible cases decreases/diminishes with the exhaustion of potential cases as it runs through the availability of the total reachable population. The growth of the totals will accelerate and then start to back off in places. It hits a barrier; it backs off; it gets to the other side; it accelerates; it runs out of people that are easy to reach; it backs off. When it finally runs out of people it can reach, then it finally stops. This is why it is called a "Pandemic," a critical mathematical honor for the virus. We haven't really seen this since 1918.

How fast is the is virus doubling in different places?

As of last edit (3/14/2020), it appears Covid-19 is doubling every three days in the U.S., but in a given location it can double every day, and even faster.

<Remember> Smaller numbers are worse! Smaller numbers mean the virus is spreading faster.

Based on the chart above - from the data on March 14, 2020 - the rate of spread of the virus in Mainland China and South Korea has drastically slowed. The rate of spread is very high right now in parts of Europe, and the U.S.

All of these numbers will change in the coming days based on the actions that are taken.

At the current rate of spread, the U.S. graph currently looks like this. This is an exponential function. That is what was happening in Italy and Iran just days ago. This is not good.

<Tip> As it pertains to this virus, a comparison to Iran and Italy should be truly terrifying given what is happening now in those countries. We need to take action immediately.

Here is what it feels like in Iran right now

From the Wall Street Journal: Iran’s Coronavirus Strategy Favored Economy Over Public Health, Leaving Both Exposed - Tehran’s efforts to sustain business activity compromised its response to the outbreak, damaging the economy it tried to protect.

The fatalities brought the country’s death toll to 724 so far, amid nearly 14,000 confirmed cases. I realize that out of context those are just numbers that don't mean that much. The take away is that the real number of infections is inevitably much higher and compounding in ways that are difficult for the health ministry to deal with, increasing the severity of the outbreak and of deaths. “If the trend continues, there will not be enough capacity,” Ali Reza Zali, who is leading the campaign against the outbreak, was quoted as saying earlier by the state-run IRNA news agency. Iran is believed to have around 110,000 hospital beds. These will likely be exhausted and many more people will die than we would expect in other places with similar population sizes.

Since they are ahead of us and have enough cases to calculate statistics, pay attention to the Iran details: Many of those who have died from the COVID-19 in Iran were otherwise healthy, a rare admission by a health authority that the virus does not only prey on the sick and elderly. Iran Health Ministry figures show that while 55% of existing fatalities were in their 60s, some 15% were younger than 40.

Some people are comparing this virus to the flu. They are wrong

The death rate of the infection varies based on several conditions, which change over time, but on average, it is showing around 3.7%.

From Al's Coda:

<Remember> This 3.7% the total death rate, which reflects those that are well-served by their leaders and medical system AND those that are under-served by their leaders and medical systems.

The reason why the death rate varies so widely between locations is that if the number of cases spikes beyond hospital capacity, then the people who need help can't get high-quality treatment, so the death rate increases substantially.

To be clear, if the hospital is overcapacity, the death rate in an area can increase from a base rate of about 1% to something more like 20%. 20% means that approximately 1 in 5 people die. Amy, Sue, Mike, Bob, Sal shake hands at a meeting, which of the five is not going to make it home, which?

To set up an equation to determine how many people will die, you need to know how many cases you will have in a location, the hospital capacity, and the death rate from the virus in cities with different levels of hospital capacity. Then you just multiply it out.

<Technical Stuff> All the nagging mathematical details can be dealt with simply. First, per location, subtract expected cases from capacity. This will give you the underserved, call them (under-served). Don't forget the well-served. They are equal to all those cases that can be met by capacity - call these (well-served). In a scenario in which there are more cases than capacity then (well-served) = (capacity). In a scenario in which there are fewer cases than capacity then (well-served) = # of total cases, and (under-served) = 0. You will need to apply a different death rate to type (well-served) and (under-served) to accurately estimate a location death rate.

For now, let's use the assumption that the Covid-19 death rate is 1.5% for the (well-served) and 20% for the (under-served). The death equation would look like this:

Deaths = (well-served) x .015 + (under-served) x .20

Capacity is a mushy term. It can mean doctors, nurses, beds, medicine, respirators, and a lot of things. In truth, you probably need to know some specifics - like how many ICU respirators and experimental anti-viral medications the hospitals in that area have. I doubt we have that data by location. For purposes of estimating our variable (capacity), "hospital beds" may suffice as a reasonable proxy. You can usually find out the number of hospital beds in a location. This becomes useful in per location simulation models, which I and others are working on now.

To pull this conversation back in, just know that the death rate is one thing if the hospital is beyond capacity (~20%), and it is entirely another if they have sufficient capacity to address the need (~1%). The variable nature of hospital capacity will make the total death rate look erratically different when comparing cities or regions. In some of the models, people are using the number of deaths to estimate the likely number of actual cases at a given time, removing lag. Well, if you are using deaths in your model need to adjust everything for the capacity problem I described above.

To forecast you have to identify the number of people in the location, then find the right curve to represent the location, then set a time location on the curve based on (Lag), adjust for (capacity), and you can make an estimate about both infection and death rates at any given time - past, present and future. That's the job to be done.

Covid-19 a lot worse than the seasonal flu

The current coronavirus is a lot worse than the seasonal flu. Below is a graph of how the death rate of this coronavirus (Covid-19) compares to the death rate of other known viral infections, including the seasonal flu.

There are two variables to worry about—the rate of infection and the death rate. See the graph below.

Notice that this coronavirus has a much higher infection rate and a much higher death rate as the seasonal flu. Seasonal flu causes one other person to infected on average per case, whereas this coronavirus infects 2.5 additional people per case. The seasonal flu death rate is only .1%, compared to coronavirus's currently expressed death rate of 3.7%. To be more conservative let's assume that this coronavirus's actual death rate provided adequate care is 1%.

These numbers may not sound like much, but if you combine the differences in the infection rate and death rate, this coronavirus (Covid-19) will be 10 to 100 times more destructive than the flu.

If 10,000 people per year die from seasonal flu it means 1,000,000 or more people could die from this virus.

You may or may not remember the panic over SARS - Covid-19 is a lot worse than SARS already

Some of the graphs above are from here: https://www.nytimes.com/interactive/2020/world/asia/china-coronavirus-contain.html

Covid-19 is highly infectious and truly a pandemic

Below is a screengrab of an interactive image representing confirmed cases of Covid-19 from John's Hopkins Cornovarius Resource Center on March 14, 2020. https://coronavirus.jhu.edu/map.html

The bubble graph is concerning and makes the point - this is bad- but it doesn't do a good job of showing change over time or help with forecasting.

The shape of the growth curve and how it is influenced by decisions

The shape of the per-location growth curve is affected by the geographic distance of the population, and the ability of the government or other coordinated actors to shut off interactions. In the graph below height represents the number of cases and width represents time. The number of cases will get tall faster in places that do not contain the virus well; however, in theory, then it should run its course quicker and not extend as wide over time. The graph is taller in densely populated regions that do not or cannot restrict travel and interaction. The graph is wider in places that are able to restrict interaction. The term people are using is "Social Distancing."

<Remember> I predict "Social Distancing" will the 'word of the year' in 2020.

Why does the shape of the curve matter?

If the graph is short but wide, there will be fewer cases at one time. Conversely, if the graph is tall but skinny, there will be more cases at one time, but the virus may run its course through the population faster.

The graph will be shorter but wider in places that apply more effort to prevent the virus from spreading, and the chart will be taller but skinnier in areas that cannot prevent the infection from spreading.

<TIP> One radical idea is that the absolute number of people impacted may be difficult to control in the end. However, what you can control is when the people show up at the hospital, and that is important for survival.

More people will die in places where the curve is taller because they will not receive sufficient critical care in the time that they need it. If you can spread the cases out over a more extended period, you have more time to get people into and out of hospital beds. A short but wide graph is not good math for the productivity of society overall, but it is much easier on the hospital system.

What we learned in 1918

This section is an excerpt from The Atlantic: Cancel Everything

[ When the influenza epidemic of 1918 infected a quarter of the U.S. population, killing hundreds of thousands nationally and millions across the globe, seemingly small choices made the difference between life and death.

As the disease was spreading, Wilmer Krusen, Philadelphia's health commissioner, allowed a huge parade to take place on September 28; some 200,000 people marched. In the following days and weeks, the bodies piled up in the city's morgues. By the end of the season, 12,000 residents had died.

In St. Louis, a public-health commissioner named Max Starkloff decided to shut the city down. Ignoring the objections of influential business people, he closed the city's schools, bars, cinemas, and sporting events. Thanks to his bold and unpopular actions, the per capita fatality rate in St. Louis was half that of Philadelphia. (In total, roughly 1,700 people died from influenza in St Louis.)

In the coming days, thousands of people across the country will face the choice between becoming a Wilmer Krusen or a Max Starkloff. ]

Three different spread scenarios depending on the action taken

I will show you three different scenarios below, from bad to good. Right now, most leaders in the U.S. can control the outcomes of their organizations and cities by making conservative decisions. They can no longer control these outcomes in 7 days.

In the graphs below, you want to get more cases below the orange line, out of the purple and into the green. You do this by slowing the transmission of the virus through area containment and social distancing.

Scenario 1 - very little area containment and social distancing

What happens if the virus spreads uncontained? The mortality rate is higher than it otherwise would be because of the failure rate of the medical system to address severe cases of the illness, which all arrive all at the same time.

Scenario 2 - some area containment and social distancing

With some late containment and social distancing, you get some hospital's overwhelmed, but it is not nearly as severe as in the first scenario.

Scenario 3 - severe area containment and social distancing

In this scenario, extraordinary measures would have to be taken before the virus is observed in the region (by prediction) or immediately upon the first case. You would likely have much fewer cases, but even if you have the same number of cases, you will get a better result. The hospitals can better treat the cases they get, and therefore fewer people die.

Observe the difference between area containment and social distancing efforts in China versus the first few countries that have come after

Notice in this graph that the more recently infected countries are not following the same pattern as China. China was more like scenario three above, whereas South Korea and Italy are more like scenario one. Also, notice when the virus first showed up in South Korea, Italy, and Iran. Reports of Iran's situation is horrific - genuinely terrifying.

Also, notice when the virus first showed up in South Korea, Italy, and Iran. This is the Lag.

Things like the following cause (Lag): a) the date when the virus first arrived, b) the date when the virus started spreading among the local community, c) the time it takes for the virus to display physical symptoms (gestation), d) the time it takes for people to go to the hospital, e) the time for testing to catch up with reality, and f) temporary measurement error.

Have a look at this graph of the cases in Hubei China to get a sense of how (Lag) works

In the graph above, you can see many things. The first thing I want to point out is that the actual number of cases is not known until later - what I'm calling (Lag). Note the difference between the blue bars and the yellow bars. The blue bars represent the date of the real onset; the yellow bars represent the date of diagnosis. There is no test for onset. Onset can only be determined later, following diagnosis. Do you see the problem? The problem is that the significant period when the virus is doubling, then doubling, then increasing many times again is before a substantial amount of diagnosis' at a location. A single diagnosis in an area should be considered a severe event. Two diagnoses portend and avalanche to follow.

The period where decisions are most critical is when you have the least information. You have to project from a small amount of impartial data that will follow. You have to take it seriously and error to caution. The consequences of not doing so are too severe.

If you were to set up a multivariate equation to predict the number of cases (a model), and if you need to be precise, you should look at cities or neighborhood data and try to see if there were a way to include all those variables that matter. You are putting numbers into several variables in on one side of the equation and then spitting a number out on the other side of the equation that gives you a line over time. Early measurement error, notably the problem of (Lag), will have an outside impact on the overall accuracy of your projection. Therefore you have to find a way to address (Lag). Your degree of precision in dealing with (Lag) will determine the overall accuracy of your result.

A more refined understanding of (Lag) can be made by studying: 1) the time it typically takes from arrival to community spread, 2) the time it typically takes for people to go to the hospital, and 3) the time it typically takes for testing to catch up with reality.

All the errors I described above can be backed out by mathematicians when they have a complete dataset. As they work with its shape of the lines used to describe the virus will become more symmetrical - so the lines will appear both more visually beautiful and predictable, relative to the partial data we have now.

Notice how efforts to contain helped china to flatten the shape of the curve from what it otherwise would have been

The second thing I want to point out from the graph above is that the flatness of China's curve, relative to other countries, is a result of their extraordinary efforts to control the virus.

Currently, in the U.S., we have not yet applied a coordinated effort to restrict the travel of the virus between states and cities (among the U.S.). We also have not yet applied a concerted approach to control the marketplaces, workplaces, and other interactions of our citizens. Therefore the virus should spread tall fast. If this continues, our virus progression will not be better than China; it will be worse. In any case, you will see that the rate of reported cases will grow increasingly fast in the U.S. as you move out of the twighlight of (Lag), and as you begin to see increasing amounts of people showing up at hospitals and increasing tests.

The most important thing you can do to control the spread of the virus is to increase isolation and social distancing BEFORE it is clear the infection is nearby at the city or work location.

Starting social distancing one day earlier can result in a 40% difference in the number of infections.

I'm not alone in the idea that you need to start social isolation now, not later. There is a very well-written piece in The Atlantic that you should read: Cancel Everything - Social distancing is the only way to stop the coronavirus. You must start immediately

What will happen in the U.S. based on different social distancing assumptions?

The New York times put together a really slick interactive chart that dramatically shows the impact of different decisions.

How Much Worse the Coronavirus Could Get, in Charts

You can find these illustrations and change the outcome by moving the time interventions are started here: https://www.nytimes.com/interactive/2020/03/13/opinion/coronavirus-trump-response.html?action=click&module=Opinion&pgtype=Homepage

How is the U.S. doing right now?

Given our geographic location in the world and the distance between our cities the U.S. trajectory started slow, but now it is rapidly accelerating towards the lines of other countries in Europe. We aren't special. You aren't special. Remember Lag?

The above graph is from https://www.macleans.ca/news/canada/could-canada-be-flattening-the-curve-of-coronavirus-cases/

H?ere is a graph from the Financial Times that helps you compare the trajectory of spread with less influence of Lag by showing data only from the day 100 cases were identified:

The people at the Financial Times are smarter than the rest of us. More from them here: https://www.ft.com/content/a26fbf7e-48f8-11ea-aeb3-955839e06441

How well have we measured cases in the U.S. to figure out where effort needs to be applied to contain the virus before it spreads beyond control?

Poorly!

Not only are we complacent because of the illusion of time (we are 7-21 days lagging behind the rest of the world), but we also don't have a reasonable estimate of the number of cases we currently have in the U.S. because of a lack of testing. In other words, more people are infected than you know, and the infected individuals are spreading it. You will not see the extent of the viruses spread until people start showing up at the hospital already in terrible shape and having already spread the virus to many other people.

You should expect:

1.) the virus to begin to stretch hospitals in large U.S. cities in 7 to 21 days.

2.) Unless the government institutes severe containment measures, the infection will spread all over the U.S., and begin to exceed capacity in some places.

3.) The results will be uneven. Different states and cities will have different results based on the size of their population, the degree they can control the movement of people, the capacity of their hospitals, and, most importantly, the decisions they make right now.

There are no do-overs on the actions taken or not taken the next few days.

The hospitals will first start to get stretched in major cities along the coasts, where it first landed, before making its way into the major cities in the middle of the U.S. where it has arrived at more recently - lagging behind those coasts.

To my friends and family. You may want to consider where you are going to be located in the next seven days before these outbreaks are realized, and travel into and out of places is restricted. I have just informed the virus just arrived in Springfield, Missouri, on the date I wrote this article. Springfield is geographically dead-center in the U.S. The entire Springfield metropolitan area and has only 150,000 people. From what I am told, Springfield has excellent hospital capacity for a city of our size. We have at least two major hospitals (Cox and Mercy) and a nearby military base. I believe we are going to be o.k. in Springfield in hospital capacity relative to the relative to the population in and around our city. Springfield doesn't have a major airport hub, but we have three nearby regional commuter airports (Springfield, Branson, and Joplin) that connect through Dallas & Chicago.

What are the ramifications of Covid-19 for commerce?

So far we don't have data from Europe but we do have data from China. Again, the Finacial Times is ahead of the rest of us:

At its peak, the virus caused a nearly 50% reduction in productivity in China. The good news is that it appears that from beginning to end China was able to get people back to work in about two months. Data is still out on whether or not Europe and the United States can recover as quickly as China did. Our early response is concerning. Based on the data we have now, the U.S. situation is likely to progress in a way that is worse than China.

This is admittedly rough analysis, but using the Financial Time's China data, 'finger to wind analysis ' suggests that the world will lose at least two months of productivity, which without recession represents a potentially 20% reduction in annual corporate revenue and GDP.

Looking at what happened in the stock market over the last month it appears the 20% drop stocks have experienced over the last month reflects expected downward revisions in revenue and GDP.

The 21% decline in the Dow Jones Index (~29,000 to ~23,000) fairly reflects the likely impact of the virus on productivity without pricing in a recession.

<REMEMBER> The DAX Friday close doesn't reflect the 10% bump the DOW got on Friday, after the President of the United States started to promise do whatever it took to use the full weight of the US treasury to provide assistance and stimulate the economy - up to and including pushing money out of airplanes if necessary. The DAX was already closed.

To be fair, the 20% loss of revenue assumption does not factor in increases in spending related to the virus (For example, increased toilet paper sales) or a post-virus surge resulting from pent up demand, or new government stimulus efforts. However, because investors don't know yet if Europe and the US will fair as well against the virus as China did, or come back to work as quickly as China did, or the reaction from other investors, or other unknown risks, you should expect continued potential downside and daily volatility for at least a month. In addition to this, individual stocks could experience dramatic changes over the next 30 days as investors use the ebb and flow of the broader market sentiment and bank lending stimulus to try to re-position their portfolio advantageously for revised near-term, and long-term growth expectations.

Corrections and Comments

Feel free to comment at the foot of the article. What have I misstated or left out? How are you thinking about it at your company (or not)?

Resources

The data project I just described is here: Coronavirus COVID-19 (2019-nCoV) Updates & Resources The document I am referencing created on a document surface called CODA. CODA is an interactive document you can embed both text and data into and do magical things with using formulas and code. Let me know if you want me to show CODA to you at some point.

Advice from an epidemiologist: What I think about COVID-19 this morning

What the smart people are saying: Cancel Everything - Social distancing is the only way to stop the coronavirus. We must start immediately (from The Atlantic)

Coronavirus: Why You Must Act Now - Politicians, Community Leaders, and Business Leaders: What Should You Do and When?

Three Optional Steps

My name is Mike West; my Linkedin profile is here: https://www.dhirubhai.net/in/michaelcwest

The People Analytics Community is here: https://www.dhirubhai.net/groups/6663060

I usually don't write about viruses. I typically focus on the topics of people in organizations. An index of my writing is here Index of my writing on people analytics on PeopleAnalyst.com