The CoronaBear Index: Monitor Anxiety on the Stock-Market resulting from the spread of the Coronavirus

Published by newspill Team | 2020-02-09

After the first cases of the Coronavirus were reported on the 10th of January 2020, American Companies highly exposed to China entered a “bear market” (as opposed to the unstoppable “bull market” that we witnessed in 2019).
Indeed, in the global effort to fight against the spread of the disease, several companies have had to shut down or slow down their operations in China.

In this study, we tried to synthesize a compound tracker that mirrors the impact of the Coronavirus outbreak through various industries including:

  • Transports as most airlines had to stop flights to and from China (e.g. United Airlines, Delta Airlines, etc.)
  • Restaurants for obvious health reasons, both for employees and clients (e.g. Starbucks, McDonald’s, etc.)
  • Entertainment as Parks and Casinos are shutting down to avoid contamination (e.g. Disney, MGM Resorts International, Las Vegas Sands Casino, etc.)
  • Industry as plants are shutting down and people staying home until further notice (e.g. Nike, Apple, etc.)

We created the CoronaBear Index to track the severity of the market reaction. We analyze the trend of this index to understand the overall rise and fall in anxiety that can be, to some extent, attributed to the Coronavirus (as the rest of the U.S. markets keeps rallying on positive 2019 earning reports and great job numbers).

Methodology & Modelization

Training Phase

We trained 3 different models:

1. Trend Baseline Model (Bullish)
This model analyses the trend of the index before the coronavirus caused its first death and assumes a similar trend in the month after the virus killed the first individual. This helps in creating a baseline trend with which we compare to the trend after the news of the virus. A simple ARIMA model is used to give us a general idea of the trend of the index.

2. Deviation Model
Due to the effects of the virus throughout the market, we observe a deviation from the forecasted performance. We observed that this deviation seemed to be increasing at a rapid rate in the earlier days of the virus, eventually reaching a saturating value of deviation. In this way, we can use this model to indicate whether the market is overreacting or is reacting “normally” with the number of deaths. We fit a curve using scipy’s optimize functionality.

3. Fatality Model
Another task is to determine the number of deaths caused by the coronavirus in the future. Based on the current pattern of deaths observed on a daily basis, the Fatality Model can predict the number of deaths on a given day based on the current rate observed. We use a polynomial fitting provided by numpy.

Testing Phase

Here, P is the Fatality Model which gives us an estimate of the number of deaths ‘x’ days from the first death caused by the virus. G is the Deviation Model that takes the number of deaths as the input and outputs the expected deviation in the market. Then we check how far off the actual deviation is from the expected deviation and if it’s outside the confidence interval, we flag it as the market overreacting.

If we observe overreaction, this could be of two types:
1. The index is moving further away from the forecast (thus causing a higher deviation) at a rate faster than the expected rate
2. The index is moving closer to the forecast (thus causing a lower deviation) at a rate faster than the expected rate

An example of 1) can be observed just before reaching 300 deaths, when the market was slightly overreacting as the deviation accelerated.
An example of 2) can be observed after reaching 450 deaths, when the market was recovering way faster than expected, even though the death toll kept increasing.

Key Take-Aways:

  • From the graph (figure 3) it can be easily observed that the market has a tendency to overreact in the early stages of the news, causing a large deviation from the forecasted trend (panic stage).
  • As the news settles in and companies, as well as authorities, start making tactical moves to deal with the harms of the virus, the deviation seems to saturate/reduce (acceptance stage).
  • However, one might point out the over-optimism of the market recovering very quickly despite the sharp acceleration in the number of cases, not even mentioning that impacted Companies’ operations are still mostly in a “wait & see” position.
  • We could expect that the market in the next coming weeks will witness another “Bear” episode, bringing the index back within the deviation model confidence interval until the economic consequences are perfectly measured - and the virus eventually contained or cured.
Data Sources:
Market Data from AlphaVantage
Social Networks Data from Stocktwits
CoronaVirus Death toll data from Wikipedia

newspill is currently in Beta-testing on Facebook Messenger

Or simply type "" in the
search bar of your Messenger App.