The American Stock Market During the COVID-19 Pandemic - The Black Swan of Our Generation

Introduction

The COVID-19 pandemic has been a very tough time for a lot of people - due to lockdowns and other restrictions put in place by government entities the economy has declined and many people have lost their jobs. On the other hand, however, the value of the stock market hasn’t reduced to the same extent as the rest of the economy - in fact, for a large portion of the pandemic, the overall market cap of the US stock market increased.

In the terms of Lebanese-American world-renowned essayist, academic and hedge fund manager Nassim Nicholas Taleb, the COVID-19 Pandemic is a classic example of a ‘Black-Swan Event’ - a term coined and observed by Taleb himself. The definition of a ‘Black Swan Event’ is an unpredictable event that is beyond what is normally expected of a situation and has potentially severe consequences (Investopedia); examples of this include the 9/11 attacks, 2008 financial crisis and others.

Since Black Swan events have a low probability of taking place, businesses, investors, and other stakeholders in the stock market do not take them into consideration when making financial models, so when they inevitably take place, very few are equipped to handle them, leading to increased volatility in the stock market.

The goal of this research is to use data science to analyse this volatility and if possible create some insights which can be used in future Black-Swan events.

More specifically, the main two questions that we are attempting to answer are as follows:

• Which variables had the most influence on the stock market during covid19?
• What effect could future black swan events have on the valuation of the stock market?

Some of the technologies that we utilized during the course of this research include Python (Pandas, Numpy, Seaborn, Matplotlib, ScikitLearn etc.), and R

Hypothesis

The stock market seemed to be extremely volatile and overvalued during this period - this is counterintuitive since the US economy was doing particularly badly. Why was this the case?

Data Collection and Variables

For all of the economic data we are collecting, we are going to be focusing on the timeframe of the COVID-19 Pandemic. Li Wen Ling, a Wuhan ophthalmologist, first alerted people of the COVID-19 virus in December of 2019. We understand that the Pandemic hasn't fully subsided just yet, but we are making the assumption that we're nearing the end of the Pandemic (fingers crossed) - the latest GDP figures are available only up until the last quarter of 2020. Since there are no GDP figures for December, we’ve decided to use the timeline: 10/01/19 - 10/01/20.

The main economic variable that we are going to be tracking is the Buffet Indicator, named after famous investor Warren Buffet. This is defined as the ratio of the value of the total US stock market to the US GDP. It measures how overvalued or undervalued the stock market is at any given point. A Buffet Indicator of 1 indicates that

The values we’re going to be using for the overall US stock market valuation is Willshire 5000 Full Cap Price Index (obtained from Federal Reserve Economic Data), which is an index that tracks the overall valuation of the US stock market and is an industry standard indicator; this indicator is a daily indicator. We also obtained quarterly US GDP figures from Federal Reserve Economic Data.

Some of the other macroeconomic indicators we chose as a part of our economic analysis include:

• Equity Market-related Economic Uncertainty
• CBOE Volatility Index (represents the market's expectations for volatility over the coming 30 days)
• TED Spread (difference between the three-month LIBOR and the three-month Treasury bill rate → used as an indicator of credit risk)
• 10-Year Treasury Constant Maturity Rate (interest rate)
• Effective Federal Funds Rate (the interest rate banks charge each other for overnight loans )
• 10-Year Breakeven Inflation Rate (market-based measure of expected inflation)

Data Preprocessing

Fortunately, most of the Data that we obtained was already structured well for a time series analysis. However, the frequencies of all of our economic variables weren’t always consistent. For example, the Willshire 5000 Full Cap Price Index is disclosed daily, while our GDP figures were disclosed quarterly. Because our economic analysis spanned approximately a year, and given we wanted as much data as possible, we attempted to extrapolate our less frequent variables data to daily variables.

Our main data preprocessing was creating a daily interpolation of the Buffet Indicator. Because GDP is disclosed quarterly, all Buffet Indicator data found on the internet are quarterly figures. Since our aim was to create a daily time series projection of the Buffet Indicator, we decided to linearly interpolate to project quarterly GDP figures to a daily indicator.

We understand that there are limitations to using linear interpolation to project Economic output, but we believed that it was the most effective way: given that GDP figures technically don’t exist between quarters, linear interpolation is an effective method of projection.

Similar steps were taken for other non-daily macroeconomic variables.

Analysis

Before looking at any of our other macroeconomic variables, let’s have a look at the Buffet Indicator during our selected period. Underneath, I plotted the Buffet Indicator and the daily percentage change in the Buffet Indicator for the selected period.

The first observation we can make is that for the duration of the entire period, the Buffet Indicator was always above 1 - implying that the US stock market has been overvalued for the entire duration of the period. This is unsurprising given that it is common knowledge that the stock market has been overvalued for a while now. What is interesting to see though is how the extent of overvaluation changes during the pandemic. From Nov 2019 to approximately the middle of Feb 2020, the overvaluation of the stock market with relation to GDP (the Buffet Indicator) was increasing at a steady rate. Some speculate that this was due to tech stocks, some of them with the largest market capitalization, becoming popular. But this growth was fairly steady and devoid of volatility - the daily pct change was relatively stable around 0 for this period.

The second major event seen in this graph is the large drop between mid Feb 2020 and mid March 2020, when the Buffet Indicator hits its lowest point. This period was also marked by high volatility in percentage change. This drop was fairly significant and also has a good explanation: January was when the first official COVID-19 case was announced in the US, but at this time, it was thought by many that the pandemic would be contained. Only by February did it become clear that the Pandemic would become as large a catastrophe as it was; furthermore, this was also the time when restrictions were beginning to be put in place in several parts of the US. This, coupled with the contradictory rhetoric put out by the CDC, the President and other arms of the government, led to large amounts of economic uncertainty, resulting in a volatile period and a reduction in the valuation of the stock market.

Yet, around mid March, we began to see another surge in the valuation of the stock market. While the period after mid March 2020 was relatively volatile - a lot more volatile than the first period - the value of the stock market did increase again. Most analysts believe that this was, again, a result of a bull market for tech stocks: tech stocks take up approximately 27% of the S&P500 implying that a major plurality of the US stock market is dominated by tech. During these months, it became increasingly clear that tech companies (e-commerce, ed-tech, food-delivery etc.) stood to gain a lot as a result of the pandemic. There was still, however, some level of volatility due to the uncertainty surrounding the pandemic and the US Government’s response to it.

The graph of the Equity Market-related Economic Uncertainty above also helps corroborate the idea that the volatility of the Stock Market during this period was related to economic uncertainty.

Stationarity

In terms of stationarity of each of the variables, after first differencing, and performing the Augmented Dicky-Fuller test on each variable, they seemed to be stationary.

In statistics and econometrics, an augmented Dickey–Fuller test (ADF) tests the null hypothesis that a unit root is present in a time series sample. The more negative the ADF value is, the stronger the rejection of the hypothesis that there is a unit root at some level of confidence. In the tables above, the ADF was done on all the macro variables listed early on in the paper with respect to both the Buffet Indicator and the CBOE Volatility Index. For each of the ADF tests there are 5 values (1 represents economic uncertainty, 2 is TED Spread, 3 is 10-Year Treasury Constant Maturity Rate, 4 is Effective Federal Funds Rate and 5 is 10-Year Breakeven Inflation Rate).

This implies that the values of the macroeconomic variables are not dependent on the time at which they are observed. This is very counterintuitive since it is generally thought that these variables are affected by the Pandemic and other events. Though more thorough analysis is needed to make stronger conclusions.

Vector Autoregression

The VAR model is used to model dynamic behavior of economic and financial time series and for forecasting - multivariate time series regression modeling. Using R’s autoregression we achieved the following results for a autoregressive model for the Buffet Indicator:

The table above implies that there was a relatively strong positive relationship between 10-Year Treasury Constant Maturity Rate and the Buffet Indicator, a positive correlation between uncertainty and the Buffet Indicator as well as Inflation and the Buffet Indicator. There is also a strong negative relationship between Effective Federal Funds Rate and the Buffet Indicator as well as negative relationships between Volatility/TED Spread and the Buffet Indicator.

Extreme Bound Analysis

Extreme Bound Analysis (EBA) is sensitivity test that examines the strength of the association between a dependent variable and a variety of possible determinants. Our dependent variable is the Buffet Indicator. The results are below:

Conclusion

The COVID-19 Pandemic was a textbook definition of a “Black Swan Event”. As evidenced by the uncertainty index during our prescribed period, the period was an event that very few saw coming, yet it had a great effect on the American economy.

One major takeaway from this commentary and analysis was that before the pandemic, the uncertainty in the economy was very low and the stock market was growing significantly - many even believe that it was significantly overvalued. When pandemic struck, the economic uncertainty and volatility of the stock market increased causing an initial panic and a crash. Yet the stock market recovered very quickly - though volatility and uncertainty remained high.

The major consensus about why the stock market was and continues to be overvalued and also why it bounced back very easily is that the overall US stock market is protected by large tech companies which flourished during the pandemic. As a result of the S&P 500 and other index funds containing large amounts of technology stock, the whole stock market seems to benefit from the confidence.

Though this is a hypothesis, perhaps this knowledge can be used again in future Black Swan Events to create strong economic returns. If a Black Swan Event ever does occur in the future, and technology stocks do not seem to be affected by it, there will likely be a large sell off initially while uncertainty is high, leading to lower prices. This is perfect financial opportunity for investors to buy at a low price and wait for Gov officials to get a handle on the situation. When this happens the economic uncertainty may reduce and the Tech industry will protect the rest of the stock market leading to another round of appreciation, albeit volatile appreciation, in the stock market.

Further research is still needed to create stronger forecasting models about stock market valuation during Black Swan Events.

Citations