Causal Factor Analysis : The Science beyond Science !!
By Pankaj Mani, CQF, FRM, AIFI
Head, CQFI Quant Community- India (Delhi)Society
GARP India Chapter
MD,RealWorldRisk
Manipankaj9@gmail.com
Note : Views expressed are personal
Factor Investing though popular needs to be looked at scientifically . Least Square Based Linear Regression establishes uperficially associated factors that need to further examined if they have really causal Scientific mechanism or not. Those associational aspects must be rejected in case of lack of causality .
Infact they exist causal factors that affect the stock price movement. They are applied as well .
History Vs Future : Real Test of Linear Least Square Regression in Time Dimension : Ignorance of Time in Statistics in Real World Applications of Finance.
History is not Future and Such Regression type analysis should be analysed in forward direction of space-time to ascertain causality mechanism. Association based factors are exposed to Black Swan Tail Risk as Investing based on historical superficial associational factors without causality can explode anytime.
Linear Least Square Based Regression Analysis should be tested on Future Data rather discovering on only Historical Data . Dimension of Time is important and so is Direction. The Real Test of Regression Analysis is in Forward Direction of Space-Time.
Linear Least Square based Regression is often done on Past Data, which is itself deeply misconception. To establish causation, we also need to establish how stable that behaves in future. How error terms behave ,which have lot of hidden information about the underlying relationship which is often ignored to establish association based relationship.
Otherwise, there can be number of false Correlation based associations. for example Let’s say there have been accidents on the road, but if we regress on the number of car accidents and number of red cars ,there might be Correlation which could be illogical or say water rainfall on Mars to road accidents on Earth.
It mostly works when we have thin-tailed data otherwise there can exist multiple line of fit having different Beta. Beta and Residual error are mostly misunderstood and misapplied.
Like Physical Space-Time , we have Economic Space-Time where a financial variable traces geometrical paths. They have a causal trajectory in that high dimensional Space-time, which needs to be found out.
Causal Falsifiable Factor Mechanism of Stock Price Movement
Behind the stock price movement , there is definitely a cause .They can be both Qualitative and Quantitative which should be looked at. There are multiple causal factors governing the dynamics .
Example :
If a company has good cash flow,profiting , product, it will create the demand factor for them eventually leading to its share price rise of that exceeds the investor’s expectation. There is a causal chain.
There is definitely the basis of causal figures as to how certain factors cause stock prices to go up .
There are hundreds of associational factors but few are causal based factors that need to be established while rejecting the false associational ones based on least square regressional ones.
For example, Why Value factors ,Momentum factors work or don’t work . How do they affect the demand supply chain to ultimately affect the price movements. Infact in Financial world , investors behavior matter a lot ,that not necessarily be rational or correct always .
Investor Human Behavior dynamics also affect the Stock Price Movement . We also need to model that emotional herd behavior scientifically.
Say for example, if we popularize a sentiment about some factor XYZ endorsed by Some Global Leader, it suddenly creates demand for that factor in investors’ herd (despite the fact that logic might be incorrect) or say like we saw Celebrated Authority endorsing some medicine for Covid, everyone started buying and storing leading to surge in demand and prices going up the stock price of the related Pharma company. (even if that might not work effectively)..once results came out not as expected, demand came down leading to price drop as the burst.
Few General Examples of Causal Mechanism behind Factors :
There exist causal factors to stock price movement which are applied by select few managers for their own use though they might not disclose publicly!!
Say for example Low Volatility of Fundamental Ratios in Stock Price is one of the causes of Low Vol of Stock Prices That’s one of the original causes among many other possible causes..
Or say Value factor low book to market prices or low price to earning ratio could lead to mean reversion of prices driven by investor’s behavior due to cheapness.
And that mean reversion could have different dynamic wavelengths that might explain the variability in the performance cycle. That cheapness might induce momentum to drive the demand due to herd behavior of investors to push the price up and then like physical mean reverting with dynamic drift, investors try to mean revert due to excessive movement of prices. It then consolidates . So, definitely there could exist deeper mechanisms that drive the stock prices based on certain factors.
Let’s take an example of ESG factors and let’s relate this to our day to day experiences . If a firm is ethical, having good social attributes, well governed, well reputed, taking care of environment, it definitely affects our company performance in the long run. That long run has its own dynamic time wavelength. May be that effect could be visible in 1 year or 5 year or 10 years. It can’t always be instantaneous. But definitely has some reasons though it might take time depending on other factors as well.
Similarly there do exist multiple causal based factors of stock price movement.
Infact Stock price is affected by the resultant of all the possible confounding causal factors not by false associational factors..
Say Stock price is genuinely affected by External and Internal Causal Factors GDP,Inflation Supply Demand Mechanism Factors, Investors’ Herd Behavior,firm’s profitability, product, policies, cash flow etc. .
The resultant of all factor causes the stock price movements at a particular moment in time like we have in say Newton’s Laws of Motion in Physics .
Infact ,at advanced scientific perspective, the motion of stock price is Geodesic in economic/financial space-time like the motion of planets in the physical space-time . But the Economic space-time is way advanced than physical space-time as it involves human investor behavior as well which doesn’t remain constant always.
Taking example of our own life : Let’s say X is an inteligent student but say due to family circumstancesz he/she couldn’t study well for few years. So, if investor/ professor realizes that X has potential value despite his poor performance due to other personal causal factors.and over the time , he/she will come up maybe it takes few years or months depending upon various other circumstantial factors. So, there are multiple sub factors affecting the functioning of the Student becoming Successful
Say we can have such student having values, who didn’t do well academically well during initial years but finally is a well successful person over the years as he or she had hidden value that reached its true level over the time .It might take longer time . We see that in our day to day lives around us.
Causal Mechanism exist in Economic / Financial Space- Time way beyond Physical Space- Time.
So, causal mechanism is always there and definitely factors enthusiasts must give the reason and logic and causal falsifiable mechanism but that would be dynamic and variable unlike straightforward like physical laws as human laws are beyond physical laws infact harder than the earlier.
Economic Space Time is why beyond Physical Space time.
The trajectory of financial or economic variable traces Geodesic in its own Economic/Financial space-time. Here the trajectory of Economic Space-time Curvature is determined by the Causal Factors not superficial spurious associational factors. That’s way more complex and advanced science beyond Space-Time in Physics ! It’s Science beyond Traditional Physics !
So, what I mean is that definitely the causal mechanism exists and investors behavior mechanism affecting the demand supply chain leading to price movements...but there is a fundamental difference between hard science and finance including human behavior. Or say event happening in Physical Space-Time and Economic Sapce-Time.
Difference Between Hard Science & Economics & Financial Science.
Economical Space-Time Science beyond Physical Space-Time.
Key Point :
Can Traditional Least Square Linear Regression Analysis capture Investor Behavior led mechanism and Randomness effectively ?? Question to Explore !!
In physics if we experiment time and again, we end up the same result but in finance/economics driven by human behavior, that may not be the case always. That poses a challenge as well. There are multiple variable factors that might be at play same time including Investors’ herd behavior.
So, relying on Regression analysis again to test those relationship needs to be taken with care. Or will the human behavior act similarly in every experiments ! Some Randomness do exist as well. So, need to be careful if these Regression Analysis really can capture those effectively!!
There would be multiple causal factors affecting the human investor behaviour and herd behavior.
It’s resultant of those factors external economic to internal company centric to investors behvaior that comes into play effectively.
Definitely, there exists some scientific mechanism but studying human and investor’s behaviour is possibly beyond traditional sciences like physics !! It definitely has its own science but in economic space-time, it’s way more complex than the normal physical space-time.
In a nutshell , proper causal factors based analysis is indeed very useful to study the mechanism of stock market performance. Traditional Factor based analysis needs to be causal rather than superficial associational. It indeed helps greatly to look at the things in a scientific approach.
Happy to discuss further in detail.
