We hope this article helps beginners, enthusiasts, and academics alike appreciate the beauty of econometric modeling. We found this work on a digital commons network and wish to honor the original author. If you are the author and would like it removed, please write to us at learneconometricsfast@gmail.com.
The Efficient Market Hypothesis (EMH) is a financial theory that holds that asset prices reflect all available information. According to this view, securities consistently trade at their true value because the market processes all available information. As a result, it is considered highly unlikely for an investor to consistently outperform the overall market through stock selection or strategic timing, as any advantage gained from new information is immediately priced in.
The core principles of the EMH can be broken down as follows:
- Information in Prices: An asset’s current price is believed to reflect all available information, from past trading data to public news and, in its strictest interpretation, even confidential insider knowledge.
- The Challenge of Outperformance: Since prices are presumed to be accurate reflections of value, there is no reliable method for securing long-term, risk-adjusted returns that exceed the market average.
- Unpredictable Price Movements: Price changes occur randomly in response to unforeseen news or events, making future trends difficult to forecast—a concept often described as a “random walk.”
- Risk as the Only Reward: In an efficient market, the sole reliable path to potentially higher returns is by accepting a greater degree of investment risk.
The EMH is commonly discussed in three distinct forms, which vary in their level of stringency:
1. Weak Form: This version asserts that all historical market data, such as prior prices and volume, is already reflected in current prices. Consequently, technical analysis, which relies on charting past patterns, is seen as ineffective for predicting future price movements.
2. Semi-Strong Form: This more comprehensive form states that prices instantly adjust to all public information, including financial reports and economic news. This implies that neither technical analysis nor fundamental analysis (the evaluation of a company’s intrinsic value) can consistently yield an edge over the market.
3. Strong Form: As the most absolute version, the strong form contends that stock prices reflect every source of information, including non-public or insider knowledge. This suggests that not even corporate insiders can reliably achieve excess returns, though profiting from such private information is illegal.
In this article by Jackson Dino, Gettysburg College, an attempt is made to study the effect of trading volume on stock price. The article can be accessed from https://cupola.gettysburg.edu/cgi/viewcontent.cgi?article=1032&context=gchq and from the https://network.bepress.com/social-and-behavioral-sciences/economics/econometrics/
Here, the EMH findings are disputed, and the author found a positive correlation between an increase in trading volume and stock prices.
Trading volume is the total number of shares of a security traded within a certain timeframe. It studies whether there exists any causation between the two.
OLS Method – Regression framework and dataset of S&P 500-listed companies from 2013 – 2018. Stock price volatility is captured by control variables addressing the impact through instrumental variables, minimizing heterogeneity. Heteroskedasticity-robust standard errors are used.
They are using 5 models of the following specifications.
Linear model with control variables, fixed effects, and an instrument.
- Stockpricei = β0 + β1 tradingvolume1 + Ɛit
- Stockpricei = β0 + β1 tradingvolume1 + β2 Highlowdiff1 + β3 Prevday1 + β4 month1 + Ɛit
- Stockpricei = β0 + β1 tradingvolume1 + Xi γ +𝛂i +λt + Ɛit
- Stockpricei = β0 + β1 tradingvolume1 + Xi γ +𝛂i +λt + Ɛit
Where, tradingvolume1 = π0 + π1 Zdow + Xi𝛉 + 𝛂i +λt + 𝓥it
- lnstockpricei = β0 + β1 tradingvolume1𠆢 + Xi γ +𝛂i +λt + Ɛit
The first model
Stockpricei = β0 + β1 tradingvolume1 + Ɛit
It is the basic framework, but it lacks the capacity to capture the dynamism in the market. The stock may have industry/sector-specific heterogeneity, or it may have a time effect.
These effects are captured in the rest of the models using control variables and the instrumental variable approach.
In the second model
Stockpricei = β0 + β1 tradingvolume1 + β2 Highlowdiff1 + β3 Prevday1 + β4 month1 + Ɛit
This framework has three control variables added. Higlowdiff1 is the control variable capturing the short-run volatility in stocks. It is the difference between the High and Low price of a stock on a given day. Including this variable takes care of the daily volatility. The variable month1 is a value coded depending on the month of the year, holding the monthly volatility factor.
Prevday1 is a lagged binary variable which is 1 if the closing price is higher than the opening price and 0 otherwise.
In the third model
Stockpricei = β0 + β1 tradingvolume1 + Xi γ +𝛂i +λt + Ɛit
Xi γ holds the three control variables of model 2.
𝛂i represents stock fixed effects, holding constant the impact of the company
λt represents daily time fixed effects, holding constant the impact of time.
These new variables control for the average differences of both observed and unobserved variation across all the companies and across time, minimizing the effect of omitted variable bias in the analysis.
In the fourth model
Stockpricei = β0 + β1 tradingvolume1 + Xi γ +𝛂i +λt + Ɛit
Where, tradingvolume1 = π0 + π1 Zdow + Xi𝛉 + 𝛂i +λt + 𝓥it
They have used the instrumental variable approach.
TradingVolume1 is an estimated instrumented variable, and Zdow is an instrument variable day of the week.
In the fifth model
lnstockpricei = β0 + β1 tradingvolume1𠆢 + Xi γ +𝛂i +λt + Ɛit
This model is a non-linear, log-linear model that captures the percentage change in stock in response to an increase in volume. This non-linearity functions as a robustness check point.
Models 3, 4, and 5 confirmed a positive relationship between trade volume and stock prices. While Models 3 and 4 used stock price levels, Model 5 uniquely analyzed stock returns and incorporated non-linear effects.