The Interweaving of Price and Volume in Factor Investing: A Complete Guide to Price-Volume Relationships
There are already a vast number of technical indicators derived from price sequences, and the complexity of comprehensively considering the relationship between volume and price has increased by more than one order of magnitude. From word-of-mouth mantras about the relationship between volume and price, to high-dimensional deep time-series neural networks, from monthly long-term volume and price trends, to instantaneous impacts on the order book, people are endlessly exploring the relationship between volume and price. This article systematically analyzes some representative volume-price relationship indicators from the perspective of factor investment, hoping to provide reference and inspiration for mining volume-price factors in this area.
Classic indicators
SemiBeta
Since the CAPM model was first proposed, \(\beta\) has been one of the most important pricing factors. \(\beta\) represents the reaction of individual stocks to the overall market trend:
$$ ER_i = R_f + \beta_i(ER_{market}-R_f) $$
Among them, \(ER\) is the expected rate of return of the risky asset, \(R_f\) is the risk-free rate of return, and \(ER_{market}\) is the expected rate of return of the market as a whole. The larger \(\beta_i\) is, the greater the elasticity of individual stock returns. Without considering risk, since the market return expectation is positive, stocks with higher elasticity can obtain higher returns. This simple model cannot fully explain the complex market, and many new explanatory variables have been added to the right side of the equation, but the position of \(\beta\) remains unshakable.
Nowadays, the classic \(\beta\) factor also has a new understanding-under different market environments, people should have different views on the elasticity of individual stocks. In the field of behavioral finance research, “Prospect Theory” points out that investors in reality are more characterized by bounded rationality. When faced with the same amount of gains and losses, they believe that losses bring more disutility, which is manifested as “loss aversion” rather than “risk aversion” (Risk Aversion). Therefore, in the face of profits, investors show risk aversion and pursue “making a profit”; in the face of losses, investors show risk preference and pursue “turning losses into profits.” For traditional Beta factors, assuming that investors only abhor fluctuations that may lead to losses, then the relevant risks should only focus on the negative return part. If investors only focus on downside risk, then the covariance of individual stocks and the market as a whole in different states of ups and downs should not be priced in equilibrium.
Based on the above research background, Bollerslev (2021) proposed the SemiBeta concept, which disassembles the traditional Beta factor into four situations based on the different directions of individual stock returns and market benchmark returns, namely
- Market decline + asset decline
- Market falls + assets rise
- Market rise + asset rise
- Market rise + asset fall
Beta factors under different circumstances have different risk premiums. The empirical results of A-shares are not the same as those of U.S. stocks, but discussing according to positive and negative yield rates is still an important improvement idea, and will be widely used in the following article.
RSI
RSI (relative strength index) is a classic technical indicator proposed by Wells Wilder in his famous book “New Concepts in Technical Trading Systems” and is used to evaluate the strength of long and short positions. RSI = increase/(increase + decrease). We can calculate the RSI indicator of the last N days as a stock selection factor. The factor is significant and the direction is negative, but the performance is relatively average. Guosheng Securities proposed to use trading volume to match RSI (that is, use daily trading volume to make a weighted average of RSI) and conduct high-frequency calculations, which greatly improved the factor effect. For details, see “How to Construct Effective Stock Selection Factors Based on RSI Technical Indicators?” “
APB
In “Measuring the Buying and Selling Pressure of Stocks Based on the Volume-Price Relationship”, Orient Securities constructed the APB indicator: the simple average of vwap in the past N days, compared with the volume-weighted average of vwap, and finally took the logarithm. The logic behind it is that long-term allocation funds determine the buying and selling direction based on prior research and judgment, and buy at low prices within the range. This will result in large trading volume when the price is relatively low within the range, and small trading volume when the price is relatively high. On the contrary, when rational investors sell, the trading volume will be large when the price is high, and small when the price is low. The author cleverly unifies the two through the average transaction price deviation, which more essentially depicts the reaction of rational funds to price than simply calculating the correlation. In addition, the APB factor is applicable to both daily and minute lines.
Volume-price correlation
Daily channel
A summary of situations such as “increasing volume” and “declining volume” that are common in technical analysis can all be attributed to the linear correlation between the price sequence P and the trading volume sequence V. If we simply calculate the recent N-day PV correlation as a stock selection factor, the direction is negative and the effect is not satisfactory. This means that volume and price move in the same direction, which means negative profit expectations. Among them, “increased volume” can be explained by short-term speculative behavior - bookmakers gradually sell out chips in fierce exchanges of hands, and speculation comes to an end. “Reduced volume” may be due to small market differences, insufficient buying pressure, and will continue to bottom out in the future. On the long side, it is consistent with the internal logic of APB.
On the basis of this PV correlation, make a difference between the two sequences, and then calculate the DP-DV correlation as a stock selection factor. The direction is negative and the performance is not bad. This can be summarized as if the increase in trading volume is highly correlated with the rise, the future excess expectations will be negative, which also describes speculation behavior.
High frequency
For intraday minute lines, PV correlation does not have significant stock selection ability, so the DP-V correlation and DP-DV correlation after difference are considered.
DP-V
|DP|-V correlation is a significant negative factor, which means that if large intraday price changes are accompanied by high trading volume, then future return expectations will be negative. Consider the staggered correlation of DP-V: the “price before quantity”/“volume before price” factor has a more significant negative stock selection ability. If large price changes and high volume follow each other, the future return expectations will be negative. However, the IC of the staggered factor is mainly contributed by short positions, and the long returns are not significant.
DP-DV
|DP|-DV has a positive factor direction during the same period, with excellent monotonicity and stability: the price change range is positively related to the marginal change in trading volume.
The situation with staggered factors is more complicated.
- Price comes before quantity
- DV>0, DP>0: The factor is negative. When the price rises, the trading volume growth follows the price rise.
- DV>0, DP<0: The factor is positive. When the price drops, the trading volume growth rate follows the price drop.
- Quantity comes first, V-|DP|
- DV>0: The factor is negative. When the transaction volume increases, it is hoped that the price change range will be in the opposite direction to the transaction volume change.
- DV<0: The factor is positive. When the transaction volume decreases, it is hoped that the price change range will be synchronized with the transaction volume change.
High frequency price autocorrelation
P-DP correlation
- dp>0: The factor is negative, and the price increase is at a high price level
- dp<0: The factor is positive and the price decline is at a low price level
|DP|Autocorrelation is a negative factor, similar to fluctuation aggregation.
High-frequency trading volume autocorrelation
High-frequency trading volume V autocorrelation/DV autocorrelation are both negative factors, which seem to be related to the short-term speculation of stock prices. IC is mainly contributed by short sellers.
V-DV correlation
- DV>0: The factor is negative, the trading volume is rising at the high level of trading volume
- DV<0: The factor is positive, the trading volume rises and falls at the low level of trading volume
Summarize
The relationship between quantity and price is complex and beyond human control. From the perspective of factor investment, a reasonable strategy should be to mine significant factors from rich data, or construct factors based on artificial logic, and finally use a nonlinear model to fit the relationship while ensuring that the factors are effective.
The Interweaving of Price and Volume in Factor Investing: A Complete Guide to Price-Volume Relationships