The Rational Foundations of Irrational Herding in Financial Markets

Herding behavior in financial markets, the tendency for investors to follow the actions of a larger group, is often dismissed as a manifestation of irrationality. However, a deeper analysis reveals that this phenomenon can arise from the aggregation of individually rational decisions. This article explores the theoretical models that explain how and why rational agents might choose to ignore their private information and join the herd, leading to market outcomes that appear disconnected from fundamental values.

The Bikhchandani, Hirshleifer, and Welch (BHW) Model

The seminal model of information cascades was developed by Bikhchandani, Hirshleifer, and Welch (1992). It provides a simple yet effective framework for understanding how herding can occur even when agents are fully rational. The model assumes that agents make decisions sequentially, observing the choices of those who acted before them. Each agent also has a private signal about the true state of the world, which is noisy but informative.

The Core Logic:

An information cascade occurs when it becomes optimal for an individual to follow the behavior of the preceding individuals without regard to his or her own private information. This happens when the public information, inferred from the actions of others, outweighs the agent's private signal.

Let's consider a simple binary choice: to invest in an asset or not. The asset's true value is either high (V=1) or low (V=0), with equal prior probability. Each agent receives a private signal, either "High" or "Low". The signal is correct with probability p > 0.5.

A Numerical Example:

Suppose the first agent receives a "High" signal and decides to invest. The second agent also receives a "High" signal and also invests. Now, the third agent comes along. Suppose this agent receives a "Low" signal. What should they do?

• The third agent knows that the first two agents invested.
• The third agent can infer that the first two agents likely received "High" signals.
• The public information is now two "High" signals.
• The third agent's private information is one "Low" signal.

Using Bayes' rule, the third agent can calculate the posterior probability of the asset's value being high, given the available information. The probability of two "High" signals and one "Low" signal if the true value is High is pp(1-p). The probability of the same signals if the true value is Low is (1-p)*(1-p)*p. Since p > 0.5, the former is greater. Therefore, it is rational for the third agent to ignore their "Low" signal and invest, following the herd.

The Fragility of Cascades:

A key insight from the BHW model is that information cascades are fragile. They can be easily broken by the arrival of new, strong public information. For example, if a public announcement is made that definitively reveals the asset's value, the cascade will immediately dissipate.

The Role of Payoff Externalities

Another rational basis for herding arises from payoff externalities. In some situations, the payoff to an agent's action increases with the number of other agents taking the same action. This can be due to a variety of factors, such as network effects, liquidity, or social validation.

Formula for Payoff with Externalities:

The payoff U for an agent i taking action A can be modeled as:

U_i(A) = V(A) + E(N_A)

Where:

• V(A) is the intrinsic value of action A.
• E(N_A) is the externality benefit, which is a function of the number of other agents N_A who also choose action A.

Example in FX Markets:

Consider a currency trader deciding whether to go long or short on a particular currency pair. If a large number of other traders are also going long, this can increase the liquidity of the long side of the market, making it easier to enter and exit positions. This liquidity benefit is a form of payoff externality that can encourage herding.

As the table shows, the total payoff increases significantly as more traders join the herd, even if the intrinsic payoff remains constant.

Reputation-Based Herding

A third rational explanation for herding is based on reputational concerns. This is particularly relevant for professional fund managers, whose compensation and career prospects depend on their perceived skill.

Scharfstein and Stein (1990) developed a model in which fund managers may choose to mimic the investment decisions of other managers, even if they believe those decisions are wrong. The logic is that if a manager deviates from the consensus and is wrong, they will be judged as unskilled. However, if they follow the consensus and are wrong, they can blame the collective failure of the group.

Actionable Example for Traders:

A trader should be aware of the potential for these rational herding mechanisms to influence their own decision-making. One practical step is to develop a disciplined trading process that emphasizes independent analysis and risk management. This can include:

• Pre-defined entry and exit criteria: This helps to avoid impulsive decisions based on market sentiment.
• Position sizing rules: Limiting the size of any single position can mitigate the impact of being caught in a herding-driven market move.
• A focus on long-term fundamentals: While short-term market movements can be dominated by herding, long-term returns are more likely to be driven by underlying economic fundamentals.

By understanding the rational foundations of herding, traders can be better equipped to identify and navigate these challenging market dynamics.