Keynes Meets Markowitz: The Tradeoff Between Familiarity and Diversification

Working Paper: CEPR ID: DP7687

Authors: Phelim Boyle; Lorenzo Garlappi; Raman Uppal; Tan Wang

Abstract: We develop a model of portfolio choice to nest the views of Keynes - who advocates concentration in a few familiar assets - and Markowitz - who advocates diversification across assets. We rely on the concepts of ambiguity and ambiguity aversion to formalize the idea of an investor?s "familiarity" toward assets. The model shows that when an investor is equally ambiguous about all assets, then the optimal portfolio corresponds to Markowitz?s fully diversified portfolio. In contrast, when an investor exhibits different degrees of familiarity across assets, the optimal portfolio depends on (i) the relative degree of ambiguity across assets, and (ii) the standard deviation of the estimate of expected return on each asset. If the standard deviation of the expected return estimate and the difference between the ambiguity about familiar and unfamiliar assets are low, then the optimal portfolio is composed of a mix of both familiar and unfamiliar assets; moreover, an increase in correlation between assets causes an investor to increase concentration in the assets with which they are familiar (flight to familiarity). Alternatively, if the standard deviation of the expected return estimate and the difference between the ambiguity of familiar and unfamiliar assets are high, then the optimal portfolio contains only the familiar asset(s) as Keynes would have advocated. In the extreme case in which the ambiguity about all assets and the standard deviation of the estimated mean are high, then no risky asset is held (non-participation). The model also has empirically testable implications for trading behavior: in response to a change in idiosyncratic volatility, the Keynesian portfolio always exhibits more trading than the Markowitz portfolio, while the opposite is true for a change in systematic volatility. In the equilibrium version of the model with heterogeneous investors who are familiar with different assets, we find that the risk premium of stocks depends on both systematic and idiosyncratic volatility, and that the equity risk premium is significantly higher than in the standard model without ambiguity.

Keywords: ambiguity; diversification; investment; portfolio choice; robust control

JEL Codes: D81; G11; G12; G23


Causal Claims Network Graph

Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.


Causal Claims

CauseEffect
Investor's familiarity across assets (G11)Optimal portfolio composition depends on relative ambiguity and standard deviation of expected returns (G11)
Low standard deviation of expected return estimate and low difference between ambiguity of familiar and unfamiliar assets (D80)Optimal portfolio includes both familiar and unfamiliar assets (G11)
High standard deviation of expected return estimate and high difference between ambiguity of familiar and unfamiliar assets (D80)Optimal portfolio contains only familiar assets (G11)
High ambiguity and high standard deviation (D80)No risky assets will be held (non-participation) (G19)
Increases in asset correlation (G19)Greater concentration in familiar assets (G19)
Keynesian portfolio (G11)More trading in response to changes in idiosyncratic volatility (G19)
Markowitz portfolio (G11)More trading in response to changes in systematic volatility (C58)

Back to index