Investor sophistication and asset prices

• Published date: 01 October 2020
• Author: Jeremy K. Page,Alok Kumar,George M. Korniotis
• Date: 01 October 2020
• DOI: http://doi.org/10.1002/rfe.1093

Rev Financ Econ. 2020;38:557–579. wileyonlinelibrary.com/journal/rfe

|

557

© 2019 University of New Orleans

1

|

INTRODUCTION

Investor sophistication plays a central role in behavioral asset pricing models.1

These models posit that demand shocks generated

by the systematic behavior of less sophisticated investors, the noise traders, can move prices away from their intrinsic values.

If arbitrage costs are high, arbitrage forces may not be very effective and prices may not move back to their fundamental values

quickly. Consequently, stock prices may exhibit excess comovement and higher degree of predictability (e.g., Alti & Tetlock,

2014). Although these theoretical models predict that investor sophistication would generate asset mispricing, there is little

empirical evidence to support this idea because it is difficult to measure investor sophistication directly. In this paper, we use

an investor-level dataset and construct a direct measure of investor sophistication, which we subsequently relate to mispricing

among stocks. Specifically, we examine if the degree of mispricing is stronger in the U.S. states with less sophisticated investors.

The potential relation between local mispricing and local sophistication is motivated by prior evidence on the geographical

segmentation in the U.S. capital markets and the cross-sectional variation in investor sophistication across regions. In particular,

Korniotis (2008) provides evidence of market segmentation. He demonstrates that since income shocks across the U.S. states

are not diversifiable (Asdrubali, Sorensen, & Yosha, 1996; Athanasoulis & van Wincoop, 2001), market returns can be better

explained using a model that treats the U.S. economy as a collection of 50 state-level investors instead of one national repre-

sentative investor. Furthermore, Becker (2007) finds that the market for bank loans and deposits is segmented, while Becker,

Ivkovich, and Weisbenner (2010) demonstrate that location of investors influence the demand for dividend-paying stocks. More

recently, Korniotis and Kumar (2013a) demonstrate that ownership and trading of local stocks has a significant local component.

Not only the U.S. capital markets are segmented, there is evidence that investor sophistication and behavioral biases vary

geographically. For example, Kumar (2009) and Kumar, Page, and Spalt (2011) find that investors’ preference for volatile, “lot-

tery-like” stocks differs across the U.S. states. In addition, Korniotis and Kumar (2011) show that investors’ ability to smooth

income shocks through their financial decision varies geographically.

Received: 12 November 2019

|

Accepted: 30 November 2019

DOI: 10.1002/rfe.1093

ORIGINAL ARTICLE

Investor sophistication and asset prices

George M.Korniotis1

|

AlokKumar1

|

Jeremy K.Page2

1Department of Finance, University of

Miami, Coral Gables, FL, USA

2Adobe Inc, Provo, UT, USA

Correspondence

Alok Kumar, Department of Finance,

University of Miami Herbert Business

School, 514 Jenkins Building, Coral Gables,

FL 33124, USA.

Email: akumar@miami.edu

Abstract

We show that geographical variation in the level of investor sophistication influences

local asset prices. Investors in less sophisticated regions exhibit stronger trading cor-

relations, and correspondingly, the returns of firms headquartered in less sophisti-

cated areas are more strongly correlated. Furthermore, we show that local economic

conditions have a greater ability to predict local stock returns in the U.S. states with

less sophisticated retail investors. These asset pricing results are driven by the so-

phistication of actual local investors, and not by the characteristics of the broader

local population.

KEYWORDS

investor sophistication, local return predictability, return comovement, state portfolios, trading

correlation

JEL CLASSIFICATION

G11; G12

558

|

KORNIOTIS eT al.

Motivated by these prior studies, we conjecture that the degree of mispricing would vary geographically since the U.S. capital

markets are segmented and investor sophistication varies across regions. To examine this local-mispricing local-sophistication hy-

pothesis, we examine whether local stocks in less sophisticated areas comove more with each other and if they are more predictable.2

Comovement among local stocks can arise since the local economy is more salient to local investors than the national economy.

Because local risk aversion is affected by local economic conditions (Korniotis & Kumar, 2013a), local economic conditions can be-

come a coordinating mechanism for the trading behavior of less sophisticated investors. The coordinated trading would generate a sys-

tematic component to local trading, which would in turn lead to stronger excess return comovement among the stocks of local firms.

Local return predictability can also arise because state-level economic conditions, a source of coordinated local demand

shocks, are persistent. Such persistence could create a slow-moving component in the investment mistakes of local investors.

For example, prolonged periods of good local economic conditions can lead to persistent local investor optimism and persistent

upward demand pressures. If arbitrage forces are limited, the persistent local demand shocks can lead to return predictability

(e.g., Alti & Tetlock, 2014). Therefore, in the U.S. states with less sophisticated investors, local economic conditions should be

strong predictors for the returns of local stocks.

To test the comovement and predictability hypotheses, we measure investor sophistication for each U.S. state. The sophis-

tication measure is based on investment decisions of retail investors at a U.S. brokerage house. Using the sophistication levels

of actual investors rather than the sophistication measures of the overall local population is a key innovation of our work.

Population demographic variables such as education, income, wealth, and so on, are not a good proxy for the sophistication

level of local investors because individuals who choose to participate in the market may be quite different from the overall local

population. Our measure is able to account for this self-selection among local investors. In fact, we show that all our results

become considerably weaker and insignificant when we use population-wide sophistication measures.

To construct the sophistication index, we focus on retail investors, rather than institutional investors, for two reasons. First,

institutional investors are broadly considered sophisticated relative to retail investors (e.g., Barber & Odean, 2008). Second, and

relatedly, prior research suggests that correlated retail investor demand can have an impact on stock prices and return comove-

ment, while correlated institutional trading seems to attenuate excess return comovement among stocks (Kumar & Lee, 2006;

Kumar, Page, & Spalt, 2013).

Using the sophistication measure, we find supporting evidence for our comovement hypothesis. Specifically, we identify

the U.S. state where a firm is headquartered (HQ state). Then, we estimate firm-level annual regressions of daily returns on a

return index of all firms in the HQ state controlling for various market factors. The beta on the HQ state index is our measure

of comovement. We find that these HQ state betas are higher for states with less sophisticated investors. Moreover, supporting

our hypothesis that local economic conditions matter for the comovement of local stocks, we find that the HQ state betas are

the highest in periods when the local economic conditions are extremely good or bad.

As a validation test, we examine whether the trading of local stocks is more correlated when local investors are less sophis-

ticated. This is an important test because return comovement should result from correlated local trading. Our measure of local

trading correlations is the partial correlation coefficient (controlling for the market return) between the buy–sell imbalance of retail

trades of a firm with the state-level buy–sell imbalance of retail trades of other firms headquartered in the same state. We find that

the local retail trading correlations (RTCs) are higher for firms in states with less sophisticated investors. Similar to our comove-

ment results, we find that the local trading correlations are the strongest when the local economic conditions are extremely good

or bad. Overall, our evidence suggests that investor sophistication enhances comovement in returns and trading among local firms.

Next, we test our return predictability hypothesis, which posits that local economic indicators should predict stock returns

more strongly in areas where investors are less sophisticated. Specifically, we estimate return predictability regressions similar to

those in Korniotis and Kumar (2013a). The dependent variables are the quarterly characteristic-adjusted state portfolio returns.

The independent variables are U.S.-level and state-level macroeconomic variables (income growth, relative unemployment,

housing collateral ratio; Lustig & Van Nieuwerburgh, 2005) as well as other U.S.-level macroeconomic variables (e.g., the

paper-bill spread, term spread, default spread, the dividend-to-price ratio, and the cay measure of Lettau & Ludvigson, 2001a).

We estimate the predictability models separately for high (above median) and low (below median) sophistication states. We

find that return predictability is stronger among the U.S. states with less sophisticated investors. Consistent with our hypothesis,

we also find that local economic predictors are stronger predictors of local returns in less sophisticated states.

Next, we focus on low sophistication states and test whether return predictability is stronger within states where investors

also exhibit stronger local bias. We expect that local mispricing to be stronger in those states where investors are less sophisti-

cated and primarily hold local stocks. Consistent with our conjecture, return predictability is stronger in these low-sophistica-

tion high-local bias states.

To examine the economic significance of the return predictability differences between high and low sophistication states, we

define Long–Short trading portfolios where the Long (Short) position is in firms located in states with poor (good) economic