BUBBLES, CRASHES, AND ENDOGENOUS UNCERTAINTY IN LINKED ASSET AND PRODUCT MARKETS*

• DOI: http://doi.org/10.1111/iere.12151
• Published date: 01 February 2016
• Date: 01 February 2016
• Author: Erik O. Kimbrough,Taylor Jaworski

INTERNATIONAL ECONOMIC REVIEW

Vol. 57, No. 1, February 2016

BUBBLES, CRASHES, AND ENDOGENOUS UNCERTAINTY IN LINKED ASSET

AND PRODUCT MARKETS∗

BYTAYLOR JAWORSKI AND ERIK O. KIMBROUGH1

Queen’s University, Canada; Simon Fraser University, Canada

In laboratory asset markets, subjects trade shares of a firm whose profits in a linked product market determine

dividends. Treatments vary whether dividend information is revealed once per period or in real time and whether the

firm is controlled by a profit-maximizing robot or human subject. The latter variation induces uncertainty about firm

behavior, bridging the gap between laboratory and field markets. Our data replicate well-known features of laboratory

asset markets (e.g., bubbles), suggesting these are robust to a market-based dividend process. Compared to a sample of

previous experiments, both real-time information revelation and endogenous uncertainty impede the bubble-mitigating

impact of experience.

1. INTRODUCTION

In financial markets, information about the value of a firm is revealed gradually as traders

observe the actions of managers and the firm’s profitability. In a class of asset pricing models,

firm value is derived from (rational) expectations about the future dividend stream provided

to shareholders. These models often assume that dividends are determined by draws from

a known distribution (or that share values follow a random walk). However, fundamentally,

financial market investments reflect beliefs about the behavior of people (e.g., employees,

managers, and CEOs), who are themselves reacting to information about the state of the firm

and market. Thus, traders of an individual firm’s stock must form expectations about the firm’s

future profitability, taking into account both the randomness of future events and the decisions

of managers. The impact of the resulting uncertainty and the timing of its resolution on the

informational efficiency of asset markets is an important open question.

In this article, we analyze laboratory financial markets in which asset values depend on the

profits earned by a firm in a linked product market. In standard experimental asset markets,

subjects trade shares of a finitely lived asset that pay a dividend determined by a draw from

a known distribution. Despite publicly available information on fundamental value, extensive

replications document the tendency of these markets to produce price bubbles.2Surprisingly, no

∗Manuscript received April 2012; revised October 2014.

1An earlier version of this article was circulated under the title “An Experimental Examination of Asset Pricing

under Market Uncertainty.” The authors thank the International Foundation for Research in Experimental Economics

Small Grants Program for providing funding for a preliminary version of this project and providing useful comments on

our ideas, and we thank the Economic Science Institute at Chapman University for the use of their laboratory during

those early stages. We also thank the Social Sciences and Humanities Research Council of Canada for funding. We

received numerous helpful comments from three anonymous referees and the editor, Hanming Fang, as well as Cary

Deck, Martin Dufwenberg, Shengle Lin, Ryan Oprea, Dave Porter, Andrew Smyth, and participants in seminars at

the Southern Economic Association Annual Conference and the Luxembourg School of Finance. We acknowledge

the able assistance of Kyle Bjordahl and Andriy Baranskyy for programming (and reprogramming) our software, and

we thank Yiqing (Phyllis) Zhou for excellent research assistance. All remaining errors are our own. Please address

correspondence to: Erik O. Kimbrough, Department of Economics, Simon Fraser University, 8888 University Drive -

WMC 4663, Burnaby, BC V5A 1S6, Canada. E-mail: ekimbrough@gmail.com

2See, e.g., Smith et al. (1988, 1993); Van Boening et al. (1993); Porter and Smith (1995); Sunder (1995); Caginalp

et al. (1998, 2000); Lei et al. (2001); Dufwenberg et al. (2005); Noussair and Tucker (2006); Haruvy and Noussair (2006);

Haruvy et al. (2007).

155

C

(2016) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

156 JAWORSKI AND KIMBROUGH

previous studies use real (market) activity to determine the dividend underlying fundamental

value.

In our environment, traders buy and sell shares of a product market monopolist who chooses

a price at which to sell units of a homogeneous good to a sequence of demand-revealing robot

buyers. At the end of each period, traders are paid a dividend based on the monopolist’s

profitability in that period. In three treatments, we vary the timing of information revelation

regarding the monopolist’s profitability and whether the monopolist is controlled by a profit-

maximizing robot or a human subject. Thus, we separately test the impact on asset prices

of varying exogenous uncertainty about product market demand and endogenous uncertainty

about the actions of firm managers. This design addresses two fundamental questions: (1) Do

traders find it easier to form common expectations when firm value is revealed in real time as

trading unfolds, and (2) how does information aggregation depend on uncertainty about the

behavior of the product market firm?

To establish a benchmark, our Base treatment closely follows the pattern of information

revelation from the standard experimental asset market environment due to Smith et al. (1988;

hereafter SSW). The firm is controlled by a profit-maximizing robot that, given the distribution

of buyer values, generates a constant expected dividend per period. Dividends are revealed

at the end of each trading period, and traders receive no within-period feedback about the

accumulation of firm profits.

Then, in our Info treatment, we reveal information about the end-of-period dividend in real

time. Specifically, we record the monopolist’s (simulated) transaction history over 10 periods

and replay the data to traders in the asset market as they buy and sell shares. Thus, exogenous

uncertainty about the dividend is slowly eliminated over the course of the trading period.

Regarding (1) above, evidence suggests that one important cause of bubbles in laboratory asset

markets is failures of common information to induce common expectations. With this treatment,

we explore whether the gradual revelation of information about the value of a share can provide

a source of common expectations and subdue speculative tendencies.

To address (2), our Human treatment maintains the information revelation process of the Info

treatment except that traders know that the monopolist is controlled by another human subject,

who may or may not set the profit-maximizing price. This design moves research on laboratory

financial markets in a direction that more closely approximates the problem faced by traders in

real-world financial markets, in which they face both relevant kinds of uncertainty. Although in

this treatment, both types of uncertainty are eventually resolved through continued observation

of the monopolist, traders must make decisions prior to this resolution if they hope to profit

from their trades. Ex ante, the expected impact of this manipulation is unclear. If endogenous

uncertainty is perceived as additional risk, then risk-averse traders will value shares at less than

fundamental value, and prices should be lower. If instead endogenous uncertainty interferes

with the ability to form common expectations, then mispricing may be more likely.

We collect data on 39 new experimental asset markets. Our findings suggest that actual

dividend realizations are a (slightly) better predictor of prices in the asset market than fun-

damental value; however, our Info treatment, which combines detailed framing with real-time

revelation of the process underlying dividends, does not appear to provide a common source

of expectations sufficient to eliminate price deviations from fundamental value. Moreover, in

comparisons to the Base and Info treatments, the impact of the Human treatment is generally

small. In general, our data replicate the stylized facts of earlier asset market experiments, sug-

gesting that bubbles are robust to the introduction of a market-based dividend. In fact, in direct

comparisons with a database of 39 previous asset market experiments, our Info and Human

treatments appear to slightly exacerbate bubbles by reducing the bubble-mitigating effects of

trading experience.

A few features of our experimental design and our findings may be better understood in

the context of the literature. For example, our decision to describe the market process that

generates dividends is informed by a recent contribution due to Kirchler et al. (2012). They

find that reframing the standard asset market experiment as buying and selling shares of a