Asymmetric Information and Bubbles: An Experiment

1
Asymmetric Information and Bubbles An Experiment

• Michael Brandner
• Jürgen Huber
• Michael Kirchler
• Matthias Sutter
• all University of Innsbruck

2
Outline

• Introduction
• Model description
• Experimental implementation
• Hypotheses
• Results and interpretation
• Concluding remarks

3
Introduction

• Starting point is the well-known basic setup of
  Smith, Suchanek and Williams (1988)
  subsequently SSW
• Outstanding characteristic declining fundamental
  value in the course of a market
• Typically, bubbles can be observed in this
  setting
• Related literature, revealing determinants for
  bubble reduction
• Less frequent dividend payments (Smith et al.
  2000)
• Constant fundamental value (Noussair et al. 2001)
• Detailed instructions (Lei et al. 2001)
• Futures markets (Noussair and Tucker 2004)
• Experience (Dufwenberg et al. 2005)
• Short selling (Haruvy and Noussair 2006)

4
Model description

• Basic features
• Assets of a virtual enterprise are traded - after
  every trading period the stock pays a dividend
  amounting to either zero or 20 MU (given equal
  probability)
• All traders know the dividend generating process,
  but nobody exhibits information about the actual
  dividend realizations before a periods
  termination
• No buy-out option is introduced so that assets
  are worthless after a rounds termination (one
  round being 10 successive periods), implying
  declining fundamental values

5
Model description

• We test for the impact of different information
  structures on bubble formation public
  information and asymmetric information
• Reference treatment (T1) symmetric (non-)
  information about future dividends (comparable
  e.g. to SSW (1988), Dufwenberg et. al (2005))
  all traders are of type I0 (just knowing the
  dividend generating process)
• Public information (T2) each trader is informed
  about the dividend realization of the current
  period from the outset all traders are type I1
• Asymmetric information (T3) on average, traders
  also know one future dividend, but this
  information is asymmetrically distributed two
  traders are of type I0, two of I1, and two are
  allocated to type I2 (so-called insiders, knowing
  the current and next periods dividend
  realizations)

6
Experimental implementation

• One market is populated by six traders -
  distribution of information levels in the three
  treatments
• We replicate three rounds in a market one round
  consisting of 10 periods (lasting for 120 seconds
  each)
• Initial endowments 3,000 MU and 10 assets/1,000
  MU and 30 assets (expected final profit is equal)
  
• Money and asset holdings are carried over from
  period to period, but reset to the respective
  starting levels after a rounds termination (no
  buy-out option, assets life-span is ten
  periods)

7
Experimental implementation

• The market is a continuous double auction with
  open order book, implemented with z-Tree
  (Fischbacher, 2007)
• Subjects can trade any quantity of assets, given
  the limitations of their asset inventory and cash
  endowment (neither loans are provided nor is
  short selling allowed)
• We conduct six markets per treatment for a total
  of 18 markets (at the University of Innsbruck)
• Participants have been business students with no
  previous experience in similar experiments

8
Trading screen
Information about future dividend(s)
9
History screen
10
Hypotheses

• H1 Public information (T2) about current
  periods dividend decreases bubble-size (compared
  to T1)
• H2 Asymmetric information (T3) about future
  dividends decreases bubble-size (compared to T1)
• H3 Informed traders will use the fundamental
  information provided so that prices increase when
  they see high dividends and fall when they see
  low dividends
• H4 Subjects in T3 have an increased incentive to
  trade across information levels

11
Results
12
Results
13
Results
14
Results
15
Results
Average price departures T1 40.6 MU T2 25.3
MU T3 18.4 MU
16
Results impact on bubble-size

• Comparing bubbles (overvaluation trading price
  minus corresponding fundamental value) in the
  three treatments pairwise (two-sided Mann-Whitney
  U-tests)
• ? H1 and H2 cannot be rejected bubbles are
  significantly smaller in markets with knowledge
  about the current dividend (T2 public
  information), and smaller again when dividend
  information is asymmetrically distributed (T3)

17
Explaining bubble-size reduction in T2 and T3

• H3 Informed traders will use the fundamental
  information provided so that prices increase when
  they see high dividends and fall when they see
  low dividends
• We examine the relationship between known
  positive/negative dividend realizations and the
  extent of overvaluation (trading price minus the
  corresponding fundamental value) correlation
  implies prices closer to fundamentals
• ? H3 cannot be rejected informed trader
  obviously concentrate more on the fundamentals of
  the stock (than on the hope for capital gains)
  with the consequence that prices track
  fundamentals closer

18
Trade activities within and across information
levels in T3

• H4 Subjects in T3 have an increased incentive to
  trade across information levels
• If subjects in T3 do indeed base their trading
  decision on their given dividend information,
  traders within any given information level should
  be rather homogeneous in their trading decisions
• The following matrix reveals pro-rata
  transactions, measured by the total amount of
  accomplished trades in T3 (n1334)

I0 I1 I2
I0 4.1 22.6 30.3
I1 - 6.3 27.7
I2 - - 9.0
19
Trade activities within and across information
levels in T3

• Trading frequency is significantly lower between
  information levels than across
• A possible explanation is that traders within
  each information level form similar beliefs based
  on the information they get and trade on these
  beliefs
• ? H4 cannot be rejected additional support that
  traders in T3 increasingly focus on fundamentals
  (thus implying that prices track fundamentals
  closer)

20
Concluding remarks

• Both changes to the standard setup public and
  asymmetric information reduce bubbles
  significantly (more pronounced effect in the
  latter treatment)
• Informed subjects in T2 and T3 base trading
  decisions on the information they receive, thus
  prices track fundamentals closer, making
  speculation less likely to occur
• Implication Real financial markets are
  undoubtedly characterized by mixed-experienced
  traders (recall Dufwenberg et. al (2005)) with
  asymmetric information about future cash flows,
  so that it should not be surprising that bubbles
  are indeed rare. The spectacular ones are less
  than five in a century (e.g. 1929, 1987 and the
  late 1990ies)!

21
Thank you for your attention!