Information Transparency and Coordination Failure: Efficiency,...

Information Transparency and Coordination Failure: Efficiency, Security, and Risk Dominance

Regina M. Anctil*

John Dickhaut*

Chandra Kanodia*

Cathleen Johnson**

*University of Minnesota **Center for Interuniversity Research and Analysis on Organizations, Montreal, QC

November 5, 2004

Very preliminary and incomplete. Comments welcome. The authors thank Frederick Rankin, Orly Sade, and participants in the 2004 American Economics Association annual meetings for helpful comments and suggestions.

Information Transparency and Coordination Failure: Efficiency, Security, and Risk Dominance

ABSTRACT. This experiment examines coordination failure among a group of privately informed decision makers. Decision makers in this study must decide individually whether to keep funds invested in a risky project, or to break ranks and liquidate their investment. The ultimate success of the project depends on a critical number of individuals keeping their funds invested. Coordination failure occurs when individuals break ranks, causing the project to fail when it otherwise would have succeeded. We are interested in finding if the quality of an individual’s private information influences the frequency of coordination failures. Prior studies have posited that improving private information can increase the incidence of coordination failure. This study provides evidence to support this claim. Our results imply that whether information improvement increases or decreases coordination failure, depends on the fundamental outlook of the project (optimistic versus pessimistic), and on the nature of the information improvement. Our results also provide strong support for the claim that individual decisions to break ranks or to stay with the project are driven by an assessment of risk dominance.

Information Transparency and Coordination Failure:

Efficiency, Security, and Risk Dominance

There are many settings where the success of a group endeavor depends upon the group

sticking together. In these settings a loss of confidence in others’ intention to stay the course can

lead to the individual defection of every member. A bank run is the classic example of the

phenomenon. But this coordination problem underlies group campaigns from the boardroom to

the battlefield, and is a key vulnerability of decentralized decision making. Scholars have long

been interested in whether such strategic uncertainty can be managed by improving the

information available to group members. So, for example, if each member of a consortium of

lenders receives periodic updates on the creditworthiness of a borrower, they are more likely to

act in concert, based on the information in the report. But it is practically impossible to prevent

individuals from pursuing their own information sources. If members of the group perceive that

others have private information, its effect on their own beliefs about the beliefs of others can

exacerbate strategic uncertainty rather than mitigate it.

In our study we test the hypothesis that it is not private information or the quality of

private information, per se, that determines whether coordination failure will occur or not.

Rather, it is the influence that private information has on each decision maker’s assessment of

risk dominance that determines whether or not coordination failure will occur. Risk dominance

captures the decision maker’s expected cost of being on the “wrong side” of prevailing play

(Harsanyi and Selten, 1988). The risk dominance criterion, predicts that coordination success or

failure will depend on each individual’s desire to minimize his loss if he guesses wrongly what

others will do. Using the loan example, a creditor that rolls over the debt is at risk of losing his

some of his principle if enough others foreclose and drive the borrower into bankruptcy.

1

However, a creditor who forecloses risks losing the income earned by those who stayed with the

project if it is successful. The risk dominance criterion would predict that the creditor would

choose the action that minimizes these two costs. We show that the risk dominance calculus

depends crucially on the quality of private information. We hypothesize that coordination failure

or success will follow risk dominance. This study extends the work of Anctil, Dickhaut,

Kanodia, and Shapiro (2003) (ADKS). Using the same experimental instrument as in that study,

with a somewhat richer information structure, we find that information quality improvement can

lead to increases or decreases in coordination failure. We also find that the risk dominance

criterion predicts the outcome in every case.

The ADKS study experimentally investigates the effect of information transparency when

multiple creditors must individually decide whether to foreclose or roll over loans that have been

made to fund a risky project in a firm. Each creditor receives idiosyncratic information that is

useful in assessing a state of nature (Bankrupt, Uncertain, or Solvent) that determines the

fundamental soundness of the project. In the Bankrupt and Solvent states the project fails or

succeeds, respectively, regardless of the creditors’ actions. In one state, the Uncertain state, the

success of the project depends on a critical number of creditors staying in. Therefore each

creditor must also weigh the possibility that the foreclosures of others will cause the project to

fail, even if the project is fundamentally sound. If a creditor believes that many other creditors

will prematurely foreclose on a loan, it becomes individually rational to foreclose, resulting in

inefficient equilibria sustained by self-fulfilling beliefs. Consequently, there is a possibility that

in equilibrium the project fails, even though failure is not warranted by the underlying economic

fundamentals.

2

Theory predicts that each creditor adopts a trigger strategy based on his private signal,

foreclosing if the value of the private signal is below a critical threshold (Morris and Shin, 1999).

In equilibrium, everyone adopts the same critical threshold. In their experiment, ADKS use three

signals, High, Medium, and Low, varying transparency by increasing the precision of the signals.

The parameters of the experiment insure that creditors receive their highest expected payoff

when each forecloses only when seeing the Low signal. This trigger strategy, if adopted by all, is

the most efficient equilibrium the group can reach. But there is another strategy, to foreclose if

the signal is Low or Medium, that can also emerge in equilibrium. We refer to this as the secure

equilibrium because it emerges when all individuals try to minimize their maximum potential

loss.

In the ADKS experiment, the Low transparency condition has one equilibrium prediction:

foreclose on the Low signal in the Optimistic setting, foreclose on either the Medium or Low

signals in the Symmetric setting. The High transparency treatment involves a symmetrical

increase in precision across the three signals. This gives rise to two possible equilibria in each

setting, the efficient equilibrium (foreclose on the low signal) and the secure equilibrium

(foreclose on the Medium or Low signals). The efficient equilibrium is characterized by no

inefficient project failure, while the secure equilibrium yields inefficient project failure 100% of

the time. They find strong support for the equilibrium predictions in the Low transparency

settings, for the efficient equilibrium in the High transparency Optimistic setting, and for the

secure equilibrium in the High transparency Symmetric setting. They are unable to conclude that

increasing information transparency influences coordination failure for better or worse because

the outcome within each setting stays the same going from Low to High transparency.

3

The results of their study are consistent with the hypothesis that risk dominance determines

whether high transparency will result in coordination failure. That is, when a secure equilibrium

and an efficient equilibrium coexist the equilibrium that arises is risk dominant. Their results are

inconclusive, however, because they do not find an overall effect for their treatment variable,

information transparency. The risk-dominance prediction in that experiment coincides with the

null hypothesis of no transparency effect.

In this study, we conduct an experiment where the treatment is again information

transparency. This study pursues the role of information transparency further by allowing for

asymmetric increases in signal transparency. In other words, we increase the precision of the

High (Low) signal, while the precision of the Low (High) signal remains constant. In this setting,

as with ADKS, there are two economic settings, an Optimistic setting and a Pessimistic setting.

In each setting, the Low information transparency treatment yields a unique, dominant-strategy

equilibrium. The efficient equilibrium arises uniquely in the Optimistic setting and the secure

equilibrium arises uniquely in the Pessimistic setting. In each setting we increase transparency in

two stages. In the first stage, we increase transparency by increasing the ability of subjects

receiving the intermediate signal to rule out one of the extreme states. In the second stage we

increase the ability of subjects receiving the intermediate signal to rule out the other extreme

state. In each successive stage, both the secure and efficient equilibria coexist. But in each

successive stage their risk-dominance order reverses. In the Optimistic setting, the first-stage

increase in transparency makes the secure equilibrium risk dominant, and the second stage

increase makes the efficient equilibrium risk dominant. The reverse is true in the Pessimistic

setting. Thus the risk-dominance hypothesis predicts that the first increase in transparency will

4

cause subjects to move away from the baseline, Low-Transparency equilibrium, and the second

increase will cause them to move back to it.

This study contributes to research on coordination among groups of creditors (Diamond

and Dybvig, 1983; Morris and Shin, 1999). Our experiment draws particularly on the model of

Morris and Shin (1999), who were the first to model the creditor coordination problem with

private information. Morris and Shin characterize when such “global games” (Carlsson and van

Damme, 1993) have a unique equilibrium where each creditor employs a trigger strategy of the

type described above. They find that for some parameter values, increased transparency shifts the

equilibrium foreclosure threshold upwards, and thereby enlarges the set of states in which

inefficient foreclosure occurs.

Our study also contributes to the experimental economics literature on equilibrium

selection and coordination failure (e.g., Cooper, DeJong, Forsythe, and Ross, 1990; Van Huyck,

Battalio, and Beil, 1990, 1991). Previous experiments have examined whether players moving

simultaneously in complete-information coordination games can implement the Pareto-optimal

equilibrium when more than one Nash equilibrium exists. Coordination failure occurs when all

players are better off in one equilibrium, but are unable to coordinate their strategies to achieve

that equilibrium. Van Huyck, Battalio, and Beil, 1991 examine equilibrium selection in an

average opinion game with secure and efficient equilibria. They find that subjects converge on

the efficient equilibrium only when the secure equilibrium is rendered non-salient, and vice

versa. When both strategies are salient, neither of the equilibria is supported. They are not able to

determine whether risk dominance is playing a role in equilibrium selection. Our study provides

a setting to test this hypothesis. Our setting will assess whether the need for coordination drives

5

players to treat their potential loss from deviating from the group norm as more salient in

deciding whether or not to take a risk, than is their expected payoff.

The remainder of this paper is organized as follows. In section we describe the setting for

the experiment. In Section II we describe the basic decision problem of the experiment. In

Section III we present the specific parameters and design for each treatment, as well as our

experimental procedures.

II. The Creditor Coordination Experiment

ADKS (2003) present a finite version of the Morris and Shin (1999) model. In both

models creditors are endowed with a loan on a risky project, and receive a noisy, private signal

regarding the viability of the project. In both models, equilibria take the form of threshold signal

values, below which a creditor would foreclose, above which a creditor would roll over. The

difference between the two models is that the ADKS version assumes there is a finite number of

creditors rather than a continuum, and that the state and signal spaces are finite. Allowing for

discontinuity in the state and signal space, ADKS cannot rule out the occurrence of multiple

equilibrium signal values, whereas, Morris and Shin are able to characterize the setting where

there is a unique equilibrium. Our experimental design is based on the ADKS adaptation.

There are eight risk-neutral creditors, each creditor must decide whether to roll over the

debt he holds or foreclose. Foreclosure yields a deterministic payoff 0>λ . Rollover yields a

payoff of λ>R if the project succeeds, and zero if the project fails. The success of the project

depends on a random variable θ representing the state of nature. In the experiment there are three

possible states, B, U, and S. In state B (Bankrupt), the project fails; anyone who rolls over

6

receives 0. In state S (Solvent), the project succeeds; anyone who rolls over receives R. Finally in

state U (Uncertain), the project succeeds if and only if at least Z out of eight creditors roll over.

If at least Z creditors roll over, all who roll over receive R. If fewer than Z roll over, all who roll

over receive 0.

Creditors have identical prior probabilities { },, SUB πππ on the set of possible states.

Before making a decision, each creditor privately observes an idiosyncratic noisy signal x of the

underlying state. There are three possible signals, L (Low), M (Medium), H (High). We denote

the probability of seeing signal L in state B as rB. While the probability of seeing signal H in state

S is given by rS. Table 1 summarizes the conditional distribution of signals in each state.

Assuming this distribution, we are able to increase or decrease signal transparency by

manipulating the values of the parameters rj, j=B, S. We discuss this in more detail below.

[INSERT TABLE 1 ABOUT HERE]

In deciding whether to roll over the loan or foreclose, each creditor compares the

expected payoff from rolling over to the deterministic payoff, λ , from foreclosing. In making his

decision, each creditor assesses the probability the project will fail conditional on his rolling over

and conditional on the state of nature. Let φ denote the assessed probability of project failure in

state U. Then the expected payoff to a creditor who rolls over is

RxUProbRxSProb )1)(|()|( φ−+ . (1)

Expression (1) shows the creditors face two sources of uncertainty in assessing the expected

payoff for rolling over, state uncertainty and strategic uncertainty. Strategic uncertainty enters

through the value of φ. If players choose a threshold signal above which they roll over, then in

equilibrium, φ represents the probability that fewer than Z-1 creditors receive signals above the

equilibrium threshold signal value.

7

In this setting there are four possible trigger strategies: FA (Foreclose regardless of the

signal); FM (Foreclose only if the signal is M or lower); FL (foreclose only if the signal is L);

RA (Rollover regardless of the signal). Given the conditional probabilities defined in Table 1,

signals L and H are perfectly informative regarding states B and S, respectively. Therefore, a

rational creditor will always foreclose upon seeing signal L, and will always roll over upon

seeing signal H, implying that strategies RA and FA cannot be supported as equilibria. There are

two candidate equilibrium trigger strategies, FL and FM. Both equilibria are independent of the

level of Z. If the equilibrium is FL, all creditors who receive signal M roll over. Since signal M is

received with probability one when the state is U, the equilibrium probability of project failure in

state U must be zero when the equilibrium is FL. If the equilibrium is FM, all creditors who

receive signal M foreclose. Since signal M is received with probability one when the state is U,

the equilibrium probability of project failure in state U must be one when the equilibrium is FM.

Thus,

=m is FMequilibriuifm is FLequilibriuif

U ,1,0

*ϕ (2)

The implication of the design of the information system is to define the coordination problem

entirely in terms of what players do upon receiving signal M. The players’ action upon seeing

signal M defines the equilibrium; it also defines successful and failed coordination. Van Huyck

et al (1990, 1991) define coordination failure as the failure to reach the superior equilibrium

when more than one equilibrium exists. If players foreclose upon seeing signal M, this represents

coordination failure (the FM equilibrium). As shown in equation (2), this coordination failure is

equivalent to a 100% project failure rate in the Uncertain state. If players roll over upon seeing

signal M, this represents coordination success (the FL equilibrium). Coordination success is

equivalent to a 100% project success rate in the Uncertain state.

8

Information Transparency:

The other major implication of the design of the information system is that the nature of

the coordination problem depends entirely on information transparency. We define the

distribution of signals in terms of the parameters rB and rS. Increasing either of these parameters

increases signal transparency in the sense that it enables the decision maker to interpret the

intermediate signal, M, more narrowly. Before we discuss how this increase will affect the

equilibrium prediction and risk dominance, we provide the formal connection between a change

in the parameter rB or rS, and the more general definition of transparency as given by the

Blackwell fineness criterion.

As shown by Kihlstrom (1984), Blackwell fineness is equivalent to expressing one set of

conditional probabilities as a stochastic transformation of another. Formally, let

represent two alternative probabilities of drawing signal conditional on state

tmtm gf ,

mtx θ , t=1,…,T,

and m=B,1,…,M,S. Let

=

TSTmTB

SmB

fff

ffff

LL

MMM

LL 111

and .

=

TSTmTB

SmB

ggg

gggg

LL

MMM

LL 111

We define to be more transparent than (finer than) f g if there is a stochastic transformation Q,

such that , whereQ , for all Qfg =

=

TTT

T

qq

qq

L

MM

L

1

111

0≥ijq ji , and for all , j , . 11 =Σ = ijTi q

The following proposition is given in terms of an increase in rB, holding rS constant. The

alternative can be shown as well.

Proposition 1:

9

Let {rB, rS} and { ' , rBr

≤ Br

S} denote the parameters of two conditional distributions as defined in

Table 1, where . Then the mapping given by {r1'0 ≤< Br B, rS} can be expressed as a

stochastic transformation of that given by { ' , rBr S} using the matrix .

−

10001'/00'/1

B

B

rdrrdr

Efficient, Secure, and Risk-Dominant Equilibria:

We are interested in examining the impact of transparency on the incidence of

coordination failure. In particular, we wish to determine if risk dominance determines when an

increase in transparency will be beneficial and when it will be detrimental. Van Huyck (1990,

1991) categorize candidate equilibria as efficient versus secure. The efficient equilibrium has the

higher expected payoff; the secure equilibrium has the higher minimum payoff. Our candidate

equilibria can be classified in the same way in the experiment. The FL equilibrium is always the

efficient equilibrium; the FM equilibrium always the secure equilibrium. Risk dominance orders

equilibria in a different way, and can apply to either efficient or secure equilibria. The intuition

underlying risk dominance is that a participant assesses his potential loss, given his strategy, if

others choose to follow a different strategy. Thus, if there are two candidate equilibria, the risk-

dominant equilibrium is that equilibrium for which defection to the other equilibrium produces

the highest loss when remaining participants continue to play the original equilibrium. 1 Risk

dominance might be a more salient criterion than either efficiency or security in this setting

because it incorporates both fundamental and strategic uncertainty, capturing the importance of

conformity.

1 Loss is formally computed multiplicatively across players, but in our symmetric payoff setting, unilateral loss works similarly. See Harsanyi and Selton (1988) for a formal treatment of risk dominance.

10

The following expressions characterize the cost of deviating from each strategy, FL and

FM.

The expected payoff from unilaterally deviating to FM when all others play FL is given by, πB(1-rB ) λ + πUλ − πS(1-rS )(1−λ) − πU . (3) The expected payoff from unilaterally deviating to FL when all others play FM is given by, − πB(1-rB ) λ − πU λ + πS(1-rS )(1−λ) . (4)

Proposition 2 defines equilibria and risk dominance using these two expressions, and enables us

to see how increasing the distributional parameters rB and rS will affect the equilibrium or

equilibria that arise, as well as risk dominance. These will enable us to choose the graduated

levels of transparency for the experimental treatments.

Proposition 2:

If (3)<0 then FL is an equilibrium. If (4)<0 then FM is an equilibrium. If (3)>0 then FM is a

dominant-strategy equilibrium. If (4)>0 then FL is a dominant-strategy equilibrium. If (3) < (4)

<0, then both FL and FM are equilibria, and FL risk-dominates FM. If (4)<(3)<0, then both FL

and FM are equilibria, and FM risk-dominates FL.

The statements in proposition 2 follow from the definition of a Nash equilibrium, a dominant-

strategy Nash equilibrium, and our working definition of risk dominance.

In the next section we present the design of the experiment and describe the procedures

for conducting it.

11

III. EXPERIMENTAL DESIGN

Experimental Design and Predictions

We designed our experiment to test whether an improvement in transparency changes the

incidence of project failure in the uncertain state, and whether risk-dominance determines the

sign of the change. To address these questions, in each treatment we set the transparency

parameters {rB,rS}at one of three levels corresponding to Low Transparency, Higher

Transparency and Highest Transparency, within each of two economic settings, Pessimistic and

Optimistic. In all treatments, we set Z=5, so success in the Uncertain state requires at least five

creditors to roll over. Under Low transparency the parameters rB and rS are both set to 0.4. In

other words, the probability of receiving signal L (H) in state B (S) equals 0.4; and the

probability of receiving signal M in states B and S equals 0.6. Under the Higher Transparency

treatment either rB or rS is increased to 0.9, which reduces the probability of receiving signal M

in the respective state to 0.1. In the high transparency setting, recipients of signal M are able to

significantly shift their expectations optimistically or pessimistically. Under Highest

transparency, both rB and rS are set at 0.9, which reduces the probability of seeing signal M in the

Bankrupt or the Solvent state to 0.1. Thus signal M is much more diagnostic of the Uncertain

state at this level of transparency.

The two economic settings differ in their relative optimism as given by the prior state

probabilities and the payoff to foreclosure. The Pessimistic setting has priors weighted toward

state B (bankrupt), and a higher foreclosure payoff λ. The Optimistic setting has priors weighted

toward state S (solvent), and a lower foreclosure payoff λ. As Proposition 2 indicates, the

equilibrium prediction(s) in each treatment will thus depend on the setting (Optimistic,

Pessimistic), and the Transparency of information (Low, Higher, Highest). Table 2 summarizes

12

the parameterization of each treatment and the predictions. Table two also summarizes each

subject’s expected payoff in equilibrium, and the payoffs for deviating given by expressions (3)

and (4).

[INSERT TABLE 2 ABOUT HERE]

Experimental Procedures

Experiment. The instructions explain (See Appendix) the experimental procedures and

information structure in the following eight steps (Table and Figure numbers contained in the list

below refer to the tables and figures in the instructions). Each participant acts as a creditor to a

company to whom he or she has lent money. Each participant decides whether to continue to

invest the money in the company until the project is complete, or to withdraw the money before

the project is complete.

1. Table 1 of the instructions explains that a participant’s payoffs in each experimental

round will depend on his or her decision (continue or withdraw) and whether the project succeeds or fails. The payoff for continuing to invest is $0.50 if the project succeeds and $0.00 if the project fails. The payoff for withdrawing from the project, regardless of project outcome, is $0.20 (Optimistic setting) or $0.30 (Pessimistic setting). The instructions state that when participants make their decisions, they will not know for certain whether the project will succeed or fail.

2. Table 2 of the instructions introduces the three states of nature (Solvent, Uncertain, and

Bankrupt), their probability of occurring in a given experimental round, and their relation to project success or failure.

3. A computer selects one of three states of nature based on the probabilities given in Table

2. Conditional on the state of nature selected, the computer then selects one private clue for each participant. The clues are labeled Low, Medium, or High. Each participant is told that his or her clues are private. Figure 1 in the instructions states the probabilities for the states of nature and private clues, and the project outcomes (success or failure).

4. Participants choose one of four decision rules to indicate their decisions in the

experiment. The decision rules are to (1) withdraw always, regardless of the clue, (2) withdraw if the clue is Low, otherwise continue, (3) withdraw if the clue is Low or Medium, otherwise continue, or (4) continue always, regardless of the clue. (Decision

13

rules (2) and (3) correspond to FL and FM in our model, respectively.) If a participant fails to select a decision rule within the allotted time for a given set, he or she will not be grouped with seven other participants and will receive $0.00 for that set of 10 rounds.

5. The decision rules are executed for a set of 10 rounds, after which participants may revise

their decision rules for the next set of 10 rounds, and so on. Participants are not told in advance how many sets will be played.

6. Participants are privately given (a) a table of their chosen decision rules and payoffs for

each set, and (b) a table indicating his or her private clue, decision, project status, and payoff for each experimental round in the current set.

7. At the conclusion of the experiment, participants are paid their cumulative payoffs plus

an additional $10 for showing up on time.

14

References

Anctil, R. M., J. Dickhaut, C. Kanodia, and B. Shapiro. 2004. Information transparency and coordination failure: theory and experiment. Journal of Accounting Research 42 (supplement): 159-196. Carlsson, H. and E. van Damme. 1993. Global games and equilibrium selection. Econometrica, 61(5), 989-1018. Cooper, R. W., DeJong, D. V., Forsythe, R., and T. W. Ross. 1990. Selection criteria in coordination games: Some experimental results. American Economic Review, 80 (1), 218-233. Diamond, D. and P. Dybvig, P. 1983. Bank runs, deposit insurance, and liquidity. Journal of Political Economy, 91 (3), 401-419. Harsanyi, J. C. , and R. Selten. 1988. A General Theory of Equilibrium Selection in Games. Massachusetts Institute of Technology Press. Kihlstrom, R. E. 1984. A “Bayesian” exposition of Blackwell’s theorem on the comparison of experiments. In M. Boyer and R. E. Kihlstrom (eds.), Bayesian Models in Economic Theory. Elsevier Science Publishers. Morris, S. and H. S. Shin. 1999. Coordination risk and the price of debt. Working Paper, Yale University. Norris, F. 2002. A new set of rules is in the works for accounting. New York Times. Oct. 22. Van Huyck, J. B., Battalio, R. C., and R. O. Beil. 1990. Tacit coordination games, strategic uncertainty, and coordination failure. American Economic Review, 80 (1), 234-249. Van Huyck, J. B., Battalio, R. C., and R. O. Beil. 1991. Strategic uncertainty, equilibrium selection, and coordination failure in average opinion games. The Quarterly Journal of Economics, 106 (3), 885-910.

15

Table 1. Conditional Distribution of Private Signals on States of Nature

STATES B U S

f(L|state) rB 0 0 f(M|state) 1-rB 1 1-rS

CONDITIONAL SIGNAL PROBS.

f(H|state) 0 0 rS

Table 1 describes the information structure for each participant in the setting. B, U, and S are the states of nature, Bankrupt, Uncertain and Solvent while L, M, and H are the possible messages a participant can receive (low, medium and high). By varying the parameters, rj, j=B, S, transparency is manipulated in the setting. Transparency is increasing in rj.

16

Table 2 Signal Distributions, Economic Settings, and Predictions Panel A: Signal distribution in the optimistic setting for Low, Higher, and Highest levels of transparency. Optimistic Setting: πB=0.3,πU=0.2,πS=0.5; R=$.50, λ=$0.20 Low Transparency Higher Transparency Highest Transparency {rU,rS}={0.4,0.4} {rU,rS}={0.4,0.9} {rU,rS}={0.9,0.9} State B U S B U S B U S Prob(L|State) 0.4 0.0 0.0 0.4 0.0 0.0 0.9 0.0 0.0 Prob(M|State) 0.6 1.0 .06 0.6 1.0 0.1 0.1 1.0 0.1 Prob(H|State) 0.0 0.0 0.4 0.0 0.0 0.9 0.0 0.0 0.9 Panel B: Signal distribution in the pessimistic setting for Low, Higher, and Highest levels of transparency. Pessimistic Setting: πB=0.5,πU=0.2,πS=0.3; R=$0.50, λ=$0.30 Low Transparency Higher Transparency Highest Transparency {rU,rS}={0.4,0.4} {rU,rS}={0.9,0.4} {rU,rS}={0.9,0.9} State B U S B U S B U S Prob(L|State) 0.4 0.0 0.0 0.9 0.0 0.0 0.9 0.0 0.0 Prob(M|State) 0.6 1.0 .06 0.1 1.0 0.6 0.1 1.0 0.1 Prob(H|State) 0.0 0.0 0.4 0.0 0.0 0.4 0.0 0.0 0.9 States: B (Bankrupt): Project fails. U (Uncertain): Projects succeeds if at lease 5 of 8 creditors roll over; otherwise it fails. S (Solvent): Project succeeds. Payoffs: R: success payoff λ: foreclosure payoff Signals: L: Low signal M: Medium signal H: High signal Transparency Parameters: rU: Probability of signal L in state B. rS: Probability of signal H in state S.

17

Table 2 Signal Distributions, Economic Settings, and Predictions Panel C: Equilibrium predictions in the optimistic setting for each transparency treatment, expected payoff per decision, opportunity cost of selecting FM(FL) when all others play FL(FM). Optimistic Setting: πB=0.3,πU=0.2,πS=0.5; R=$.50, λ=$0.20

Transparency

Equilibrium Trigger

Strategyc

Expected Payoff Per Decision

Incremental Payoff per

Decision for Defecting to FM

When All Others Play FL


Decision for Defecting to FL

When All Others Play FM

Low: {rU,rS}={0.4,0.4}

FL $3.70 -$0.11 $0.10

FL $3.70 Higher: {rU,rS}={0.4,0.9} FM* $3.40

-$0.40 -$0.60

FL* $4.00 Highest: {rU,rS}={0.9,0.9} FM $3.40 -$0.70 -$0.30

*: Risk-dominant equilibrium Panel D: Equilibrium predictions in the pessimistic setting for each transparency treatment, expected payoff per decision, opportunity cost of selecting FM(FL) when all others play FL(FM). Pessimistic Setting: πB=0.5,πU=0.2,πS=0.3; R=$0.50, λ=$0.30

Transparency

Equilibrium Trigger

Strategyc

Expected Payoff Per Decision


Decision for Defecting to FM

When All Others Play FL


Decision for Defecting to FL

When All Others Play FM

Low: {rU,rS}={0.4,0.4}

FM $3.20 $0.10 -$0.11

FL* $3.90 Higher: {rU,rS}={0.9,04} FM $3.20

-$0.60 -$0.40

FL $3.90 Highest: {rU,rS}={0.9,0.9} FM* $3.50 -$0.30 -$0.70

*: Risk-dominant equilibrium

18

Table 3. Number of Participants Who Most Frequently Chose Each Decision Rule in the Low Transparency Settings.

Decision Rule

Low Transparency Optimistic setting (Foreclose if receive Low Signal (FL) is Bayesian Nash)

Low Transparency Pessimistic setting (Foreclose if receive Medium Signal (FM) is Bayesian Nash)

Roll over always (RA) 1 0 Foreclose if receive

Low Signal (FL) 23 2

Foreclose if Receive Low or Medium Signal (FM) 0 19

Foreclose Always (FA) 0 0 In the Optimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .30), Uncertain (πU = .20), and Solvent (πS = .50). In the Pessimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .50), Uncertain (πU = .20), and Solvent (πS = .30). The Low Transparency Optimistic setting had 24 participants. The Low Transparency Pessimistic setting had 21 participants. If a participant chose more than one strategy with equal frequency, half a point is given to each decision rule.

19

Table 4. Number of Participants Who Most Frequently Chose Each Decision Rule in Higher Transparency and Highest Transparency Settings Panel A. Higher Transparency Optimistic and Higher Transparency Pessimistic Settings.

Decision Rule

Higher Transparency Optimistic setting (Foreclose if receive Medium Signal (FM) is the Risk-Dominant decision rule)

Higher Transparency Pessimistic setting (Foreclose if receive LowSignal (FL) is the Risk-Dominant decision rule)


Low Signal (FL) 4 20.5

Foreclose if Receive Low or Medium Signal (FM) 20 2.5

Foreclose Always (FA) 0 1 Panel B. Highest Transparency Optimistic and Highest Transparency Pessimistic Settings.

Decision Rule

Highest Transparency Optimistic setting (Foreclose if receive Low Signal (FL) is the Risk-Dominant decision rule)

Highest Transparency Pessimistic setting (Foreclose if receive Medium Signal (FM) is the Risk-Dominant decision rule)


Low Signal (FL) 19 7

Foreclose if Receive Low or Medium Signal (FM) 0 17

Foreclose Always (FA) 0 0 In the Optimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .30), Uncertain (πU = .20), and Solvent (πS = .50). In the Pessimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .50), Uncertain (πU = .20), and Solvent (πS = .30). Both the Higher Transparency Optimistic and Pessimistic settings had 24 participants. The Highest Transparency Optimistic setting had 21 participants, two of whom did not respond after set three. The Highest Transparency Pessimistic setting had 24 participants. If a participant chose more than one strategy with equal frequency, half a point is given to each decision rule.

20

Table 5. Selected Decision Rules by Decision Period

Panel A. Number of Participants Selecting Each Decision Rule by Decision Period in the Optimistic Settings

Low Transparency Optimistic (n= 24 per decision period) Prediction: FL

Higher Transparency Optimistic (n= 24 per decision period) Prediction: FL or FM*

Highest Transparency Optimistic (n= 21 per decision period) Prediction: FL* or FM

Continue Always (RA)

Foreclose if Low Signal (FL)

Foreclose If Low or Medium Signal (FM)

Foreclose Always

(FA)


Foreclose if Low Signal


Foreclose Always




Foreclose Always

(FA) 1 3 15 6 0 1 12 11 0 0 17 3 0 2 1 18 5 0 1 10 13 0 0 17 4 03 3 20 1 0 0 9 15 0 1 19 1 04 0 23 1 0 0 7 17 0 2 17 0 05 2 21 1 0 0 6 18 0 0 19 0 06 1 21 2 0 1 4 19 0 1 18 0 07 0 23 1 0 1 5 18 0 1 17 1 08 2 21 1 0 0 4 20 0 0 19 0 09 0 23 0 1 1 0 23 0 0 19 0 010 0 24 0 0 1 3 20 0 1 18 0 0Total 12 209 18 1 6 60 174 0 6 180 9 0 In the Optimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .30), Uncertain (πU = .20), and Solvent (πS = .50). The Highest Transparency Optimistic setting had 21 participants, however, one participant did not respond in set one and two participants did not respond in sets four through ten. * Risk-dominant equilibrium.

21

Table 5 (Continued) Panel B. Number of Participants Selecting Each Decision Rule by Decision Period in the Pessimistic Settings

Low Transparency Pessimistic (n= 21 per decision period) Prediction: FM

Higher Transparency Pessimistic (n= 24 per decision period) Prediction: FL* or FM

Highest Transparency Pessimistic (n= 24 per decision period) Prediction: FL or FM*


Foreclose if Low Signal (FL)


Foreclose Always

(FA)




Foreclose Always




Foreclose Always

(FA) 1 0 10 11 0 0 14 10 0 0 13 9 1 2 0 11 10 0 0 12 11 1 0 12 12 03 1 8 12 0 0 15 7 1 1 11 11 14 1 4 15 1 0 19 4 1 0 12 12 05 0 5 16 0 0 20 3 1 0 10 14 06 0 4 17 0 0 19 4 1 0 5 19 07 0 3 18 0 0 19 4 1 0 5 19 08 0 2 19 0 0 22 1 1 0 8 16 09 0 1 20 0 0 22 1 1 0 2 22 010 0 0 21 0 0 22 1 1 0 4 20 0Total 2 48 159 1 0 184 46 9 1 82 154 2 In the Pessimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .50), Uncertain (πU = .20), and Solvent (πS = .30). The Highest Transparency Pessimistic setting had 24 participants, however, one participant did not respond in set one. * Risk-dominant equilibrium.

22

Table 7. Payoffs Panel A. Frequency of Payoffs and Mean Payoff per Decision Rule in the Optimistic Settings

Low Transparency Payoffs (n= 2,400 participant rounds)

Higher Transparency (n= participant rounds)

Highest Transparency (n= participant rounds)

Decision Rule $0.00 $0.20 $0.50 Mean Pay $0.00 $0.20 $0.50 Mean Pay $0.00 $0.20 $0.50 Mean PayForeclose always (FA) 0 10 0 0.200 0 0 0 0 0 0 0 0Foreclose if Low or Medium Signal (FM) 0 156 34 0.254 0 997 733 0.327 0 66 34 0.302Foreclose if Low Signal (FL) 411 295 1374 0.359 231 57 322 0.283 48 385 1357 0.422Roll over always (RA) 48 0 72 0.300 32 0 28 0.233 12 0 48 0.400Non-responding 0 0 0 0 0 0 0 0 150 0 0 0 In the Optimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .30), Uncertain (πU = .20), and Solvent (πS = .50).

23

Table 7 (Continued) Panel B. Frequency of Payoffs and Mean Payoff per Decision Rule in the Pessimistic Settings

Low Transparency Payoffs (n= participant rounds)

Higher Transparency (n= participant rounds)

Highest Transparency (n= participant rounds)

Decision Rule $0.00 $0.30 $0.50 Mean Pay $0.00 $0.30 $0.50 Mean Pay $0.00 $0.30 $0.50 Mean PayForeclose always (FA) 0 10 0 0.300 0 90 0 0.300 0 20 0 0.300Foreclose if Low or Medium Signal (FM) 0 1381 209 0.326 0 412 48 0.320 0 1044 476 0.363Foreclose if Low Signal (FL) 226 82 172 0.230 127 849 854 0.373 165 296 379 0.331Roll over always (RA) 16 0 4 0.100 10 0 10 0.250 7 0 3 0.150Non-responding 0 0 0 0 0 0 0 0 10 0 0 0 In the Pessimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .50), Uncertain (πU = .20), and Solvent (πS = .30).

24

Table 7 (continued)

Panel C. Total Payoff Per Participant, Excluding $10 Show-Up Fee N Min Max Mean S.D.

Low Transparency 24 $29.50 $36.80 $34.68 $1.87Higher Transparency 24 $28.20 $33.80 $31.35 $1.29

Optimistic Settings

Highest Transparency 21 $12.20 $42.60 $38.56 $8.90Low Transparency 21 $23.40 $33.00 $30.21 $2.18Higher Transparency 24 $29.80 $38.90 $35.89 $2.10

Pessimistic Settings

Highest Transparency 24 $30.90 $36.80 $34.88 $1.65

In the Optimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .30), Uncertain (πU = .20), and Solvent (πS = .50). In the Pessimistic setting, the prior probabilities (π) for the three states of nature are as follows: Bankrupt (πB = .50), Uncertain (πU = .20), and Solvent (πS = .30).

25

APPENDIX

ID #: ______

Instructions

Thank you for participating in this experiment. Your identification number is printed at the top of this page. This experiment is concerned with decision making in a market setting. You are guaranteed to receive $10 for showing up on time. By following the instructions carefully and making good decisions, you may earn a considerable amount of money in addition to the $10. This additional money will range from approximately $20 to $50. The actual amount of additional money that you may earn will depend on your decisions and the decisions of other participants. Your money will be paid to you in cash after the experiment ends. You will need to sign and date a compensation receipt form before you receive your payment. Throughout the experiment, please observe the following rules during the instructions and experiment:

1. Do not talk to any other participant at any time. The screens between the participants are to maintain privacy.

2. You will use your mouse to select decisions during the experiment. Do not use your

mouse or keyboard to play around with the software running on your computer. If you unintentionally or intentionally close the software program running on your computer, you will be asked to leave. If this happens, you will receive only your $10 fee for showing up.

3. If you have any questions, raise your hand. An experimenter will answer your questions

privately. 4. At the end of the instructions, you will take a short quiz to verify that you understand the

experimental procedures. For easy reference, your computer screen will display the tables and figures in these instructions. Right now, your computer screen should be blank.

1

Your Role in the Experiment

Throughout the experiment, you will be randomly grouped with 7 other participants. You and the other participants will each be a creditor to a company to whom you have lent money. The company has borrowed your money in order to fund a project. Each creditor (including you) will decide whether to (1) Continue to invest his or her money in the company until the project is complete, or (2) Withdraw his or her money before the project is complete. The experiment will consist of multiple sets. Each set has 10 rounds. For example, two sets include a total of 20 rounds, and 40 sets include a total of 400 rounds. You will not know how many sets will be played until the experiment ends. Your Payoffs As shown in Table 1, your payoffs in each round will depend on your decision (Continue or Withdraw) and whether the project succeeds or fails. When you make your decision, you will not know for certain whether the project will succeed or fail. Please take a moment to review Table 1:

Table 1: Payoff Schedule (Per Round)

Project Succeeds Project Fails You Continue $0.50 $0.00 You Withdraw $0.20 $0.20

Table 1 appears in the upper left corner of your computer screen.

Below, please write down your answers to the following questions on this page about Table 1. In a few moments the experimenter will review the correct answers with you. 1. If you withdrew your money in an experimental round, how much money would you receive? ________________________________________________________ 2. If you continued to invest your money in an experimental round, what would your payoff depend on? _______________________________________________ __________________________________________________________________

2

Project Success or Failure

There are three possible states of nature: Solvent, Uncertain, and Bankrupt. Project success is guaranteed when the state of nature is Solvent. Project failure is guaranteed when the state of nature is Bankrupt. When the state of nature is Uncertain, the project succeeds only if at least 5 creditors continue to invest. Only one state of nature will occur in each experimental round. Table 2 shows the percentage of times each state of nature will occur, and the project outcome for each state of nature. Table 2 appears below Table 1 in the upper left corner of your computer screen. Please take a moment to review this table:

Table 2: Percentage of times each state of nature will occur and the project outcome for each state of nature.

State of nature Percentage of times each state of nature will occur Project Outcome

Solvent 55% Success is guaranteed

Uncertain 25% Project succeeds if at least

5 out of 8 creditors continue to invest

Bankrupt 20% Failure is guaranteed

Table 2 appears below Table 1 in the upper left corner of your computer screen. Below, please write down your answers to the following questions on this page about Table 2. In a few moments the experimenter will review the correct answers with you. 1. If you were to participate in this experiment for 100 rounds, how many times would you expect

A. the Solvent state of nature to occur? ________ times B. the Bankrupt state of nature to occur? ________ times C. the Uncertain state of nature to occur? ________ times

2. If the state of nature is Solvent and 4 out of 8 creditors continued to invest their money, what would be the project outcome? (check one:) _____ Success _____ Failure 3. If the state of nature is Bankrupt and 6 out of 8 creditors continued to invest their money, what would be the project outcome? (check one:) _____ Success _____ Failure 4. If the state of nature is Uncertain and 6 out of 8 creditors continued to invest their money, what would be the project outcome? (check one:) _____ Success _____ Failure 5. If the state of nature is Uncertain and 4 out of 8 creditors continued to invest their money, what would be the project outcome? (check one:) _____ Success _____ Failure

3

Your Clues about the State of Nature Each experimental round, a computer will randomly select one of the three states of nature based on the percentages shown in Table 2. You will not observe which state of nature the computer has selected. Instead, you will observe one of three clues (Low, Medium, or High) about the state of nature. Each creditor will observe only his or her clue. After you receive your clue, you will decide whether to continue to invest or withdraw your money. Figure 1 shows the clues and project outcomes for each state of nature. Figure 1 appears below and on a separate sheet of paper for your convenient reference. Please follow along as we describe Figure 1:

.

States of Nature Clues Project Outcome

Bankrupt

Uncertain

Solvent

Low

Medium

Medium

Medium

High

Failure

Failure

Success

Success

At least 5 out of 8 creditors continue

Fewer than 5 out of 8 creditors continue

Failure

Success

20%

25%

55%

100%

80%

20%

80%

20%

Figure 1. States of Nature, Clues, and Project Outcomes

4

Below, please write down your answers to the following questions on this page about Figure 1. In a few moments the experimenter will review the correct answers with you. 1. If you receive the Low clue, what state of nature occurred?

a. Solvent b. Uncertain c. Bankrupt d. cannot be determined with certainty

2. If you receive the Medium clue, what state of nature occurred?


3. If you receive the High clue, what state of nature occurred?


4. If you receive the Low clue,

A. is it possible that others will receive the Medium clue? _____ Yes _____ No

B. is it possible that others will receive the High clue? _____ Yes _____ No

5. If you receive the Medium clue,

A. is it possible that others will receive the Low clue? _____ Yes _____ No

B. is it possible that others will receive the High clue? _____ Yes _____ No

6. If you receive the High clue,

A. is it possible that others will receive the Low clue? _____ Yes _____ No

B. is it possible that others will receive the Medium clue? _____ Yes _____ No

5

Making Your Decision to Continue or Withdraw

You will choose a decision rule to indicate your decisions in the experiment. Your decision rule will specify which clues will lead you to continue to invest your money and which clues will lead you to withdraw your money. After you receive your private clue in an experimental round, your decision rule will be executed by the computer. You can choose one of four decision rules:

Decision Rule Select a Decision Rule by Clicking the Appropriate Box 1 Continue always, regardless of your clue 1 Withdraw if your clue is Low, otherwise continue 1 Withdraw if your clue is Low or Medium, otherwise continue 1 Withdraw always, regardless of your clue

The decision rules are displayed in the upper right corner of your computer screen. During the experiment you will indicate your choice by clicking on the appropriate box. Choose your decision rule carefully because it will remain in effect for 10 experimental rounds. After that, you may revise your decision rule for the next set of 10 rounds, and so on.

How Your Decision Rule Will Be Executed

Your decision rule will be executed in the following manner for each experimental round:

1. The computer will use the percentages in Table 2 and Figure 1 to generate a state of nature and private clues for each creditor.

2. The computer will compare your private clue with your decision rule to determine

whether you continue to invest or withdraw your money for that round.

3. The computer will tabulate how many creditors continue to invest, determine whether the project fails or succeeds, and calculate your payoff for that round.

4. After 10 such rounds, you may revise your decision rule for the next 10 rounds.

6

Below, please write down your answers to the following questions on this page about the decision rules. In a few moments the experimenter will review the correct answers with you. 1. For how many experimental rounds will your decision rule apply? ______ rounds 2. Suppose you selected the decision rule which states “Withdraw if your clue is Low, otherwise continue.” A. If you receive the Medium clue, what would your decision be?

_____ Continue _____ Withdraw

B. If you receive the Low clue, what would your decision be? _____ Continue _____ Withdraw

3. Suppose you selected the decision rule which states “Withdraw if your clue is Low or Medium, otherwise continue.” A. If you receive the Low clue, what would your decision be?

_____ Continue _____ Withdraw

B. If you receive the High clue, what would your decision be? _____ Continue _____ Withdraw

C. If you receive the Medium clue, what would your decision be? _____ Continue _____ Withdraw

7

Market Activity and Payoff Report The middle of your computer screen will report the market activity and payoff information in Table 3. This table will be displayed as shown below. For this example, suppose that you chose the decision rule which states “Withdraw if your clue is Low, otherwise continue” and that Round 4 has just been completed. Please take a few moments to verify that you understand the information in Table 3:

Table 3: Market Activity and Payoff Report

Round

Your Clue

Your Decision

Project Status

Your Payoff

1 2 3 4 5 6 7 8 9 10

High Low

Medium Medium

Continue Withdraw Continue Continue

SuccessFailure Failure Success

$0.50 $0.20 $0.00 $0.50

Table 3 is displayed in the middle of your computer screen and will be updated after each experimental round.

8

Set Choices and Payoffs After each experimental round, the computer will report your Set Choices and Payoffs in Table 4. This table will appear at the bottom of your computer screen. The table will be displayed as follows:

Table 4: Set Choices and Payoffs

Set Decision Rule

Payoff Per Set

Experiment Cumulative Payoff

1

A brief description of your decision rule will appear in this column

Your payoffs for the most recent set will be updated in this column after each experimental round.

Your cumulative payoff for the experiment will be updated in this column after each experimental round.

2 etc.

Table 4 is displayed at the bottom of your computer screen.

9

Summary of Experimental Procedures The experiment consists of multiple sets of 10 rounds. You will not know how many sets will be played until the experiment ends. Each set of 10 rounds will consist of the following steps:

1. You will be randomly and anonymously grouped with seven other creditors. For each set of 10 rounds you will be grouped with a different set of seven creditors. If the number of creditors exceeds an exact multiple of 8, the computer will randomly select the remaining creditors to serve as alternates for that set. The computer will randomly select different alternates for each set. Each alternate will be randomly grouped with seven other creditors in order to receive payoffs as fully participating creditors.

2. You will privately select your decision rule for the entire set of 10 rounds. 3. In each of the 10 rounds, the computer will select the state of nature and your private

clue. Your investment decision will be executed on your behalf based on your decision rule.

4. For each round, your payoff will be computed based on your investment decision and

whether the project succeeded or failed. 5. At the end of each round, you will receive a Market Activity and Payoff Report.

6. Steps 3 –5 will be repeated for each of the 10 rounds in the set.

Limited Time to Make Decisions

Before the first set of 10 rounds commences, you will have 2 minutes to select a decision rule. For subsequent sets, you will have 1 minute to select a decision rule. You may change your selection any time during this allotted time period. Your computer monitor will display a reminder message thirty seconds before your allotted decision time expires. When the allotted time expires, the computer will execute your decision rule on your behalf for 10 rounds. If you fail to select a decision rule within the allotted time for a particular set, you will not participate in that particular set and your payoff for that set will be $0.00.

Conclusion of the Experiment At the conclusion of the experiment, an experimenter will ask you to please quietly remain seated. To receive your earnings, you must first hand in your instructions booklet (including the questionnaire on the following pages) and sign a payment receipt form.

10

ID #: ______________

QUESTIONNAIRE ON EXPERIMENTAL PROCEDURES The amount of money you will earn in this experiment will depend on how well you understand the market procedures. Please take the following quiz to verify that you understand the procedures. Raise your hand when you are finished. The experiment will begin after each participant completes the quiz. 1. In an experimental round, how much money will you earn if you continue to invest your money in the company and

a. the project succeeds? $__________

b. the project fails? $__________

2. In an experimental round, how much money will you earn if you withdraw your money from the company and

a. the project succeeds? $__________

b. the project fails? $__________ 3. For an experimental round, when will you find out whether the project actually succeeded or failed?(circle one answer)

a. At the beginning of the experimental round b. When you receive your private clue about the state of nature in that experimental round. c. When your decision rule determines whether you will continue or withdraw your money

in that experimental round. d. When the Market Activity and Payoff Report is updated in Table 3 in the middle of your

computer screen. 4. Complete the following table, assuming that you chose the decision rule “Withdraw if your clue is Low or Medium, otherwise continue”, 6 out of 8 creditors in your group continued to invest, and the state of nature was Uncertain:


Round

Your Clue

Your Decision

Project Status

Your Payoff

4

Medium

11

12

5. Complete the following table, assuming that you chose the decision rule “Withdraw if your clue is Low, otherwise continue”, 4 out of 8 creditors in your group continued to invest, and the state of nature was Uncertain:


Round

Your Clue

Your Decision

Project Status

Your Payoff

16

Medium

6. Complete the following table, assuming that you chose the decision rule “Withdraw if your clue is Low, otherwise continue”, 6 out of 8 creditors in your group continued to invest, and the state of nature was Bankrupt:


Round

Your Clue

Your Decision

Project Status

Your Payoff

21

Medium

7. Complete the following table, assuming that you chose the decision rule “Withdraw always”, 6 out of 8 creditors in your group continued to invest, and the state of nature was Solvent:


Round Your Clue

Your Decision

Project Status

Your Payoff

58

High

8. Complete the following table, assuming that you chose the decision rule “Continue always”, five creditors continued to invest, and the state of nature was Bankrupt:


Round

Your Clue

Your Decision

Project Status

Your Payoff

97

Low

AFTER THE EXPERIMENT ENDS, YOU WILL BE ASKED TO HAND IN YOUR INSTRUCTIONS BOOKLET AND THIS QUESTIONNAIRE BEFORE YOU RECEIVE YOUR PAYMENT.

Information Transparency and Coordination Failure: Efficiency,...

Documents

Transcript of Information Transparency and Coordination Failure: Efficiency,...