An experimental investigation of trust and sequential trade.
Deck, Cary A.
1. Introduction
To study a complex phenomenon, economists often reduce it to a
tractable abstract problem. For example, the trust game of McCabe and
Smith (2000) is a model of sequential trade with no external contract
enforcement as described by Coricelli, McCabe, and Smith (2000). In the
standard trust game, the first mover can decide to either end the game
with both parties receiving $10 or continue the game with the total
payoff increased to $40. If the first mover does not end the game, the
second mover can decide to either keep the entire $40 or keep $25 and
return $15 to the first mover. The first mover can be viewed as a
stylized seller who incurs a $10 cost to provide an item valued by the
buyer at $30 in exchange for a $15 payment.
The voluntary trade is mutually welfare improving, but it requires
the first party to forego something of value and risk not having the
transaction completed; thus the first mover must trust that the second
mover is trustworthy. (1) Behavior in the trust game is fairly robust;
Cox and Deck (2005) and Gillies and Rigdon (2008) replicate the results
of McCabe and Smith (2000). Approximately half of the first movers trust
and about two-thirds of the second movers are trustworthy. However,
second-mover behavior is sensitive to the level of social distance. Cox
and Deck (2005) report a reversal of second-mover behavior when the
experimenters could not identify who took what action using a
double-blind protocol.
What does behavior in the trust game say about behavior in
naturally occurring trading opportunities? A comparison of the
well-understood simplified version of the game and the original problem
of interest identifies additional complicating features appropriate for
further investigation in an attempt to answer this question. The trust
game differs from naturally occurring trades in several ways. When
people consider a trade, they weigh the gain from a successful trade
against the loss from an unsuccessful one. (2) The price determines the
size of the potential gain and loss. In the naturally occurring economy,
prices are formed through some endogenous process; yet prices are
exogenously fixed by the experimenter in the trust game. In the
laboratory, payoff information is usually public; whereas in practice a
seller does not know the willingness to pay of the buyer, nor does the
buyer know the seller's cost. Experiments by McCabe, Rassenti, and
Smith (1998) and Gillies and Rigdon (2008) demonstrate that the effect
of payoff privacy is behavior more consistent with the traditional
self-interested model than what is typically observed with public
information. Further, previous trust game experiments have been
intentionally abstract. Eckel and Grossman (1996) argue "...that
social and psychological factors affect economic decision making, and
the importance of social factors can only be introduced by abandoning,
at least to some extent, abstraction" (p. 189). Direct evidence of
the effect of a buyer-seller framing in the related ultimatum game is
mixed (see Hoffman et al. 1994; Cox and Deck 2005). (3)
This article moves towards a more complete story of sequential
trade by imbedding endogenously determined variations of the trust game
into a richer experimental environment with payoff privacy and
buyer-seller framing. The next section details the experimental design
and a separate section presents the behavioral results, which differ
from previous work. A fourth section explores which factors cause the
behavioral shift and a final section concludes.
2. Experimental Design and Procedures
The experiments involve variations of the trust game and a related
game where the order of play is reversed. In the standard trust game,
the second-mover buyer providing payment makes a one for one transfer to
the first-mover seller. When the order is reversed so that the buyer
initiates trade, then the second-mover seller makes a one for three
transfer to the first mover. The difference is that surplus is generated
when the good is exchanged, but not when the payment is made. (4) Figure
1 describes the seller-first trust game (left panel) and the buyer-first
modified trust game (right panel) in terms of endowments of the buyer
and seller ([E.sub.B] and [E.sub.s], respectively), value to the buyer
(V), cost to the seller (C), and the price (P). In the experiments, the
induced value of the good was V = $15, and the induced cost of producing
and "shipping" the good was C = $5. The buyer was endowed with
[E.sub.B] = $15, and the seller was endowed with [E.sub.s] = $10. (5) As
was explained to the subjects, all amounts are in U.S. dollars.
[FIGURE 1 OMITTED]
The experiment was framed as an opportunity for a buyer and seller
to trade. The directions used the terms buyer, seller, value, cost,
price, and so on. (6) The trust game was presented in extensive form
with the branches labeled "Pay" and "Not Pay" for
the buyer and "Ship" and "Not Ship" for the seller.
Figure 2 shows an example screen for a buyer who is moving second.
Payoff information was private. That is, a buyer knew her own endowment
and value but not the endowment or cost of the seller. (7) Similarly, a
seller knew his cost and endowment but not the buyer's value and
endowment. As shown in Figure 2, buyers observed E and C, denoting the
seller's endowment and cost. The buyer knew that her failure to pay
resulted in the seller earning E - C while payment resulted in the
seller receiving E - C + the price.
Previous trust game experiments have assumed a fixed price. In the
current experiments, prices were determined through a bargaining
process. Subjects were randomly matched with someone in the opposing
role and had five minutes to negotiate a price. During the bargaining
process, subjects could adjust the game displayed on their screen to
reflect any price between 0 and 25. Figure 2 shows the decision tree for
a price of P = 5. There was no imposed order of offers, nor was there an
improvement rule, but sellers could not suggest a price below the
current price proposed by the buyer, nor could a buyer suggest a price
above the current price proposed by the seller. Prices were required to
be whole dollar amounts, and a no bankruptcy condition was imposed.
Sellers could not post or agree to a price below C = $5, and buyers
could not post or accept a price above V = $15, but this was private
information. A contract was only reached when one party accepted the
price put forward by the other party. If time expired without a contract
being reached, both subjects received their respective endowments.
[FIGURE 2 OMITTED]
A price agreement depends upon the first mover's beliefs that
the trade will be completed at that price as well as the potential
payoffs associated with the price. If first movers anticipate that
second movers are more likely to complete trades where the price favors
the second mover, then first movers find themselves in a situation
similar to a first price auction; one could ask for a larger but less
likely payoff or one that is smaller and more likely. Such a belief is
commonly expressed by the old adage that "if a deal looks too good
to be true, then it probably is." At the same time, the second
mover must agree to the negotiated price as well, and thus the price
reflects the second mover's distributive preferences and intended
action. A second mover that plans to defect would be willing to agree to
any price but would want to act like a well-intentioned second mover so
as not to arouse the suspicion of the first mover. Given the exploratory
nature of this work, no formal prediction is made with respect to the
prices that are likely to emerge.
Even without price predictions, one can hypothesize how price will
impact behavior in the endogenously formed trust game conditional on
price. Consider a second-mover buyer's decision to make a payment.
Cox and Deck (2005) find that second-mover cooperation is higher when
the stakes are lower. Thus, a buyer may be more likely to send a $6
payment than a $14 payment once the good is received. More generally,
one would expect that the lower the agreed-upon price, the more likely a
second-mover buyer is to send payment. It would also seem reasonable
that the larger the payment that has been received, the more willing a
second-mover seller may be to forgo $C due to reciprocal motivations or
increased social indebtedness.
Based upon previous research, one would expect that second-mover
sellers are more likely to complete initiated trades than are
second-mover buyers. Andreoni and Vesterlund (2001) and Andreoni and
Miller (2002) study dictator games where $1 given up by the dictator
results in the recipient receiving $[alpha], where the conversion rate
[alpha] varies between 1/3 and 3. They find that dictators give more as
[alpha] increases. By making the payment, a second-mover buyer increases
the seller's payoff by SP, exactly the amount that the buyer
forgoes. On the other hand, second-mover sellers give up $C to transfer
$V to the buyer. As long as the trade is strictly mutually beneficial,
it must be that $V > $P > $C. Hence a seller knows that each
foregone dollar generates more than one dollar for the buyer. In Deck
(2009) the conversion rate, [alpha] = V/C = 3, was known, and sellers
were found to be more likely to complete a trade. This pattern may
explain why buyers move first in many trades, as denoted by the common
expression: "Sorry, no CODs." In the current study, a seller
only knows [alpha] = V/C > 1, and thus, this study provides a
robustness test of the relationship between generosity and the
conversion rate as previous studies have found that payoff privacy leads
to more self-interested behavior and may therefore mitigate the
conversion rate effect.
The rate of trust by first movers is approximately 50% in the
standard trust game. Payoff privacy encourages self-interested behavior,
while the negotiated prices might encourage trust since the first mover
has already agreed to the price before acting. The buyer-seller
framework may encourage trust since defection is explicitly reneging on
an agreement. There is no a priori basis for predicting the relative
influence of these factors, and no explicit predictions are made,
reflecting the exploratory nature of this work.
The experiments were conducted at a state university, and subjects
were drawn from the population of business school undergraduates. For
each of the 12 sessions, groups of between 8 and 12 subjects entered the
laboratory and were allowed to sit at any active computer station. (8)
Subjects participated in a single session and had not participated in
any related experiments. The computer stations were separated by privacy
dividers so that no one could see any other subject or any other active
computer screen. Subjects read role--and order-specific directions.
After all subjects had completed the directions and were given the
opportunity to ask questions, additional directions explaining the
double-blind payoff procedures were distributed. The subjects drew
sealed envelopes containing mailbox keys with which they would
anonymously collect their cash earned at the conclusion of the
experiment. Once the experimenters left the room, subjects opened their
envelopes and entered the identification codes on the mailbox keys in
their computers. (9) After all codes were entered, the experiment began.
When the payoffs were determined, money was placed in envelopes, which
were in turn placed in mailboxes. Subjects were allowed to collect their
earnings outside the view of the experimenter and then leave.
Sessions lasted approximately 20 minutes, although subjects were
recruited for half-hour sessions. As was made explicit in the
directions, subjects went through this process once, in only one role
and one order. Failure to trade did not result in another opportunity to
trade with the same or another party.
[FIGURE 3 OMITTED]
3. Experimental Results
A total of 62 subject pairs completed the experiment; 31 pairs per
trading order. Of these 62 pairs, 23% completed a mutually beneficial
trade. The overall rate of trade success did not differ by order based
upon a two-sample proportions test with a null hypothesis of equality
against the two-sided alternative (p = 0.224). Figure 3 shows the
frequency of each outcome by treatment pooling across prices.
In this experiment, a subject pair could fail to make a successful
trade for three reasons: The second mover may not complete an initiated
trade, the first mover may decide to not initiate a trade, and the
parties may not be able to agree upon a price. This third reason is not
present in the typical trust game and accounted for 16% of the total
pairs in the experiment. Four pairs failed to reach a price agreement
when the seller moved first, and six pairs failed to reach an agreement
when the buyer moved first, a difference that is not significant based
upon a two-sample proportions test of equal proportions against the
two-sided null (p = 0.490). It is interesting to note that 79% of the
observed agreements occurred as the result of the second mover accepting
the first mover's price. (10) This proportion is larger than would
be expected to occur randomly (p < 0.001) and is intuitive since the
first mover bears all of the risk.
The majority of trade failures were where the second mover failed
to complete the transaction. Of the second-mover buyers who received the
item, only 38% (9 of 9 + 15) subsequently sent payment. For second-mover
sellers who received payment, only 29% (5 of 5 + 12) went ahead and
shipped the item. Second-mover behavior did not vary with trading order
(p = 0.591 for the test of equal proportions against the one-sided
alternative that sellers were more likely to complete a trade). The
nominal pattern of sellers being less trustworthy contradicts previous
research by Andreoni and Vesterlund (2001), Andreoni and Miller (2002),
and Deck (2009). However, the general pattern that less than half of
second movers are trustworthy is similar to what has been reported
previously (Cox and Deck 2005; Gillies and Rigdon 2008; and Deck 2009).
First movers were considerably more likely to trust their
counterpart and initiate a trade as compared with previously reported
results. Of the first-mover sellers who reached a price agreement, 89%
(15 + 9 of 15 + 9 + 3) shipped the item. Of the first-mover buyers who
reached a price agreement, 68% (12 + 5 of 12 + 5 + 8) sent the payment.
For comparison, these numbers were 56% and 35% for first-mover sellers
and buyers, respectively, in Deck (2009). More than half of the subjects
exhibit trust (p < 0.001, = 0.071 for first-mover sellers and buyers,
respectively). The differences in percentages in the current study and
those reported in Deck (2009) are highly significant (p = 0.002 and
0.004 for first-mover sellers and buyers, respectively). (11) Here order
has a significant impact on trust; sellers were more likely to initiate
trade (p = 0.065). Given that the price could not be less than C = $5,
first-mover sellers risked a smaller loss than did first-mover buyers,
which could explain this result. As discussed above, first movers
overwhelmingly set the price determining the risk and reward from
exchange. First movers who set the price are nominally, but not
statistically, more likely to trust the second mover. The probability
that a first mover will trust is 79% when the first mover sets the price
and 67% when the first mover accepts a price set by the second mover (p
= 0.421 in the two-sided two-sample proportions test).
Figure 4 shows the behavior observed during the experiment for both
orderings at the negotiated prices. The average agreed-upon price is
$10.78 when the seller moves first, and it is $11.36 when the buyer
moves first. This difference is not significant (t-statistic = 0.656).
For comparison, the original trust game provides 75% of the gain to the
second-mover buyer, which translates to a price of $7.50 in this
context. A price of $10 would split the gains evenly regardless of
order. While prices were higher than the equal split level in both
conditions, the difference is not dramatic in economic terms. For
first-mover buyers who actually sent their payment, the average
agreed-upon price was $10.41, and for first-mover sellers who actually
shipped the good it was $11.29. This difference remains insignificant
(t-statistic = -0.940).
Based upon the probit regression results shown in Table 1, there is
evidence that the price impacts the likelihood that a first mover will
initiate trade regardless of role. Sfirst is a dummy variable for the
treatment and takes the value of 1 if the seller moves first and 0
otherwise. Surplus is the dollar amount the first mover would gain from
a successful trade. For a first-mover buyer this variable equals $V
minus the negotiated price and for a first-mover seller it is the
negotiated price minus $C. Thus the estimation results reveal that the
larger the potential return, the more likely subjects are to engage in
trust and this effect does not differ by role.
Although price does impact the likelihood that a trade is
initiated, it does not appear to impact the likelihood that a trade is
completed. The average price that second-mover buyers paid for received
goods was $11.11 while the average price for which second-mover buyers
did not send payment upon receipt of the good was $10.40. This
difference is not significant (U = 122, p = 0.3942). (12) The average
price received by second-mover sellers for goods that were subsequently
delivered was $10.80, and the average payment for orders left unfilled
was $10.25. Again, this difference is not significant (U = 47, p =
0.4165). The probit analysis shown in Table 2 also confirms the
conclusion that price does not impact the likelihood that an initiated
trade is successfully completed. Ssecond is a dummy variable for the
treatment and takes a 1 if the seller moves second and 0 otherwise.
Surplus is measured as before. Based upon estimation results neither of
these variable impacts the likelihood of a successful trade. The
expression that "if a deal looks too good to be true, then it
probably is" holds, technically. Such deals are unlikely to be
true, but less favorable deals are also unlikely to be true.
[FIGURE 4 OMITTED]
4. Further Exploration
The experiments reported in the previous section differ from
standard trust game experiments in several ways including introducing a
buyer-seller frame and negotiated prices. The combined effects of these
changes are minimal for second movers. The observed behavior is similar
to what has been previously reported for games with double-blind
procedures and public payoff information (Cox and Deck 2005) and for
games with single-blind payoff procedures and private payoff information
(Gillies and Rigdon 2008). Data from these papers are shown in Table 3
along with data from the seller first treatment of Deck (2009). It
appears that either payoff privacy or the double-blind procedures leads
to low levels of trustworthiness, and these effects dominate any
tendency to increase cooperation that the buyer-selling framing or
negotiated prices might engender, although second movers are slightly
more trustworthy with the buyer-seller framing.
Where the observed behavior does differ dramatically from previous
experiments is the level of "trust" exhibited by the first
mover. While this rate is normally around 50% even for experiments with
double-blind payoff procedures, Gillies and Rigdon (2008) report that
approximately two thirds of subjects do not trust when payoffs are
private. In the current experiments the opposite behavior occurs; the
vast majority of subjects are trusting.
Both the framing and the negotiated prices could explain the
increase in trust. By framing the decision as a buyer-seller
interaction, a social connotation is associated with not completing the
transaction. First movers may (incorrectly) believe that second movers
are unlikely to explicitly steal. Alternatively, the negotiation process
which requires first movers to agree to the price may increase trust by
setting the level of risk at (or below) a level the first mover is
willing to tolerate. To explore the relative impacts of price
negotiations and the buyer-seller framework, an additional set of
seller-first experiments was conducted with 21 new subject pairs. (13)
These experiments were similar to those reported in the previous section
with the exception that the price was fixed at p = $10.
The results of the additional experiments are shown in the
right-most column of Table 3. A nominally higher, but statistically
indistinguishable 81% of the first movers trusted as compared with the
case with negotiated prices (p = 0.431). For comparison, the rate of
trust is also higher than that observed by Cox and Deck (2005) with
double-blind payoffs, fixed prices, public information, and neutral
framing (p = 0.037); Rigdon and Gilles (2008) with single-blind payoffs,
fixed prices, neutral framing, and private information (p < 0.001);
and Deck (2009) with fixed price, neutral framing, and public
information (p = 0.053). While one should always be cautious of
comparisons across studies, this pattern suggests that the substantial
increase in trust is due to the buyer-seller framing and not the
negotiated prices. (14) Approximately 40% of the second-mover buyers
actually sent payment conditional on receiving the good. As expected
given the lack of a price response by second movers, this is
statistically indistinguishable from the case of negotiated prices (p =
0.812) indicating that negotiated prices do not affect trustworthiness.
The 40% figure is also not statistically different from second-mover
behavior in Cox and Deck (2005), Rigdon and Gilles (2008), or Deck
(2009) with p-values of 0.388, 0.197, and 0.478, respectively.
5. Conclusion
Laboratory experiments are intended to inform researchers about
behavior in the naturally occurring world. The trust game and the
related investment game have received considerable attention from
experimental economists. While previous work has discussed the trust
game as a model of sequential trade, this article reports experiments
that explore this interpretation more directly. Here the game is framed
to the subjects as the interaction of a buyer and a seller. Costs and
values that would normally be private information in a naturally
occurring trade are kept private. The price, which determines how the
potential gains from trade are to be shared, is determined endogenously
in practice and in the experiment. There are some features of the
experiment that are distinct from naturally occurring trades. One is the
order of moves, which serves as a treatment in this study. Others
differences include the double-blind payoff procedure, the use of
induced values, and the total inability of first movers to retaliate.
As compared with the findings of previous studies, first movers
were far more likely to trust regardless of role. Based upon additional
experiments, this surprising result is found to be the result of the
buyer-seller framing rather than the negotiated prices. Second movers
were not trustworthy. This is similar to previous results from
experiments with double-blind payoff procedures or payoff privacy. In
the experiments price did not impact the likelihood that a trade was
successfully completed. Contradicting previous results, the role of the
second mover did not impact the probability that an initiated trade was
completed in this environment where traders could not observe the social
gains from exchange. This suggests that people need precise rather than
vague information about how a choice will benefit others before
undertaking a costly action.
This article demonstrates the need to closely consider the
abstractions that are made when simplifying a problem as argued by Eckel
and Grossman (1996). However, the norm in experimental economics
continues to be that individual choice experiments rely upon
"neutral" language. The abstract trust game has been widely
used to study individual choice behavior in bilateral exchange. The
current results suggest that previous studies have underestimated the
amount of trust that occurs in naturally occurring trading
opportunities. This also suggests that subjects do not see the trust
game as an exchange no matter how well that interpretation seems to fit
in theory. Researchers need to be cautious in interpreting lab results,
not only for the trust game but for other stylized games as well. This
should not be misconstrued as an argument that experiments with abstract
games do not provide valuable insight. The very consistency of behavior
in such games is evidence that they reveal real patterns of human
decision making. Clearly, additional work on the effect of context is
warranted. Hoffman et al. (1994) found a similar behavioral effect by
placing a market frame on the ultimatum game, but this result was not
replicated by Cox and Deck (2005).
Appendix 1: Subject Directions--First-Mover Seller with Price
Negotiation
You are going to participate in an experiment like the one pictured
below in which a "buyer" and "seller" have the
opportunity to trade a fictitious good. In the experiment, you will have
to make decisions that will have a direct impact on your cash
payoff'. The numbers represent the $US amounts that you and your
randomly selected counterpart will be paid at the end of the experiment,
including the $5 show-up fee that each of you is receiving for
participating in this experiment.
You have been assigned the role of seller and your counterpart has
been assigned the role of buyer. Your role will be clearly indicated on
the top of your screen. To assist you, items on your screen that refer
to you are highlighted in yellow, while information regarding your
counterpart is in blue.
[ILLUSTRATION OMITTED]
The seller (you) is endowed with $10 and the buyer (your
counterpart) is endowed with $E. By incurring a cost of $5, you can
produce and ship a good which the buyer values at $V. This information
is contained in the tables at the top of the screen. Your counterpart
does not know your endowment or cost, and since no one knows their
counterpart's identity, "CP" is displayed on the
buyer's table.
As the seller, you will first have to decide to "Ship" or
"Not Ship" the fictitious good. If you decide to "Not
Ship" you will receive $10, and the buyer will receive $E. If you
decide to "Ship" you will incur the $5 cost, and the buyer
will receive $V. At this point, the buyer would have to decide to
"Pay" or "Not Pay." If the buyer decides to
"Pay," you will receive $10 endowment--$5 cost + price = price
+ $5 and the buyer will receive SE + $V - price. If the buyer decides to
"Not Pay," you will receive $10 endowment - $5 cost = $5 and
the buyer will receive SE + $V.
The decision tree on the bottom portion of your screen contains all
of this payoff information. The seller has the first decision and thus
is at the first node of the decision tree. If the seller chooses to
"Ship", this leads to the second node at which the buyer will
have to make a decision. If and when you need to make a decision,
"???" will appear beside each of your choices on the decision
tree. To make a decision, you click on the desired branch in the
decision tree, which will highlight your selection in green. To confirm
your decision you must click the green "Confirm" button that
will appear on your screen.
So how is the price determined? It is determined by you and the
buyer. To suggest a price, you can type it in the blue box at the top of
the screen and press the "Propose Price of" button. After you
do this, the price you proposed will be displayed as your proposed
price, and the buyer will be able to accept it. Any price the buyer
proposes will be displayed in the purple box at the top of your screen,
and you can accept it by clicking on the "Accept Price of"
button. If either of you accepts the price the other person proposed,
the decision tree will reflect the agreed-upon price and the experiment
will proceed as described above. While determining a price, you can see
the decision tree for any price by using the dropdown tool just above
the decision tree. The example screen image shown above has a price of
$15. if neither of you accept a price within five minutes, you will
receive $10, your counterpart will receive $E, and neither of you will
have any further decisions to make in this experiment. A clock counts
down the remaining seconds.
You will only go through this process only once during this
experiment. After all participants' earnings have been determined
as described above, you will receive your money and be dismissed from
the experiment. If you have any questions, please raise your hand;
otherwise please wait quietly.
References
Andreoni, James, and John Miller. 2002. Giving according to GARP:
An experimental test of the consistency of preferences for altruism.
Econometrica 70:737-53.
Andreoni, James, and Lise Vesterlund. 2001. Which is the fair sex?
Gender differences in altruism. Quarterly Journal of Economies
116:293-312.
Coricelli, Giorgio, Kevin McCabe, and Vernon Smith. 2000.
Theory-of-mind mechanism in personal exchange. In Affective minds,
edited by G. Hatano, N. Okada, and H. Tanabe. Amsterdam, Netherlands:
Elsevier Science, pp. 249-60.
Cox, James C., and Cary A. Deck. 2005. On the nature of reciprocal
motives. Economic Inquiry 43:623-35.
Deck, Cary A. 2009. An experimental analysis of cooperation and
productivity in the trust game. Experimental Economics 12:1-11.
Eckel, Catherine, and Philip Grossman. 1996. Altruism in anonymous
dictator games. Games and Economic Behavior 16:181-91.
Gillies, Anthony, and Mary Rigdon. 2008. Epistemic conditions and
social preferences in trust games. University of Michigan Working Paper.
Hoffman, Elizabeth, Kevin A. McCabe, Keith Shachat, and Vernon L.
Smith. 1994. Preferences, property rights, and anonymity in bargaining
games. Games and Economic Behavior 7:346-80.
Houser, Dan, Daniel Schunk, and Joachim Winters. 2006. Trust games
measure trust. University of Munich Discussion Paper 2006-38.
McCabe, Kevin A., Stephen J. Rassenti, and Vernon L. Smith. 1998.
Reciprocity, trust, and payoff privacy in extensive form bargaining.
Games and Economic Behavior 24:10-24.
McCabe, Kevin A., Mary L. Rigdon, and Vernon L. Smith. 2003.
Positive reciprocity and intentions in the trust game. Journal of
Economic Behavior and Organization 52:267-75.
McCabe, Kevin A., and Vernon L. Smith. 2000. A comparison of naive
and sophisticated subject behavior with game theoretic predictions.
Proceedings of the National Academy of Sciences 97:3777-81.
Cary A. Deck, Department of Economics, Walton College of Business,
University of Arkansas and Economic Science Institute, Chapman
University, 402 WCOB, 1 University of Arkansas, Fayetteville, AR 72701,
USA; E-mail cdeck@walton. uark.edu.
The author wishes to thank two anonymous referees and Bart Wilson
for extremely helpful suggestions and Taylor Jaworski for research
assistance. Support from the Center for Retailing Excellence at the Sam
M. Walton College of Business is gratefully acknowledged.
Received March 2008; accepted June 2009.
(1) Various institutions have evolved to protect against such
risks. These institutions include the court system, escrow accounts, and
reputation mechanisms such as the one used on eBay. However, there are
numerous (potential) trading situations in which these institutions are
not available or practical. For example, when buying a ticket from an
illegal scalper before a sporting event, the seller could simply pocket
the money, leaving the buyer little recourse. A person purchasing
narcotics on the street runs the risk that the received goods are
impure. Undocumented workers may not receive promised wages. A handyman
may receive money for supplies and never complete a project. Restaurant
owners are exposed to the risk that a customer will "dine and
dash" without paying. The cost of legal proceedings makes it
unlikely one party would sue if an item priced at a few dollars turned
out to be defective. Measuring the value of trades that are not
completed or even initiated due to such incomplete contracting concerns
would be difficult.
(2) Recent evidence by Houser, Schunk, and Winter (2006) indicates
that subjects view the decision to trust differently from a lottery with
a similar payoff structure.
(3) In the ultimatum game the first mover proposes an allocation of
a fixed pie. The second mover can accept this allocation or reject it,
in which case both sides receive 0. The first mover can be considered a
seller (buyer) setting a price, and the second mover can be considered a
buyer (seller) making a purchase (sell) decision.
(4) See McCabe, Rigdon, and Smith (2003) for a variation of the
trust game in which the transaction can be thought of as barter with
surplus generated at both stages.
(5) In the original trust game of McCabe and Smith (2000), the
seller moved first and the parameters were [E.sub.B] = [E.sub.S] = $10,
V = $30, C = $10, and P = $15. See Deck (2009) for a discussion of the
impact on behavior of changes to these parameters including asymmetric
endowments.
(6) The Appendix contains the directions for a subject in the role
of a first mover seller. Copies of the other directions arc available
from the author upon request.
(7) Subjects did not receive a separate participation payment; the
$5 promised in recruiting was included in their endowments as was made
explicit in the directions (see Appendix). In this respect, it was
public information that there was a minimum endowment.
(8) Each session contained male and female subjects, although
precise demographic information was not collected.
(9) From an adjoining room the experimenter could monitor that
subjects did not talk during the experiment without compromising
privacy.
(10) Due to an error in the computer program, the negotiation data
were lost for some sessions.
(11) The subject pool, laboratory, and double-blind payoff
procedures were the same as those used in Deck (2009).
(12) The comparison of first mover behavior is based upon a t-test.
Given the relatively small number of observations, comparisons of
second-mover behavior are based upon the nonparametric Mann-Whitney
test. In both cases, the null hypothesis is no effect, and the
alternative is two-sided.
(13) The purpose of the additional experiments is to explore the
effects of various aspects of the experimental design. The seller-first
format is comparable to existing literature, thereby allowing more
general comparisons. It is possible that payoff privacy, price
negotiations, or the buyer-seller framing interact with the player
order, although there is no a priori reason to suspect they do.
(14) It is possible that both the framing and negotiations effects
are the same but have no interaction effect.
Table 1. Probit Regression for Likelihood That First Mover Initiates
Trade
Variable Coefficient Standard Error z-Statistic p-Value
Sfirst -0.21 0.84 -0.25 0.80
Surplus 0.19 0.09 2.09 0.04
Sfirst x Surplus 0.21 0.24 0.87 0.39
Constant -0.15 0.39 -0.41 0.68
Table 2. Probit Regression for Likelihood That Second Mover
Completes Trade
Standard
Variable Coefficient Error z-Statistic p-Value
Ssecond -0.48 0.81 -0.60 0.55
Surplus -0.08 0.10 -0.85 0.40
Ssecond x Surplus 0.12 0.15 0.82 0.41
Constant -0.26 0.44 -0.59 0.56
Table 3. Comparison of Behavior Across Studies
Cox Gillies
and and
Deck Rigdon Deck
(2005) (2008) (2009)
Double blind Yes No Yes
Private payoffs No Yes No
Fixed price Yes Yes Yes
Buyer seller framing No No No
Percent of "trusting"
first movers 52% 33% 56%
Percent of "trustworthy"
second movers 29% 20% 30%
Current Current
Article Article
Double blind Yes Yes
Private payoffs Yes Yes
Fixed price No Yes
Buyer seller framing Yes Yes
Percent of "trusting"
first movers 77% 81%
Percent of "trustworthy"
second movers 38% 41%
