Abstract
In this paper we analyze a cheap-talk model with a partially informed receiver. In clear contrast to the previous literature, we find that there is a case where the receiver’s prior knowledge enhances the amount of information conveyed via cheap talk. The point of departure is our explicit focus on the “dual role” of the sender’s message in our context: When the receiver’s prior belief is subject to higher-order uncertainty, the sender’s message provides information not only about the true state but also about the reliability of the receiver’s private information. Building on this result, we argue that whether information acquisition and communication are complements or substitutes depends crucially on the extent of uncertainty regarding the information source.
Similar content being viewed by others
Notes
For expositional clarity, we say that information acquisition and communication are complements (substitutes) when more precise prior knowledge of the receiver facilitates (obstructs) communication.
An exception is Moreno de Barreda (2013) who shows that information acquisition and communication can be complements when the receiver is sufficiently risk averse. As we will detail later, the underlying mechanism of this paper is totally different as it has nothing to do with the receiver’s risk aversion. Also, in settings that are somewhat different from CS, some previous works show that the receiver can extract more information as she becomes more informed (Seidmann 1990; Watson 1996; Olszewski 2004). Seidmann (1990) and Watson (1996) consider an environment in which the sender’s payoff function does not satisfy the single-crossing property. They present examples in which full information revelation can be attained when the receiver is endowed with her own private information whereas it is impossible otherwise. Olszewski (2004) shows that the sender’s concern to be perceived as honest, combined with the receiver’s private information, induces truth-telling as a unique equilibrium.
In this context, we say that the receiver is more informed when her signal is more likely to come from the more informative source.
As in most cheap-talk models, we assume that the sender knows the true state with precision.
In a discrete-state model, Ishida and Shimizu (2016) also show that the receiver’s prior information becomes an impediment to efficient communication. In contrast, there are also a few works which analyze a case where the sender (or senders) has only partial information (Agastya et al. 2014; Ambrus and Lu 2014).
Note, however, that the main focus of her analysis is to show the existence of non-monotone equilibria and investigate whether the receiver can truthfully report to the sender about her signal.
To analyze a setup with many senders (many-to-one communication), Morgan and Stocken (2008) assume that signals are distributed according to a Beta distribution. Galeotti et al. (2013) also extends this setting to a network (many-to-many communication) and obtain a similar conclusion in the case of private communication.
The key to their argument is that the state space is truncated into a shorter interval after the receiver consults one of the senders. This is a particular form of what we refer to as the information effect, which by itself lowers the quality of communication. In fact, when the senders have like biases, there is never a monotonic equilibrium that is informationally superior. They show, however, that the receiver can combine the senders’ messages to extract more information when the senders have opposing biases.
With a similar reasoning, Kawamura (2011) shows that the most informative equilibrium converges to binary communication as the number of respondents increases because binary messages do not allow exaggeration.
Aside from this, however, our information structure differs substantially from Mandler (2012) due to the difference in objectives: strategic information transmission in our model and information aggregation through strategic voting in Mandler (2012). The most notable difference is that in our model, there is another player—the sender—who observes the true state, so that the receiver can learn from the sender’s message not only about the true state but also about which distribution the observed signal is more likely to come from. This draws clear contrast to Mandler (2012) where all voters face the same (though uncertain) signal distribution.
In contrast, we assume that the sender has perfect information about the realized state. This is mainly for expositional clarity, as the main logic of the model, especially the confirmation effect, survives even when the sender’s information is noisy in the same way as the receiver’s.
The need for this focus arises from a special feature of cheap-talk models with an informed receiver: as demonstrated by Chen (2009), there may exist a non-monotone equilibrium when the receiver is endowed with some information of her own which is correlated with the sender’s. We rule out this possibility given the question we set out to solve, although it is certainly intriguing as a theoretical possibility. Note that CS shows that in the case of \(q =0\), there exists only monotone partition equilibria.
To be more precise, Moreno de Barreda (2013) obtains this result when the noise distribution is also uniform. Also, as can easily be expected, the risk effect may dominate the information effect as the sender becomes more risk averse. This is in fact what is found in Moreno de Barreda (2013) who shows through a numerical example that the risk effect can dominate the information effect when the sender is sufficiently risk averse.
To see this, suppose that the more accurate information source is also subject to noise, and consider a boundary type who is exactly on the threshold. In this case, this boundary type has less to “confirm” because she has no strong belief about which message is more likely even if her signal comes from the more accurate source, rendering the confirmation effect weaker and more ambiguous. In general, there is more to confirm for types in the interior of an interval.
Among other things, the continuous state space implies that the receiver’s signal can convey information with infinitely fine precision, i.e., each player can unambiguously distinguish a state t from any arbitrarily close one \(t+\varepsilon \). We argue that it is somewhat implausible in many applications.
This condition is satisfied by any CS partition.
The second inequality holds under the assumption that \(b <0.5 \).
\( A _{ n + 1 } \) is also considered as a function of \( \lambda _{ n + 1 } \)
References
Agastya, M., Bag, P.K., Chakraborty, I.: Communication and authority with a partially informed expert. RAND J. Econ. 45(1), 176–197 (2014)
Ambrus, A., Lu, S.E.: Almost fully revealing cheap talk with imperfectly informed senders. Games Econ. Behav. 88, 174–189 (2014)
Austen-Smith, D.: Interested experts and policy advice: multiple referrals under open rule. Games Econ. Behav. 5(1), 3–43 (1993)
Chen, Y.: Communication with Two-Sided Asymmetric Information. Mimeo, (2009). https://doi.org/10.2139/ssrn.1344818
Chen, Y.: Value of public information in sender–receiver games. Econ. Lett. 114(3), 343–345 (2012)
Crawford, V.P., Sobel, J.: Strategic information transmission. Econometrica 50(6), 1431–1451 (1982)
Galeotti, A., Ghiglino, C., Squintani, F.: Strategic information transmission in networks. J. Econ. Theory 148(5), 1751–1769 (2013)
Ishida, J., Shimizu, T.: Cheap talk with an informed receiver. Econ. Theory Bull. 4, 61–72 (2016)
Kawamura, K.: A model of public consultation: why is binary communication so common? Econ. J. 121(553), 819–842 (2011)
Kawamura, K.: Eliciting information from a large population. J. Public Econ. 103, 44–54 (2013)
Krishna, V., Morgan, J.: A model of expertise. Quart. J. Econ. 116(2), 747–775 (2001)
Lai, E.K.: Expert advice for amateurs. J. Econ. Behav. Organ. 103, 1–16 (2014)
Mandler, M.: The fragility of information aggregation in large elections. Games Econ. Behav. 74(1), 257–268 (2012)
Moreno de Barreda, I.: Cheap Talk with Two-Sided Private Information. Mimeo (2013)
Morgan, J., Stocken, P.C.: Informal aggregation in polls. Am. Econ. Rev. 98(3), 864–896 (2008)
Olszewski, W.: Informal communication. J. Econ. Theory 117(2), 180–200 (2004)
Seidmann, D.J.: Effective cheap talk with conflicting interests. J. Econ. Theory 50(2), 445–458 (1990)
Watson, J.: Information transmission when the informed party is confused. Games Econ. Behav. 12(1), 143–161 (1996)
Author information
Authors and Affiliations
Corresponding author
Additional information
This paper was previously circulated under the title “Can More Information Facilitate Communication?” We thank Shinsuke Kambe and Dan Sasaki for helpful comments on earlier versions of the manuscript. The first author acknowledges financial support from JSPS KAKENHI Grant-in-Aid for Scientific Research (S) 15H05728, (A) 20245031, and (C) 24530196 as well as the program of the Joint Usage/Research Center for Behavioral Economics at ISER, Osaka University. The second author acknowledges financial support from JSPS KAKENHI Grant-in-Aid (C) 26380252, (C) 17K03624 and the Kansai University Domestic Researcher, 2014. Of course, any remaining errors are our own.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Appendices
Appendix A: the proofs
Proof of Proposition 1
The proof is quite standard and closely follows that of Proposition 1 of Moreno de Barreda (2013). We relegate the proof to Online Appendix. \(\square \)
Proof of Proposition 2
The local incentive condition is necessary for an MPE but not sufficient. To show the existence of an MPE, we need to show that the sender has an incentive to follow the equilibrium strategy for all t.
First, consider \(t\in T_n\). If \(\lim _{ t \uparrow t _n } {\partial \Delta \over \partial t} \ge 0\), then \({\partial \Delta \over \partial t} \ge 0\) holds and the sender has no incentive to deviate for any \(t\in T_n \) since \({\partial ^2 \Delta \over \partial t^2}\le 0\). Taking derivative of \(\Delta \) with respect to t yields
which presents an additional condition to be satisfied:
Second, for \(t\in T_{n+1}\), taking derivative of \(\Delta \) with respect to t yields
Note that \( \lim _{ t \downarrow t _n } \frac{\partial \Delta }{\partial t } > 0 \). Then, the sender has no incentive to deviate for any \( t \in T _{ n + 1 } \) since \(\frac{\partial ^2 \Delta }{\partial t ^2 } \ge 0 \).
Lastly, for \( t \in T \backslash ( T _{ n } \cup T _{ n + 1 } ) \), \( \Delta \) can be written as
It is also verified that
Then, \( \Delta ( t ; t _{ n - 1 } , t _{ n } , t _{ n + 1 } ) \le 0 \) for \( t \in T _{ n ' } \) and \( n ' < n \) if and only if
This holds if and only if
since \( \lim _{ t \uparrow t _{ n ' } } \Delta ( t ; t _{ n - 1 } , t _{ n } , t _{ n + 1 } ) = \lim _{ t \downarrow t _{ n ' } } \Delta ( t ; t _{ n - 1 } , t _{ n } , t _{ n + 1 } ) \) for \( n ' < n - 1 \). This is guaranteed by \(\underline{G}(t_{n-1}, t_n , t_{n+1})\le 0\) and (3).
Similarly, \( \Delta ( t ; t _{ n - 1 } , t _{ n } , t _{ n + 1 } ) \ge 0 \) for \( t \in T_{ n ' } \) and \( n ' > n + 1 \) if and only if
This holds if and only if
since \( \lim _{ t \downarrow t _{ n ' } } \Delta ( t ; t _{ n - 1 } , t _{ n } , t _{ n + 1 } ) = \lim _{ t \uparrow t _{ n ' } } \Delta ( t ; t _{ n - 1 } , t _{ n } , t _{ n + 1 } ) \) for \( n ' > n + 1 \). This is guaranteed by \(\overline{G}(t_{n-1}, t_n , t_{n+1})\ge 0\). \(\square \)
Proof of Proposition 4
We have shown that the incentive conditions in our setting are less stringent than those in CS. Moreover, if \( \beta _{ n + 1 } + \beta _{ n } - 2 t _{ n } - 2 b = 0 \) holds, then
This means that the equilibrium partition in our setting can be more flexible than that in CS. Thus, the first statement of the proposition is verified.
As for the second statement, fix any MPE-0 \( {\mathbf {t} } = ( t _0 , \dots , t _N ) \) for \( N \ge 2 \). Note that
It then follows that there exists \( \bar{\varepsilon } > 0 \) such that for any \( \varepsilon \in ( 0 , \bar{\varepsilon } ) \) a partition \( \bar{\mathbf {t}} ( \varepsilon ) \) satisfying
constitutes an MPE-q and \( V ^R ( \bar{\mathbf {t}} ( \varepsilon ) , 0 ) > V ^R ( \mathbf {{t}} , 0 ) \). Also, the latter implies that \( V ^S ( \bar{\mathbf {t}} ( \varepsilon ) , 0 ) > V ^S ( \mathbf {t} , 0 ) \) by (2). \(\square \)
Proof of Proposition 5
Throughout this proof we assume that the state is distributed over [0, 1] . In this case \( \underline{G} \) and \( \overline{G} \) are rewitten as follows:
Then, the equilibrium conditions are simplified and obtained as follows.
Lemma 1
For any \(\tau _n \in (0, 1)\) and \(q\in [0, 1)\),
- (i)
There exists \(\underline{\tau }(\tau _n, q)\) and \(\overline{\tau }(\tau _n, q)\) such that \(\overline{G}(t_{n-1},t_n,t_{n+1};q) \ge 0\ge \underline{G}(t_{n-1},t_n,t_{n+1};q)\) if and only if \(\overline{\tau }(\tau _n, q) \ge \tau _{n+1}\ge \underline{\tau }(\tau _n, q)\);
- (ii)
\(\underline{G}(t_{n-1},t_n,t_n+\underline{\tau }(\tau _n,q);q) = 0 \) and \(\overline{G}(t_{n-1},t_n,t_n+\overline{\tau }(\tau _n,q);q) = 0\);
- (iii)
\(\overline{\tau }(\tau _n, q)\ge \tau _n + 4b\ge \underline{\tau }(\tau _n,q )> 2b\) where
$$\begin{aligned} q = 0 \&\Leftrightarrow \ \overline{\tau }( \tau _n , q ) = \tau _n + 4 b = \underline{\tau }( \tau _n , q ) , \\ q> 0 \&\Leftrightarrow \ \overline{\tau }(\tau _n,q)> \tau _n + 4b > \underline{\tau }(\tau _n,q). \end{aligned}$$
Moreover, \(\overline{G}(t_{n-1},t_n,t_{n+1};q) \ge 0\ge \underline{G}(t_{n-1},t_n,t_{n+1};q)\) for \(n = 1, \dots , N - 1\) is both necessary and sufficient for an MPE with N intervals.
Proof
We first make the following observations by direct calculation. \(\square \)
Observation 1
For any \( \tau _n \in [0, 1] \) and \(\tau _{n+1}\ge 0\),
Moreover,
where the last condition holds with equality if and only if \(q = 0\).
Observation 2
For any \( \tau _n \in [0, 1] \) and \(\tau _{n+1}\ge 0\),Footnote 17
Moreover,
where the latter holds with equality if and only if \(q = 0\).
It follows from Observation 1 that for any \( \tau _n \in [0, 1]\) and \(\tau _{n+1} \ge 0\), \(\underline{G}(t_{n-1},t_n,t_{n+1})\le 0\) if and only if \(\tau _{n+1}\ge \underline{\tau }(\tau _n, q)\) where \(\underline{\tau }(\tau _n, q)\) is defined as the unique solution \(\tilde{\tau }\in [0,\infty ) \) of \(\underline{G}(t_{n-1},t_n,t_n+\tilde{\tau }) = 0 \). Moreover, \(\tau _n + 4b\ge \underline{\tau }(\tau _n, q) > 2b\) where \(\underline{\tau }(\tau _n,q) = \tau _n + 4b\) if and only if \(q = 0\). Similarly, it follows from Observation 2 that for any \(\tau _n\in [0,1] \) and \(\tau _{n+1}\ge 2b\), \(\overline{G}(t_{n-1},t_n,t_{n+1}) \ge 0\) if and only if \(\tau _{n+1} \le \overline{\tau }(\tau _n , q) \) where \( \overline{\tau }(\tau _n,q)\) is defined as the unique solution \(\tilde{\tau }\in [2b,\infty )\) of \(\overline{G}(t_{n-1},t_n,t_n+\tilde{\tau }) = 0\). Moreover, \(\overline{\tau }(\tau _n,q) \ge \tau _n + 4b\) which holds with equality if and only if \(q=0\).
Finally, we show that \(\overline{G}(t_{n-1},t_n,t_{n+1};q) \ge 0\ge \underline{G}(t_{n-1},t_n,t_{n+1};q)\) for \(n = 1, \dots , N - 1\) is both necessary and sufficient for an MPE with N intervals when the state is uniformly distributed. To see this, since \(\beta _n={t_n+t_{n-1}\over 2}\) and \(\tau _n=t_n-t_{n-1}\) under the uniformed distribution, condition (ii) of Proposition 2 can be written as
This condition holds because \(\tau _{n+1}>2b\).\(\square \)
Given this result, we prove the proposition by a series of lemmas.
Lemma 2
Given \( { \mathbf t } = ( t _0 , \dots t _N ) \), \( { \mathbf t ' } = ( t ' _0 , \dots t ' _N ) \), if there exists \(\hat{n}\) such that
\( t_{\hat{n}} > t'_{\hat{n}}\),
\( t _n = t ' _n \) for \( n \not = \hat{n} \), and
\( \tau _{ \hat{n} } < \tau _{ \hat{n} + 1 } \).
then \(\mathbf{t}\) is more efficient than \(\mathbf{t'}\) at \( q = 0 \), i.e., \(V^i(\mathbf{t},0)>V^i(\mathbf{t'},0)\).
Proof
We denote
Then, since it is verified
we obtain the lemma. \(\square \)
Lemma 3
Given b and \( q > 0 \), if \( {\mathbf {t}} \) is the most efficient MPE and \( \tau _{ n } < \tau _{ n + 1 } \) for \( n = 1 , \dots , N - 1 \), then \( \tau _{ n + 1 } = \underline{\tau }( \tau _n , q ) \) for \( n = 1 , \dots , N - 1 \).
Proof
Suppose the contrary. Define \( \hat{n} \) such that
Then, \( \hat{n} = 1 \) or
Thus, we can find another MPE \( {\mathbf {t}} ' \) where \( t ' _{ \hat{n} } \) is slightly higher than \( t _{ \hat{n} } \), \( t ' _{ n } = t _{ n } \) for \( n \not = \hat{n} \), and the presupposition of Lemma 2 holds at \( n = \hat{n} \). Then the lemma is obtained from Lemma 2. \(\square \)
Lemma 4
There exists \( \hat{q} > 0 \) such that \( \underline{\tau }( \tau , q ) > \tau \) for any \( \tau \in ( 0 , 1 ) \) and \( q \in [ 0 , \hat{q} ] \).
Proof
By Lemma 1, \( \underline{\tau }( \tau , 0 ) = \tau + 4 b > \tau \) for any \( \tau \in ( 0 , 1 ) \). Then, there must exist \( \hat{q} > 0 \) such that \( \hat{\tau }( \tau , q ) > \tau \) for any \( \tau \in ( 0 , 1 ) \) and \( q \in [ 0 , \hat{q} ] \). \(\square \)
Lemma 5
There exists \(\check{q} > 0\) such that \( \displaystyle \frac{ \partial \underline{\tau }}{ \partial q } < 0 \) for any \( \tau \in ( 0 , 1 ) \) and \( q \in ( 0 , \check{q} ] \).
Proof
Since \( \frac{\partial \underline{G}}{ \partial t _{n+1}} < 0 \) for \(\tau _{n+1} > 2b\) (see the proof of Lemma 1), \(\frac{ \partial \underline{G}}{ \partial t_{n+1}}\vert _{ \tau _{ n + 1 } = \underline{\tau }( \tau _n ) } < 0\).
On the other hand,
Then, Observation 1 (in the proof of Lemma 1) implies
Therefore, there exists \( \check{q} \) such that \( \frac{\partial \underline{G}}{ \partial q } \vert _{ \tau _{ n + 1 } = \underline{\tau }( \tau _n ) } < 0 \) for any \( q \in ( 0 , \check{q} ] \).
We now return to the Proof of Proposition 5. Letting \( \overline{q} := \min \{ \hat{q} , \check{q} \} \), the proposition then follows from Lemmas 2, 3, 4, and 5. \(\square \)
Appendix B: The discrete-state model and convergence
In this appendix, we establish that the equilibrium of the discrete-state model converges to that of the continuous-state model as \(L\rightarrow \infty \). In the process, we also derive the equilibrium conditions for a sufficiently large L which allows us to isolate the confirmation effect.
In what follows, we use the following notations:
\( m _n \) is a typical element of the messages sent by \( t \in T ^L _n \),
\( T ^L _n = \{ t _{ I _{ n - 1 } + 1 } , \dots , t _{ I _{ n } } \} \) for \( n = 1 \dots , N \) where \( I _0 = - 1 \),
\( \gamma _n = \frac{ I _n + I _{ n - 1 } + 1 }{ 2 L } \),
\( \lambda _n = \frac{ I _{ n } - I _{ n - 1 } }{ L } \), and
\( \ell = \frac{ 1 }{ L } \).
Define
A sufficient condition for an MPE with \( \{ T ^L _n \} _{ n = 1 ,\dots , N } \) is that
holds for any \( n = 1 , \dots , N - 1 \). Note that \(\Delta ^L_n(t_{I_n}) \le 0 \) and \(\Delta ^L_n ( t _{ I _{ n } + 1 } ) \ge 0 \) are sufficient for (4) whenever L is sufficiently large.
Proposition 6
There exists \( \varLambda \) such that for any finite model with \( L > \varLambda \), \( \Delta ^L _n ( t _{ I _{ n } } ) \le 0 \) and \( \Delta ^L _n ( t _{ I _{ n } + 1 } ) \ge 0 \) are sufficient conditions for an MPE with \( \{ T ^L _n \} _{ n = 1 , \dots , N } \).
Proof
The receiver’s posterior is written as
Then, the receiver’s best response is written as
For \( t \in T ^L _n \), we obtain
For \( t \in T ^L _{ n + 1 } \),
For \( t \notin T ^L _{ n } \cup T ^L _{ n + 1 } \),
We show that \( \Delta ^L _n ( t _{ I _{ n } } ) \le 0 \) implies \( \lambda _{ n + 1 } > 2 b \) whenever L is sufficiently large. First, \( \Delta ^L _n ( t _{ I _{ n } } ) \) is rewritten as
When \( \Delta ^L _n ( t _{ I _{ n } } ) \) is considered as a function of \( \lambda _{ n + 1 } \)Footnote 18, we obtain
It then follows that \( \Delta ^L _n ( t _{ I _{ n } } ) \le 0 \) implies \( \lambda _{ n + 1 } > 2 b \) for any sufficiently large L.
We are in a position to verify that \( \Delta ^L _n ( t _{ I _{ n } } ) \le 0 \) and \( \Delta ^L _n ( t _{ I _{ n } + 1 } ) \ge 0 \) are sufficient. We do it by dividing the situation into the following cases.
Case 1 For \( i = 0 , \dots , I _{ n - 1 } - 1 \), it is verified that
Case 2 It is verified that
Case 3 For \( i = I _{ n - 1 } + 1 , \dots , I _{ n } - 1 \),
Case 4 For \( i = I _{ n } + 1 , \dots , I _{ n + 1 } - 1 \), it is verified that
Case 5 It is verified that
Case 6 For \( i = I _{ n + 1 } + 1 , \dots , L - 1 \), it is verified that
From Cases 1–6, if \( \Delta ^L _n ( t _{ I _{ n } } ) \le 0 \), \( \Delta ^L _n ( t _{ I _{ n } + 1 } ) \ge 0 \), and L is sufficiently large, then we have
Therefore, we conclude (4). \(\square \)
Given this, we now explicitly derive the equilibrium conditions. With some computation, we obtain
where \( A _n := \frac{ q ( 1 + \ell ) }{ q ( 1 + \ell ) + ( 1 - q ) \lambda _n } \). The third line of each equation represents the confirmation effect, which tends to relax the equilibrium conditions.
Finally, to motivate the use of the discrete-state model, we confirm that our baseline model is indeed the limit of a sequence of discrete-state models as \(L \rightarrow \infty \). More precisely, we define \(T^L_n = T_n \cap T^L\) for \( n = 1 , \dots , N \). It is then verified that
leading to the following statement (the proof omitted).
Corollary 1
Consider the baseline model where the state is uniformly distributed. Given any MPE with \( \{ T _n \} _{ n = 1 , \dots , N } \) of the baseline model, which satisfies the equilibrium conditions with strict inequalities, i.e., \(\overline{G}(t_{n-1}, t, t_{n+1})> 0 > \underline{G}(t_{n-1}, t, t_{n+1})\) for \( n = 1 , \dots , N - 1 \), then there exists \( \varLambda \) such that for any discrete-state model with \( L > \varLambda \), there exists an MPE with \( \{ T ^L _n \} _{ n = 1 , \dots , N } \) such that \( T ^L _n \subset T _n \) for any n.
Rights and permissions
About this article
Cite this article
Ishida, J., Shimizu, T. Cheap talk when the receiver has uncertain information sources. Econ Theory 68, 303–334 (2019). https://doi.org/10.1007/s00199-018-1123-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00199-018-1123-y