The social value of information with an endogenous public signal


This paper analyzes the equilibrium and welfare properties of an economy characterized by uncertainty and payoff externalities using a general model that nests several applications. Agents receive a private signal and an endogenous public signal, which is a noisy aggregate of individual actions, and causes an information externality. Agents in equilibrium underweight private information for a larger payoff parameter region in relation to when public information is exogenous. In addition, the socially optimal endogenous degree of coordination is lower than the socially optimal exogenous degree of coordination. The welfare effect of increasing the precision of the noise in the public signal has the same sign with endogenous or exogenous public information, but its magnitude differs. The social value of private information may be overturned in relation to when public information is exogenous: from positive to negative if agents in equilibrium coordinate more than is implied by the socially optimal exogenous degree of coordination, and the opposite if they coordinate less.

This is a preview of subscription content, access via your institution.


  1. 1.

    See, for example, McAfee and Brynjolfsson (2012) and Einav and Levin (2014).

  2. 2.

    See, for example, Avdjiev et al. (2012).

  3. 3.

    Clark and Polborn (2006) also studied this issue in a binary choice context.

  4. 4.

    AP refer to \(\alpha \) as the equilibrium degree of coordination.

  5. 5.

    This is not the only way in which the public signal can be made endogenous. Refer to Sect. 2 for a discussion on the alternative formulations used in the literature.

  6. 6.

    If \(\tau _{v}\rightarrow \infty \) then the REE would be fully revealing.

  7. 7.

    Refer to Sect. 2 for further details of this literature.

  8. 8.

    The paper closely follows AP’s notation, except from the weight to private information. In this paper, the weight to private information with an endogenous public signal is \(\gamma ^{m}\), while the weight to private information with an exogenous public signal is \(\gamma _{\mathrm{exo}}\). AP use \(1-\gamma \) to denote the weight to private information with an exogenous public signal. The reason for this alternative notation is that the mathematical expressions become significantly more compact since most of the results throughout the paper depend primarily upon the weight given to private information.

  9. 9.

    This discussion also applies to the logic of the comparative statics with respect to \(\tau _{\theta }\).

  10. 10.

    I have added the word exogenous to “socially optimal exogenous degree of strategic coordination” since \(\alpha ^{*}\) is no longer optimal with an endogenous public signal. Notice that \(\alpha ^{*}\) only depends on the parameters of the utility function. For further details refer to Proposition 4.

  11. 11.

    This result has also been found in Bru and Vives (2002) focus on a pure prediction (\(\alpha ^{*}=\alpha =0\)) with endogenous public information. Models that consider a payoff structure with \(\alpha =\alpha ^{*}\) but in which price has an additional allocation role, such as in Vives (2017), and where agents use price contingent strategies, lead to different results as those presented in Proposition 3. For example, agents in equilibrium can give too much weight to private information whenever actions are strategic substitutes since the market price has dual information and allocation roles.

  12. 12.

    An implication is that the equilibrium precision of the endogenous public signal, \(\tau ^{m}\), may be too high or too low in relation to the efficient level, \(\tau ^{*}\).

  13. 13.

    I thank a referee for suggesting that this result should be discussed.

  14. 14.

    Specifically, \(E[W_{K}(\kappa (\theta ),0,\theta )(K-\kappa (\theta ))]=\mid W_{KK}\mid E[(K-\kappa (\theta ))\kappa (\theta )]\Big (\frac{E[\kappa (\theta )(\kappa ^{*}(\theta )-\kappa (\theta ))]}{E[(\kappa (\theta ))^{2}]}\Big )\), where \(\phi =\Big (\frac{E[\kappa (\theta )(\kappa ^{*}(\theta )-\kappa (\theta ))]}{E[(\kappa (\theta ))^{2}]}\Big )\).

  15. 15.

    I assume that the welfare planner cannot change the precision of the ex ante fundamentals, \(\tau _{\theta }.\)

  16. 16.

    For completeness, define \(\hat{X}=-\infty \) if \(\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }}=0\).


  1. Amador, M., Weill, P.O.: Learning from prices: public communication and welfare. J. Polit. Econ. 118(5), 866–907 (2010)

    Article  Google Scholar 

  2. Angeletos, G.M., Pavan, A.: Efficient use of information and social value of information. Econometrica 75(4), 1103–1142 (2007)

    Article  Google Scholar 

  3. Angeletos, G.M., Pavan, A.: Policy with dispersed information. J. Eur. Econ. Assoc. 7(1), 11–60 (2009)

    Article  Google Scholar 

  4. Angeletos, G.M., Werning, I.: Crises and prices: information aggregation, multiplicity, and volatility. Am. Econ. Rev. 96(5), 1720–1736 (2006)

    Article  Google Scholar 

  5. Avdjiev, S., McGuire, P., Tarashev, N.: The social value of policy signals. BIS Working Paper No. 386 (2012)

  6. Banerjee, A.V.: A simple model of herd behavior. Q. J. Econ. 107(3), 797–817 (1992)

    Article  Google Scholar 

  7. Bergemann, D., Morris, S.: Robust predictions in games with incomplete information. Econometrica 81(4), 1251–1308 (2013)

    Article  Google Scholar 

  8. Bikhchandani, S., Hirshleifer, D., Welch, I.: A theory of fads, fashion, custom, and cultural change in informational cascades. J. Polit. Econ. 100(5), 992–1026 (1992)

    Article  Google Scholar 

  9. Bond, P., Goldstein, I.: Government intervention and information aggregation by prices. J. Finance 70(6), 2777–2812 (2015)

    Article  Google Scholar 

  10. Bru, L., Vives, X.: Informational externalities, herding, and incentives. J. Inst. Theor. Econ. 158(1), 91–105 (2002)

    Article  Google Scholar 

  11. Choi, H., Varian, H.: Predicting the present with Google Trends. Econ. Rec. 88(s1), 2–9 (2012)

    Article  Google Scholar 

  12. Clark, C., Polborn, M.K.: Information and crowding externalities. Econ. Theory 27(3), 565–581 (2006)

    Article  Google Scholar 

  13. Colombo, L., Femminis, G., Pavan, A.: Information acquisition and welfare. Rev. Econ. Stud. 81(4), 1438–1483 (2014)

    Article  Google Scholar 

  14. Diamond, D., Verrecchia, R.E.: Information aggregation in a noisy rational expectations economy. J. Financ. Econ. 9(3), 221–235 (1981)

    Article  Google Scholar 

  15. Einav, L., Levin, J.: Economics in the age of big data. Science 346(6210), 1243089 (2014)

    Article  Google Scholar 

  16. Grossman, S., Stiglitz, J.: On the impossibility of informationally efficient markets. Am. Econ. Rev. 70(3), 393–408 (1980)

    Google Scholar 

  17. Hauk, E., Hurkens, S.: Secret information acquisition in Cournot markets. Econ. Theory 18(3), 661–681 (2001)

    Article  Google Scholar 

  18. Hellwig, M.: On the aggregation of information in competitive markets. J. Econ. Theory 22(3), 477–498 (1980)

    Article  Google Scholar 

  19. Hellwig, C.: Heterogeneous information and the benefits of transparency. UCLA Economics Online Papers 283 (2005)

  20. Klemperer, P., Meyer, M.: Supply function equilibria in oligopoly under uncertainty. Econometrica 57(6), 1243–77 (1989)

    Article  Google Scholar 

  21. Kovalenkov, A., Vives, X.: Competitive rational expectations equilibria without apology. J. Econ. Theory 149(1), 211–235 (2014)

    Article  Google Scholar 

  22. McAfee, A., Brynjolfsson, E.: Big data: the management revolution. Harv. Bus. Rev. 90(10), 60–68 (2012)

    Google Scholar 

  23. Morris, S., Shin, H.S.: Social value of public information. Am. Econ. Rev. 92(5), 1521–1534 (2002)

    Article  Google Scholar 

  24. Morris, S., Shin, H.S.: Central bank transparency and the signal value of prices. Brook. Pap. Econ. Act. 36(2), 1–66 (2005)

    Article  Google Scholar 

  25. Myatt, D., Wallace, C.: Cournot competition and the social value of information. J. Econ. Theory 158(Part B), 466–506 (2015)

    Article  Google Scholar 

  26. Myatt, D.P., Wallace, C.: Information use and acquisition in price-setting oligopolies. Econ. J. (2017). doi:10.1111/ecoj.12410

    Article  Google Scholar 

  27. Radner, R.: Team decision problems. Ann. Math. Stat. 33(3), 857–881 (1962)

    Article  Google Scholar 

  28. Shorish, J.: Functional rational expectations equilibria in market games. Econ. Theory 43(3), 351–376 (2010)

    Article  Google Scholar 

  29. Ui, T., Yoshizawa, Y.: Characterizing social value of information. J. Econ. Theory 158(Part B), 507–535 (2015)

    Article  Google Scholar 

  30. Vives, X.: Aggregation of information in large Cournot markets. Econometrica 56(4), 851–876 (1988)

    Article  Google Scholar 

  31. Vives, X.: Trade association disclosure rules, incentives to share information, and welfare. RAND J. Econ. 21(3), 409–430 (1990)

    Article  Google Scholar 

  32. Vives, X.: On the possibility of informationally efficient markets. J. Eur. Econ. Assoc. 12(5), 1200–1239 (2014)

    Article  Google Scholar 

  33. Vives, X.: Endogenous public information and welfare in market games. Rev. Econ. Stud. 84(2), 935–963 (2017)

    Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Anna Bayona.

Additional information

I thank the editor, the associate editor and two anonymous referees for providing very useful comments which have greatly improved the paper. I am grateful to my thesis advisor, Xavier Vives, for the guidance and feedback. I would also like to thank Jose Apesteguia, Ariadna Dumitrescu, Emre Ekinci, Sebastian Harris, Juan Imbett, Angel L. Lopez, Carolina Manzano, Margaret Meyer, Manuel Mueller-Frank, Rosemarie Nagel, Morten Olsen, and Alessandro Pavan for constructive comments.



Proof of Proposition 1

Focusing on linear arbitrary strategies of the form \(k(x_{i},z)=b'+ax_{i}+c'E\left[ \theta \mid z\right] \) and using the definition of the public signal, I obtain: \(w=\int k_{i}+v=b'+a\theta +c'w+v\), the informational content of which can be seen to be the same as that of \(z=a\theta +v\). From the properties of the normal distribution: \(E\left[ \theta |z\right] =\frac{\tau _{v}az+\tau _{\theta }\bar{\theta }}{\tau _{\theta }+a^{2}\tau _{v}}\) and \((var\left[ \theta |z\right] )^{-1}=\tau (a)=\tau _{\theta }+a{}^{2}\tau _{v}\). Hence \(E\left[ \theta |x_{i},z\right] =\frac{\tau _{\xi }}{\tau _{\xi }+\tau (a)}x_{i}+\frac{\tau (a)}{\tau _{\xi }+\tau (a)}E\left[ \theta \mid z\right] \).

Using the maximization condition in the REE definition, the best response strategy \(k'\) satisfies \(E\left[ U_{k}(k',K,\sigma _{k},\theta )\mid x_{i},z\right] =0\) and hence \(k'=E[(1-\alpha )\kappa (\theta )+\alpha K\mid x_{i},w]\). The second-order condition is satisfied since \(U_{kk}<0\). Equating coefficients, I obtain: \(b=\kappa _{0}\), \(a=\frac{\kappa _{1}(1-\alpha )\tau _{\xi }}{(1-\alpha )\tau _{\xi }+\tau (a)}\) and \(c=\frac{\kappa _{1}\tau (a)}{(1-\alpha )\tau _{\xi }+\tau (a)}\). Hence, \(k(x_{i},z)=\kappa _{0}+\kappa _{1}(\gamma x_{i}+(1-\gamma )E\left[ \theta |z\right] )\), where \(a=\kappa _{1}\gamma \) and \(\gamma =\frac{(1-\alpha )\tau _{\xi }}{(1-\alpha )\tau _{\xi }+\tau _{\theta }+\kappa _{1}^{2}\gamma ^{2}\tau _{v}}\). The equilibrium weight to private information, \(\gamma ^{m}\), is then a solution of \(\Gamma (\gamma ^{m})=0\), where \(\Gamma (\gamma )=\gamma ^{3}\kappa _{1}^{2}\tau _{v}+\gamma ((1-\alpha )\tau _{\xi }+\tau _{\theta })-(1-\alpha )\tau _{\xi }.\) By the Descartes’ Rule of signs, I note that there is only one sign change in \(\Gamma (\gamma )\), and therefore, there is only a positive real root. Checking \(\Gamma (-\gamma )\), I notice that there are no sign changes, and therefore, there are no negative real roots. Additionally, \(\Gamma (0)=-(1-\alpha )\tau _{\xi }<0\) and \(\Gamma (1)=\kappa _{1}^{2}\tau _{v}+\tau _{\theta }>0\), and the positive real root is between 0 and 1. \(\square \)

Proof of Corollary 1

Differentiating \(\Gamma (\gamma ^{m})=0\) with respect to the various information and payoff parameters, and noting that \(0<\gamma ^{m}<1\), I obtain:

$$\begin{aligned} \frac{\partial \gamma ^{m}}{\partial \tau _{\xi }}= & {} \frac{(1-\alpha )(1-\gamma ^{m})}{(1-\alpha )\tau _{\xi }+\tau _{\theta }+3(\gamma ^{m})^{2}\kappa _{1}^{2}\tau _{v}}>0,\\ \frac{\partial \gamma ^{m}}{\partial \tau _{\theta }}= & {} \frac{-\gamma ^{m}}{(1-\alpha )\tau _{\xi }+\tau _{\theta }+3(\gamma ^{m})^{2}\kappa _{1}^{2}\tau _{v}}<0,\\ \frac{\partial \gamma ^{m}}{\partial \tau _{v}}= & {} \frac{-(\gamma ^{m})^{3}\kappa _{1}^{2}}{(1-\alpha )\tau _{\xi }+\tau _{\theta }+3(\gamma ^{m})^{2}\kappa _{1}^{2}\tau _{v}}<0,\quad \hbox {and}\\ \frac{\partial \gamma ^{m}}{\partial \alpha }= & {} \frac{-(1-\gamma ^{m})\tau _{\xi }}{(1-\alpha )\tau _{\xi }+\tau _{\theta }+3(\gamma ^{m})^{2}\kappa _{1}^{2}\tau _{v}}<0. \end{aligned}$$

\(\square \)

Proof of Corollary 2

The proof follows since: \(\frac{d\tau ^{m}}{d\tau _{\xi }}=\frac{(2\kappa _{1}^{2}\tau _{v}\gamma ^{m})(1-\alpha )(1-\gamma ^{m})}{(1-\alpha )\tau _{\xi }+\tau _{\theta }+3(\gamma ^{m})^{2}\kappa _{1}^{2}\tau _{v}}>0\), \(0<\frac{d\tau ^{m}}{d\tau _{v}}=\kappa _{1}^{2}\gamma ^{2}(\frac{(1-\alpha )\tau _{\xi }+\tau _{\theta }+(\gamma ^{m})^{2}\kappa _{1}^{2}\tau _{v}}{(1-\alpha )\tau _{\xi }+\tau _{\theta }+3(\gamma ^{m})^{2}\kappa _{1}^{2}\tau _{v}})<\kappa _{1}^{2}\), and \(\frac{d\tau ^{m}}{d\tau _{\theta }}=\frac{(1-\alpha )\tau _{\xi }+\tau ^{m}}{(1-\alpha )\tau _{\xi }+\tau _{\theta }+3(\gamma ^{m})^{2}\kappa _{1}^{2}\tau _{v}}\), which implies that \(0<\frac{d\tau ^{m}}{d\tau _{\theta }}<1\). \(\square \)

Proof of Proposition 2

The welfare planner internalizes all externalities and maximizes ex ante utility, or equivalently minimizes welfare loss. Focusing on a linear arbitrary efficient strategy of the form \(k^{*}(x_{i},z)=b^{*}+a^{*}x_{i}+c^{*}E\left[ \theta \mid z\right] \). By a similar argument as that used in Proposition 1, \(b^{*}=\kappa _{0}^{*}\) and \(k^{*}(x_{i},z)=\kappa _{0}^{*}+\kappa _{1}^{*}(\gamma x_{i}+(1-\gamma )E\left[ \theta |z\right] )\), where \(\gamma \) is an arbitrary weight to the private signal, and \(\kappa _{0}^{*},\kappa _{1}^{*}\) were found by AP and are reported in the text. The first-order condition satisfies \(\frac{dL^{*}}{d\gamma }|_{\gamma =\gamma ^{*}}=0\). Hence, a candidate efficient strategy with weight to private information \(\gamma \) and precision given by \(\tau (\kappa _{1}^{*}\gamma )=\tau _{\theta }+(\kappa _{1}^{*}\gamma )^{2}\tau _{v}\) solves

$$\begin{aligned} \frac{\gamma }{\tau _{\xi }}-\frac{(1-\alpha ^{*})(1-\gamma )}{\tau (\kappa _{1}^{*}\gamma )}-\frac{(\kappa _{1}^{*})^{2}\tau _{v}\gamma (1-\alpha ^{*})(1-\gamma )^{2}}{(\tau (\kappa _{1}^{*}\gamma )){}^{2}}=0, \end{aligned}$$

where the last term is due to the endogeneity of public information. The previous expression can be rewritten as

$$\begin{aligned} \gamma =\frac{(1-\alpha ^{*})\tau _{\xi }}{(1-\alpha ^{*})\tau _{\xi }+\tau _{\theta }+(\kappa _{1}^{*})^2\gamma ^{2}\tau _{v}-(1-\alpha ^{*})(1-\gamma )^{2}\frac{\tau _{v}\tau _{\xi }(\kappa _{1}^{*})^2}{(\tau _{\theta }+(\kappa _{1}^{*})^2 \gamma ^{2} \tau _{v})}}. \end{aligned}$$

Then, \(\gamma ^{*}\) is a solution of a quintic equation \(\psi (\gamma ^{*})=0\), where

$$\begin{aligned} {\begin{matrix} \psi (\gamma )&{}=(\gamma )^{5}\tau _{v}^{2}(\kappa _{1}^{*})^{4}+2(\gamma )^{3}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\theta }+(\gamma )^{2}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\xi }(1-\alpha ^{*})\\ &{}\quad +\gamma (\tau _{\theta }^{2}+\tau _{\xi }(1-\alpha ^{*})(\tau _{\theta }-\tau _{v}(\kappa _{1}^{*})^{2}))-\tau _{\xi }\tau _{\theta }(1-\alpha ^{*}). \end{matrix}} \end{aligned}$$

The SOC guarantees that the denominator of (A2) is positive and hence \(\gamma ^{*}>0\). Applying Descartes’ Rule of signs to find out the number of real roots, I find that there is only one change in sign in \(\psi (\gamma )\), and therefore, this polynomial has only one real positive root. Furthermore, \(\psi (0)=-\tau _{\xi }\tau _{\theta }(1-\alpha ^{*})<0\), while \(\psi (1)=\tau _{v}^{2}(\kappa _{1}^{*})^{4}+2(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\theta }+\tau _{\theta }^{2}>0\) and the unique positive real root is between 0 and 1. \(\square \)

Proof of Corollary 3

Differentiating \(\psi (\gamma ^{*})=0\) with respect to the various information and payoff parameters and noting that \(0<\gamma ^{*}<1\), I find:

$$\begin{aligned} \frac{\partial \gamma ^{*}}{\partial \tau _{\xi }}= & {} \frac{\gamma ^{*}(1-\gamma ^{*})(1-\alpha ^{*})\tau _{v}(\kappa _{1}^{*})^{2}}{5(\gamma ^{*})^{4}\tau _{v}^{2}(\kappa _{1}^{*})^{4}+6(\gamma ^{*})^{2}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\theta }+2\gamma ^{*}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\xi }(1-\alpha ^{*})}>0, \\ \frac{\partial \gamma ^{*}}{\partial \tau _{\theta }}= & {} \frac{-2\gamma ^{*}\tau ^{*}+\tau _{\xi }(1-\alpha ^{*})(1-\gamma ^{*})}{5(\gamma ^{*})^{4}\tau _{v}^{2}(\kappa _{1}^{*})^{4}+6(\gamma ^{*})^{2}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\theta }+2\gamma ^{*}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\xi }(1-\alpha ^{*})}<0,\quad \hbox {and}\\ \frac{\partial \gamma ^{*}}{\partial \alpha ^{*}}= & {} \frac{-(\gamma ^{*}\tau _{\xi }\tau _{v}(\kappa _{1}^{*})^{2}+\tau _{\xi }\tau _{\theta })(1-\gamma ^{*})}{5(\gamma ^{*})^{4}\tau _{v}^{2}(\kappa _{1}^{*})^{4}+6(\gamma ^{*})^{2}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\theta }+2\gamma ^{*}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\xi }(1-\alpha ^{*})}<0. \end{aligned}$$

Notice that the sign of \(\frac{\partial \gamma ^{*}}{\partial \tau _{v}}=\kappa _{1}^{2}\gamma ^{*}(\frac{-2(\gamma ^{*})^{2}\tau ^{*}+\tau _{\xi }(1-\alpha ^{*})(1-\gamma ^{*})}{5(\gamma ^{*})^{4}\tau _{v}^{2}(\kappa _{1}^{*})^{4}+6(\gamma ^{*})^{2}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\theta }+2\gamma ^{*}(\kappa _{1}^{*})^{2}\tau _{v}\tau _{\xi }(1-\alpha ^{*})})\) is ambiguous. \(\square \)

Proof of Proposition 3

The efficient strategy \(\gamma ^{*}\) satisfies \(\frac{dL^{*}}{d\gamma }|_{\gamma =\gamma ^{*}}=0\), where \(L^{*}\) is a strictly convex function of \(\gamma \). Hence \({{\mathrm{sign}}}\{\gamma ^{m}-\gamma ^{*}\}\) can be found by inspecting the sign of \(\frac{dL^{*}}{d\gamma }\mid _{\gamma =\gamma ^{m}}\). Substituting for the equilibrium strategy in \(\frac{dL^{*}}{d\gamma }\mid _{\gamma =\gamma ^{m}}\), I obtain that

$$\begin{aligned} {{\mathrm{sign}}}\left\{ \frac{dL^{*}}{d\gamma }\mid _{\gamma =\gamma ^{m}} \right\} ={{\mathrm{sign}}}\left\{ \gamma ^{m}-\gamma ^{*}\right\} = {{\mathrm{sign}}}\left\{ \alpha ^{*}-\alpha -\frac{(1-\alpha ^{*})(1-\alpha ) (\kappa _{1}^{2}\tau _{v}\tau _{\xi })}{((1-\alpha )\tau _{\xi }+\tau ^{m})^{2}}\right\} , \end{aligned}$$

which implies that \({{\mathrm{sign}}}\{\gamma ^{m}-\gamma ^{*}\}={{\mathrm{sign}}}\{\hat{\alpha }-\alpha \},\) where \(\hat{\alpha }=\alpha ^{*}-\frac{(1-\alpha ^{*})(1-\alpha )(\kappa _{1}^{2}\tau _{v}\tau _{\xi })}{((1-\alpha )\tau _{\xi }+\tau ^{m})^{2}}\). \(\square \)

Proof of Proposition 4

Follows immediately from Propositions 2 and 3. \(\square \)

Proof of Proposition 5

Substituting \(\gamma ^{m}\) into (9), I obtain:

$$\begin{aligned} WL=\frac{\mid W_{\sigma \sigma }\mid \kappa _{1}^{2}}{2}\Big (\frac{(1-\alpha )^{2}\tau _{\xi }+(1-\alpha ^{*})\tau ^{m}}{((1-\alpha )\tau _{\xi }+\tau ^{m})^{2}}+\frac{2\phi (1-\alpha ^{*})\tau ^{m}}{(1-\alpha )\tau _{\xi }+\tau ^{m}}\Big ). \end{aligned}$$

Comparative statics of ex ante utility with respect to \(\tau _{v}\) are equivalent to the opposite comparative statics of WL. Since \(\frac{dWL}{d\tau _{v}}=\Big (\frac{\partial WL}{\partial \tau ^{m}}\Big )_{\gamma \,\hbox {cons}.}\frac{d\tau ^{m}}{d\tau _{v}}\) then

$$\begin{aligned} \Big (\frac{\partial WL}{\partial \tau ^{m}}\Big )_{\gamma ^{m}\,\hbox {cons}.}=\frac{\mid W_{\sigma \sigma }\mid (\kappa _{1})^{2}}{2}\Big (\frac{\iota \tau _{\xi }+\rho \tau ^{m}}{(\tau _{\xi }(1-\alpha )+\tau ^{m})^{3}}\Big ), \end{aligned}$$

where \(\iota \) and \(\rho \) are defined in Proposition 5. Recall from Corollary 2 that \(0<\frac{d\tau ^{m}}{d\tau _{v}}\). Hence, I find that \(\iota \gtrless 0\Leftrightarrow I=\frac{2\alpha -\alpha ^{*}-1}{2(1-\alpha ^{*})}\gtrless \phi \) and \(\rho \gtrless 0\Leftrightarrow R=\frac{-1}{2}\gtrless \phi \). \(\square \)

Proof of Proposition 6

Noting that \(\frac{dWL}{d\tau _{\xi }}=\Big (\frac{\partial WL}{\partial \tau _{\xi }}\Big )_{\tau ^{m}\,\hbox {cons}.}+\Big (\frac{\partial WL}{\partial \tau ^{m}}\Big )_{\gamma \,\hbox {cons}.}\frac{d\tau ^{m}}{d\tau _{\xi }}\), I obtain:

$$\begin{aligned} \Big (\frac{\partial WL}{\partial \tau _{\xi }}\Big )_{\tau ^{m}\,\hbox {cons}.}=\frac{\mid W_{\sigma \sigma }\mid (\kappa _{1})^{2}}{2}\Big (\frac{\varphi \tau _{\xi }+\chi \tau ^{m}}{(\tau _{\xi }(1-\alpha )+\tau ^{m})^{3}}\Big ), \end{aligned}$$

where \(\varphi \) and \(\chi \) are defined in Proposition 6. Rearranging the previous equation, I obtain \(\frac{d(E[u])}{d\tau _{\xi }}\gtrless 0\Leftrightarrow (\varphi +\iota \frac{d\tau ^{m}}{d\tau _{\xi }})\tau _{\xi }+(\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }})\tau ^{m}\lessgtr 0\). Furthermore, \((\varphi +\iota \frac{d\tau ^{m}}{d\tau _{\xi }})\gtrless 0\Leftrightarrow F=\frac{-(1-\alpha )^{2}+(2\alpha -\alpha ^{*}-1)\frac{d\tau ^{m}}{d\tau _{\xi }}}{2(1-\alpha ^{*})((1-\alpha )+\frac{d\tau ^{m}}{d\tau _{\xi }})}\gtrless \phi ,\) and \((\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }})\gtrless 0\Leftrightarrow C=\frac{(1-\alpha )(2\alpha ^{*}-\alpha -1)-(1-\alpha ^{*})\frac{d\tau ^{m}}{d\tau _{\xi }}}{2(1-\alpha ^{*})((1-\alpha )+\frac{d\tau ^{m}}{d\tau _{\xi }})}\gtrless \phi \). Notice that the denominators of F and C are always strictly positive. \(\square \)

Proof of Proposition 7

  1. (i)

    Cutoffs between types of games. Using the definitions of the cutoffs, I obtain the following. If \(\alpha >\alpha ^{*}\) then \(I>F>R>C\); if \(\alpha ^{*}>\alpha \) then \(C>R>F>I\); and if \(\alpha ^{*}=\alpha \) then \(C=R=F=I\). Manipulating the expressions previously defined, I obtain: \(F>F_{\mathrm{exo}}\), \(C>C_{\mathrm{exo}}\) \(\Leftrightarrow (\alpha -\alpha ^{*})\frac{d\tau ^{m}}{d\tau _{\xi }}>0\). Hence, \({{\mathrm{sign}}}\{\alpha -\alpha ^{*}\}={{\mathrm{sign}}}\{F-F_{\mathrm{exo}}\}={{\mathrm{sign}}}\{C-C_{\mathrm{exo}}\},\) and the result follows.

  2. (ii)

    Cutoffs within types of games. Games of type \(+{ III}\) have: \(\iota>0,\varphi +\iota \frac{d\tau ^{m}}{d\tau _{\xi }}>0\) and \(\rho<0,\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }}<0,\chi <0\). Hence \(\hat{X}=\frac{-\Big (\varphi +\iota \frac{d\tau ^{m}}{d\tau _{\xi }}\Big )}{\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }}}>\frac{-\varphi }{\chi }=\hat{X}_{\mathrm{exo}}\) since \((\varphi \rho -\iota \chi )\frac{d\tau ^{m}}{d\tau _{\xi }}>0\) because \(0<\varphi \rho -\iota \chi \). Similar arguments follow for games of types \(+{ IV},-{ III},-{ IV}\) since: \(+{ IV}\) that have \(\iota>0,\varphi +\iota \frac{d\tau ^{m}}{d\tau _{\xi }}>0\) and \(\rho >0,\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }}<0,\chi <0\); games of type\(-{ III}\) that have \(\iota<0,\varphi +\iota \frac{d\tau ^{m}}{d\tau _{\xi }}<0\) and \(\rho>0,\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }}>0,\chi >0\); and games of type \(-{ IV}\) that have \(\iota<0,\varphi +\iota \frac{d\tau ^{m}}{d\tau _{\xi }}<0\) and \(\rho <0,\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }}>0,\chi >0\).

\(\square \)

Proof of Corollary 4

(i) Follows from Proposition 3 by noting that \(\gamma ^{m}<\gamma ^{*}\) will always hold since \(\alpha >\alpha ^{*}\). (ii) The parameters which determine the social value of information are as follows: \(\rho =-1<0\), \(\iota =-(1-\alpha )(1+2\alpha )\), \(\varphi =-(1-\alpha )(1+\alpha )\), \(\chi =-(1-\alpha )^{2}<0\) since \(\phi =\frac{\alpha }{1-2\alpha }\) and \(\alpha ^{*}=2\alpha<\alpha <0\). Note that \(R=\frac{-1}{2}<\phi <0\). From Table 1, note that the game can be of types \(+I,+{ II},+{ III}\). If \(-\frac{1}{2}\le \alpha <0\) then both \(\hat{Y},\hat{X}<0\) and the game is of type \(+I\); if \(\alpha <-1\) then both \(\hat{Y},\hat{X}>0\) and the game is of type \(+{ III}\). Due to the endogenous public signal, the range defined by \(-1\le \alpha <\frac{-1}{2}\) gets divided into two: (1) if \(\hat{X}_{\mathrm{exo}}<\hat{X}\le 0<\hat{Y}\) then the game is of type \(+{ II}\); (2) if \(\hat{X}_{\mathrm{exo}}\le 0<\hat{X}<\hat{Y}\) then the game is of type \(+{ III}\). The cutoff between types \(+{ II}\) and \(+{ III}\) changes with endogenous public information and is given by \({{\mathrm{sign}}}\{\varphi +\iota \frac{d\tau ^{m}}{d\tau _{\xi }}\}={{\mathrm{sign}}}\{F-\phi \}\), where \(F=\frac{-(1-\alpha )^{2}-\frac{d\tau ^{m}}{d\tau _{\xi }}}{2(1-2\alpha )((1-\alpha )+\frac{d\tau ^{m}}{d\tau _{\xi }})}\). Then \(\phi>F\Leftrightarrow \Delta (\alpha )=(1-\alpha ^{2})+\frac{d\tau ^{m}}{d\tau _{\xi }}(1+2\alpha )>0\). Note that \(\Delta (-1)<0\), \(\Delta (-\frac{1}{2})>0\) and \(\Delta (\alpha )\) is a strictly increasing function of \(\alpha \) since some regularity conditions regarding \(\frac{d\tau ^{m}}{d\tau _{\xi }}\) are satisfied. Hence, there a exists a unique \(\vartheta \) such that \(\Delta (\vartheta )=0\), where \(-1<\vartheta <\frac{-1}{2}\). Then, if \(\vartheta \le \alpha <\frac{-1}{2}\) the game is of type \(+{ II}\), and if \(\alpha <\vartheta \) then the game is of type \(+{ III}\). From Proposition 7, the cutoff within a type \(+{ III}\) game changes due to endogenous public information since in this region: \(\hat{X}_{\mathrm{exo}}<\hat{X}\). \(\square \)

Proof of Corollary 5

  1. (i)

    Follows from Proposition 3 and by noting that \(\alpha ^{*}>\alpha \). When the information externality is large, the following occurs: there exists an \(\alpha ^{o}=\alpha ^{*}-\frac{(1-\alpha ^{*})(1-\alpha )(\kappa _{1}^{2}\tau _{v}\tau _{\xi })}{((1-\alpha )\tau _{\xi }+\tau ^{m})^{2}}>0\) because \(\gamma (\alpha )\) and \(\gamma ^{*}(\alpha ^{*})\), respectively, are both strictly decreasing functions of \(\alpha \), which range from 1 to 0 as \(\alpha ,\alpha ^{*}\rightarrow -\infty \) to \(\alpha ,\alpha ^{*}\rightarrow 1\). Using the intermediate value theorem and Proposition 3, a unique intersection exists between the two curves, \(\alpha ^{o}\), which satisfies \(0<\alpha ^{o}<1\). Hence, \({{\mathrm{sign}}}\{\gamma ^{m}-\gamma ^{*}\}={{\mathrm{sign}}}\{\alpha -\alpha ^{o}\}\).

  2. (ii)

    Since \(\phi >0\) and \(\alpha ^{*}>\alpha >0\) then the game has:

    $$\begin{aligned} \rho= & {} \frac{-c(2c+1)b^{2}+4c(1+c)b-(c+1)(2c+1)}{(2c+1)(1+c)}, \\ \iota= & {} \frac{-((2-b)+2c(1-b))((b+1)+c(3-b^{2})+2c^{2}(1-b^{2}))}{2(1+c)^{2}(2c+1)},\\ \varphi= & {} \frac{-(b-2-2c+2bc)^{2}((b+2)+2c(3-2b)+4c^{2}(1-b))}{8(2c+1)(1+c)^{3}},\\ \chi= & {} \frac{-(2c(1-b)+(2-b))^{2}(2c(1-2b)+1)}{4(1+c)^{2}(2c+1)}, \end{aligned}$$

    where \(\rho \),\(\iota \), \(\varphi \) are negative, and \(\chi \) might be positive or negative. \(\chi >0\) if and only if parameters (bc) belong to set B defined in Corollary 5. With endogenous public information, there exists a non-empty region such that \(\chi >0\) and \(\chi +\rho \frac{d\tau ^{m}}{d\tau _{\xi }}<0\) if and only if the payoff parameters (bc) and information parameters \((\tau _{\xi },\tau _{\theta },\tau _{u})\) jointly belong to set A defined in Corollary 5. Note that \(A\subset B\). Hence, the game is of type \(-{ IV}\) if \(A^{c}\cap B\), and of type \(+I\) if \(A\cup B^{c}\).

\(\square \)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Bayona, A. The social value of information with an endogenous public signal. Econ Theory 66, 1059–1087 (2018).

Download citation


  • Payoff externalities
  • Information externalities
  • Rational expectations equilibrium
  • Private information
  • Welfare analysis

JEL Classification

  • D62
  • D82
  • G14