I am having the following setup of privately known signals and I am trying to understand an assumption. Here, I quote the setup.

Consider two agents idexed by $i=\{1,2\}$ and each one observes some privately noisy signal, that is \begin{align}s_i=\tilde{v}+\tilde{u_{i}}\end{align} where $(\tilde{v},\tilde{u_{i}})$ are identically distributed, with $\tilde{v}\sim N(v,\sigma_{v}^2)$ stand for the payoff of the risky asset and $\tilde{u_{i}}\sim N(0,\sigma_{u_{i}}^2)$ denotes the error term of each of the privately known signals. Also, it holds that $\sigma_{(u_1,u_2)}\neq 0$ and $\sigma_{(v,u_1)}=0=\sigma_{(v,u_2)}$.

The assumption that I am trying to understand following the aforementioned setup is the following:

``Every informed agent does know the cross sectional distribution of the privately knonw noisy signal that the other agent observes." Also in the market, there are uninformed traders etc. Does this mean, that in fact they know exactly the signal that each other observes or do I catch it wrong?

  • 1
    $\begingroup$ Hi: It sounds to me like an informed trader $j$ say, knows $v$, $\sigma^2_{v}$ and $\sigma^2_{u_{i}}$ of every other trader $i$, whether $i$ is informed or uninformed. So, this is not the exact signal because these are just the parameters of the normal distributions of the two components of the unknown signal. $\endgroup$
    – mark leeds
    Commented Aug 2, 2020 at 3:47
  • $\begingroup$ Well, I believe that you are right. I have the same feeling, after I read it again and again. $\endgroup$ Commented Aug 2, 2020 at 5:12
  • $\begingroup$ if it's a paper, then probably best to confirm with the author but I think that's right. $\endgroup$
    – mark leeds
    Commented Aug 2, 2020 at 19:53
  • $\begingroup$ I have not seen it in some paper...so that is why i am not making any reference...but I think that in Glosten (1989) you can find a similar assumption. Specificaly in Glosten's paper it says ``A typical market maker is assumed to be risk neutral, facing no inventory costs or constraints and ignorant of the endowment vector and the realization of the information. He does know the crosssectional distribution of these variables"... $\endgroup$ Commented Aug 3, 2020 at 6:39
  • 1
    $\begingroup$ Gotcha. In that case, I think what we're saying is correct. $\endgroup$
    – mark leeds
    Commented Aug 4, 2020 at 11:48

1 Answer 1


Yes, that assumption just means that every informed agent knows the true distributions where signals come from (i.e. roughly speaking, they know the parameters that define them).

The terminology "cross-sectional distribution" comes from the usual interpretation that instead of having literally 2 agents, in reality, you have a continuum of agents that come in 2 types. If that is the case, due to the law of large numbers, the empirical distribution of signals at any point in time will be equal to the theoretical distribution of signals. Therefore, knowing the cross-sectional distribution is the same as knowing the theoretical distribution. Since these distributions are normal, the latter corresponds exactly with knowing the mean and variance of signals of the other (type of) agent.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.