I am studying a rather standard DSGE Model with a standard Utility function $U = f(c_t, n_t, M_t)$ subject to a budget constraint. Now, tosolve the intertemporal optimization problem I have, or everyone has, to deal with an expectation term; often similar to $E_t \left[\lambda_{t+1}\right]$, where $\lambda_t$ is the standard Lagrangian multiplier. Is there an anlytical trick to solve this? It seems to me the that usually the Expectation is just erased (without a justification) or studied using MLE estimators or Kalman Filters and surrogates. Thanks to whoever can enlighten me on the issue. Cheers

Edit 1: $\max_{c_t, n_t, i_t, M_t, K_t, B_t}{\mathbb{E}_t}\left[ \sum_{t=0}^{\infty} \beta^t \left( \log{(c_t^i)} - \gamma_i(n_t^i)^{2}+ \log \left(\frac{B_t}{P_t} \right) \right) \right]$ under the constraint: $\frac{M_t}{P_t}+ c_t + i_t + \frac{B_t}{P_t(1+r_t)} \leq u_t n_t + \frac{Q_t}{P_t}K_t + \frac{B_{t-1}}{(1+\pi_t)P_{t-1}} +\frac{M_{t-1}}{(1+\pi_t)P_{t-1}}$ The issue is wrt to a term which comes up in both the maximisation wrt to Bonds and Money: $\beta {\mathbb{E}}_{t}\left[\frac{\lambda_{t+1}}{(1+\pi_{t+1})}\right]$ - how could I solve it? (with theother terms it would lead $ c_{t}^{-1}=\beta {\mathbb{E}}_{t}\left[\frac{c_{t+1}^{-1}}{(1+\pi_{t+1})}\right]$

  • 1
    $\begingroup$ Can you provide more information regarding the optimization problem? Budget constraint, choice variables, et cetera $\endgroup$
    – Brennan
    Oct 31 '19 at 18:16
  • $\begingroup$ Hi Brennan, I edited my questions with more infos, Cheers $\endgroup$
    – dadelutz
    Nov 4 '19 at 17:41
  • $\begingroup$ What is stochastic at time $t$ and what are your state variables? You've just provided detail on the period $t$ problem and not what the uncertainty and dynamics are about. $\endgroup$
    – user26098
    Apr 2 '20 at 19:47

It really depends on the problem. As to why it would disappear, sometimes you can express the multiplier in terms of variables at the present or past, as in $\lambda_{t+1}=\lambda_{(c_t,n_t,M_t)}$ , which means it's an expectation over values that are known at that time. Therefore, you have $E_t[\lambda_{t+1}]=E_t[\lambda_{(c_t,n_t,M_t)}]=\lambda_{(c_t,n_t,M_t)}=\lambda_{t+1}$

But, as said above, it depends on the problem.

  • $\begingroup$ Thanks, do you think your idea would be appliable to my issue? I edited my question with more infos. $\endgroup$
    – dadelutz
    Nov 4 '19 at 17:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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