I try to solve the following maximization problem of a representative household with dynamic programming. However, my last result is not similar to the solution. Could any one help me?

$$\max\limits_{C_{t},H_t,N_t} E_0 \sum_{t=0}^{\infty} \beta^t\Bigg[logC_t+jlogH_t-\dfrac{(N_t)^\eta}{\eta}\Bigg]$$ $C_t,H_t,N_t$ represent consumption at time t, housing stock and working hours, respectively. Household budget constraint:

$C_t + b_t +q_t(H_t-H_{t-1})=\dfrac{R_{t-1} b_{t-1}}{\pi_t}+w_t N_t+F_t$

$b_t$ denotes bank deposits, $R_t$ is the return from deposit, $q_t$ is the price of housing, $\pi_t$ is the inflation rate and $w_t$ is the real wage rate. $F_t$ denotes profits received from firms. To sole this problem, I use dynamic programming with these steps: Step 1: Bellman equation:

I identify $b_{t-1}$ as the state variable, since the household must know the deposit in the previous period. The budget constraint can be rewrite as: $b_t=f(C_t,b_{t-1},H_t,H_{t-1},N_t,F_t,q_t,R_{t-1},\pi_t)$

Step 1: Setting up the Bellman equation:

$V(b_{t-1})=\max\limits_{C_{t},H_t,N_t} \Bigg\{logC_t+jlogH_t-\dfrac{(N_t)^\eta}{\eta} +\beta E_tV(b_t) \Bigg\}$

Evolution of the costate variable $b_t$, using envelope theorem:

$$\dfrac{\partial V(b_{t-1})}{\partial b_{t-1}} =\beta E_t \dfrac{\partial V(b_{t})}{\partial b_{t}} \dfrac{\partial b_t}{\partial b_{t-1}} = \beta E_t V'(b_t) \dfrac{R_{t-1}}{\pi_t}$$ (1)

Step 2: FOC

First order condition (FOC1) $$\dfrac{\partial V(b_{t-1})}{\partial C_t}=\frac{1}{C_t}+\beta E_t V'(b_t)\dfrac{\partial b_t}{\partial C_t}=0$$ $\dfrac{\partial b_t}{\partial C_t}=-1$

$\dfrac{1}{C_t}=\beta E_t V'(b_t)$ (2)

Inserting (2) into (1) twice to get the Euler equation, we have:

$$\dfrac{1}{C_t}=\beta E_t \Bigg(\dfrac{R_t}{\pi_{t+1 C_{t+1}}}\Bigg)$$

This result is similar to the solution

FOC2: $$\dfrac{\partial V(b_{t-1})}{\partial N_t}=-(N_t)^{\eta-1}+\beta E_t V'(b_t)w_t=0$$ Inserting (2), we have: $\dfrac{1}{C_t} w_t = -(N_t)^{\eta-1}$, This result is similar to the solution

FOC3: The third FOC involves taking the partial derivative of $V(b_{t-1})$ with $H_t$, but I cannot get the same result with the solution: The solution is: $$\dfrac{j}{H_t}=\dfrac{1}{C_t}q_t-\beta E_t \dfrac{1}{C_{t+1}}q_{t+1}$$

I only end up with:

$$\dfrac{\partial V(b_{t-1})}{\partial H_t}=\dfrac{j}{H_t}+\dfrac{1}{C_t}(-q_t)=0$$ Did I miss something? Did I correctly identify the state variable? I suspect that it has something to do with $H_t$ and $H_{t-1}$, but I cannot figure it out


1 Answer 1


$$V(b_{t-1})=\max\limits_{C_{t},H_t,N_t} \Bigg\{\ln C_t+j\ln H_t-\dfrac{(N_t)^\eta}{\eta} +\beta E_tV(b_t) \Bigg\}$$


$$\beta E_tV(b_{t})=\beta E_t\left[\max\limits_{C_{t+1},H_{t+1},N_{t+1}} \Bigg\{\ln C_{t+1}+j\ln H_{t+1}-\dfrac{(N_{t+1})^\eta}{\eta} +\beta E_{t+1}V(b_{t+1}) \Bigg\}\right]$$

But forwarding the budget constraint

$$C_t + b_t +q_t(H_t-H_{t-1})=\dfrac{R_{t-1} b_{t-1}}{\pi_t}+w_t N_t+F_t$$

we get after re-arranging,

$$C_{t+1} =- b_{t+1} -q_{t+1}(H_{t+1}-H_{t})+\dfrac{R_{t} b_{t}}{\pi_{t+1}}+w_{t+1} N_{t+1}+F_{t+1}$$

So $\beta E_tV(b_{t})$ is also a function of $H_t$ through $C_{t+1}$.


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.