Macro II - Fluctuations - ENSAE, 2024-2025
2025-03-19
What are the classical reasons to open economy to trade?
RBC models have been very successful at matching Business Cycles
It didn’t take long before the same methodology was applied to International Business Cycles
Seminal Paper:
Very successful methodology:
Moments
From Kehoe,Kydland (1995)
Domestically:
Internationally:
Can we replicate these moments with a BC model?
Representative agents maximizes: \[\max_{c_t} \sum_{t=0}^{\infty} \beta^t u(c_t)\] \[c_t + a_{t} \leq y_t + (1+r) a_{t-1} \]
Endowment economy:
Small open economy:
We solve this problem with the terminal conditions:
\(a_{-1}\) given1
\(\lim_{T\rightarrow\infty} \frac{a_{T}}{(1+r)^T}\geq0\)
The no-ponzi condition will in effect eliminate diverging solutions.
In a first order approximation, it selects the right eigenvalues.
We get the lagrangian:
\[\mathcal{L}= \sum_{t=0}^{\infty} \beta^t u(c_t) + \sum_{t=0}^{\infty} \beta^t \lambda_t \left(y_t + (1+r) a_{t-1} - c_t-a_{t} \right)\]
First order conditions:
\[\begin{align} u^{\prime}(c_t)& =& \lambda_t \\ \lambda_t &=& \beta (1+r) \lambda_{t+1} \end{align}\]
Under the technical assumption \(\beta (1+r)=1\) we get \(c_t=c_{t+1}\) then
\[c_0 = \frac{r}{1+r}\left\{ (1+r) a_{-1} + \sum_{t=0}^{\infty} \frac{y_t}{(1+r)^t}\right\}\]
Reminders on Current Account
The trade balance is exports-imports (here \(y_t-c_t\))
The current account is trade balance + net factor payments (here \(y_t-c_t+r a_{t-1}\))
Positive current account: additional lending to the rest of the world.
Using the formula from before
\[CA_0 = a_{-1} r + (1-\frac{r}{1+r}) y_0 - \frac{r}{1+r}\left\{ \sum_{t\geq1}^{\infty} \frac{y_t}{(1+r)^t}\right\}\]
How does the current account reacts to income shocks?
current account responds positively to temporary shock in income
and to news about future income shocks:
Still with the same formula: \[c_0 = \frac{r}{1+r}\left\{ (1+r) a_{-1} + \sum_{t=0}^{\infty} \frac{y_t}{(1+r)^t}\right\}\]
What is the effect of an increase in \(a_{-1}\)?
- consumption rises permanently
- $a_t$ is constant, equal to $a_{-1}$
- agent consumes small amount $r$ corresponding to interests paid forever on $a_{-1}$
We add capital and production to our endowment economy: \[y_t = z_t k_{t-1}^\alpha\] \[k_t = (1-\delta) k_{t-1} + i_{t}\]
The aggregate resource constraint becomes:
\[a_{t} + c_t + i_t = (1+r) a_{t-1} + y_t\]
Now maximize \(\sum_t \beta^ t U(c_t)\)
We get first order conditions
\[\lambda_t = \beta \lambda_{t+1} (1+r)\] \[\lambda_t = \beta \lambda_{t+1}\left[ (1-\delta) + z_{t+1} f^{\prime}(k_{t}) \right]\]
where \(\lambda_t\) is lagrange multiplier associated to budget constraint.
Since \(\lambda_t>0\) (constraint is always binding), we get:
\[(1-\delta) + z_{t+1} f^{\prime}(k_{t}) = 1+r\]
\[k_{t} = \left( \frac{r+\delta}{\alpha z_{t+1}}\right)^{\frac{1}{\alpha-1}}\]
and investment \[i_t = \left( \frac{r+\delta}{\alpha z_{t+1}}\right)^{\frac{1}{\alpha-1}}- (1-\delta)\left( \frac{r+\delta}{\alpha z_{t}}\right)^{\frac{1}{\alpha-1}}\]
Here investment is fully determined by productivity shocks
A possible solution: change the resource constraint such that adjusting capital is costly
For instance:
\[a_{t} + c_t + i_t + \frac{\omega}{2}\frac{(k_{t}-k_{t-1})^ 2}{k_t} = (1+r)a_{t-1} + z f(k_{t-1})\]
\[k_{t} = (1-\delta) k_{t-1} + i_t\]
where \(\omega\) is an adjustment friction.
Typically, \(\omega\) is chosen so that the model replicates \(\frac{Var(i_t)}{Var(y_t)}\) from the data.
🔜 Cf tutorial.
Closing Small Economy Models, Schmitt Grohe and Uribe (2003), JIE
\[\max_{c_t, n_t} \sum_{t=0}^{\infty} \beta^t u(c_t, n_t)\]
\[c_t + k_{t} + a_{t} = y_t + g_t - \frac{\omega}{2}(k_{t}-k_{t-1})^2 +(1-\delta) k_{t-1} + (1+r^{\star}+{\color{red}\pi(a_{t-1})})a_{t-1}\] \[y_t = f(k_{t-1}, n_t, z_t)\]
\[z_{t+1} = \rho z_t + \epsilon_{t+1}\]
and \[u(c, n) = \frac{1}{1-\sigma}\left(c^{\psi}(1-n)^{1-\psi} )\right)^{1-\sigma}\]
The term \(\color{red}\pi\) is there to make the model stationary.
The solution of the model exhibits a unit root:
\[a_t = a_{t-1} + ... \text{other variables in t-1} + \text{shocks in t}\]
Problem:
This raises practical issues (notably for estimation) for the linear model.
General idea:
Schmitt Grohe and Uribe (2003) consider many options:
SGU show that the choice of the stationarization device has little effect for the dynamics (moments) of most variables
Parameters | Values |
---|---|
\(σ\) | 2 |
\(ψ\) | 1.45 |
\(α\) | 0.32 |
\(ω\) | 0.028 |
\(r\) | 0.04 |
Parameters | Values |
---|---|
\(δ\) | 0.1 |
\(ρ\) | 0.42 |
\(σ²\) | 0.0129 |
\(A^{\star}\) | -0.7442 |
\(χ\) | 0.000742 |