Research interests:
Macroeconomics. computational economics, labor market rigidities, heterogeneous agent models, portfolio choice models.


"Méthodes de simulation des modèles stochastiques d'équilibre général (in french)" (with Michel Juillard), Economie et Prévision (Numéro Spécial), pp.115-126, Vol.2-3, 2008.


This paper presents numerical methods, which are commonly used today to solve dynamic stochastic general equilibrium models. Once we have introduced a canonical model of dynamic optimization which is in the heart of these problems, we review the methods of value function iteration, projection, parameterized expectation approach (PEA) and the perturbation method. The linearization, which is very popular in this literature, is presented as a special case of the perturbation method.

Working papers

"Nonlinearities in a DSGE Model with Search and Matching."


Empirical studies have shown that there are strong asymmetries in the growth of various labor market time series. Especially unemployment, employment, and vacancies have different dynamics during recessions and expansions of the economy. In order to generate the asymmetries, I investigate the role played by real rigidities in the labor market. The benchmark model used here resembles those described in Andolfatto (American Economic Review 1996) and Merz (Journal of Monetary Economics 1995). In this paper, I solve the model using first and second order perturbations in levels and projection methods with ordinary polynomials. This allows me to quantify the role of nonlinearities implied by the search and matching framework. The accuracy of the solution is verified by using parameterizations for the United States and the EURO area.

"Solving Dynamic Models with Heterogeneous Agents and Aggregate Uncertainty with Dynare or Dynare++ (with Wouter den Haan)"


This paper shows how models with heterogeneous agents and aggregate uncertainty can be solved using Dynare or Dynare++ software that implements a perturbation approach. Using the explicit aggregation algorithm (XPA) to obtain aggregate laws of motion is possible by combining a Dynare program with a very simple Matlab program. We calculate and compare 1st and 2nd-order numerical solutions using both algorithms. These numerical procedures are also compared with the algorithm that solves the individual policy rules with a projection instead of a perturbation procedure. Finally, we discuss a procedure that efficiently chooses which cross-sectional moments to include as aggregate state variables when nonlinearities are important and the mean is not a sufficient statistic.

Here are the simplified codes for this project: main.m Dimension4PF.mod fnConvergence.m fnSimulateAAKS.m. These codes require Dynare to be installed on your computer.

"Can Lower Order Perturbation Methods accurately describe Wealth Dynamics?"


This paper reveals the sensitivity of the solution method described in the previous chapter to the specification and parameterization of the borrowing constraint.
We find that the recent contributions that deal with perturbation methods in heterogeneous agent models ignore this aspect of sensitivity which can be detected in the behaviour of the tails. This leads us to conclude that more work is needed on the choice of borrowing constraints. We also briefly discuss the economic policy implications that the dynamics observed in the tails may have for instance in implementing fiscal and social policies that target the lower part of the wealth and income distributions.

"How does firm-level volatility affect household consumption?"


This paper studies the propagation mechanism of time-varying volatility from the production sector to the household sector in the presence of asymmetric adjustment costs. The model is an extension of the standard neoclassical growth model, in which we introduce heterogeneity on the production side, similarly to Bloom (Econometrica 2009). We use this framework to analyze the role of financial markets in households’ effort to smooth out consumption in the presence of second moment shocks. In order to isolate this effect, we assume the presence of households that do not have access to risky assets.
In this model, the distribution of capital is a high-dimensional state variable, which complicates the solution for the decision rules. The paper applies an iterative hybrid algorithm which combines third-order perturbation methods with the parameterized expectations approach, and approximates the stationary distribution through simulations.

"Limited participation and International Risk-Sharing (with Nicolas Coeurdacier and Hélène Rey)."


The "consumption real exchange rate anomaly" is one of the most subborn puzzle in open economy macroeconomics.We make substantial progress by introducing limited participation in an otherwise standard model. In particular, one of our methodological contribution is to show how we can solve for the characteristics and the dynamics of optimal portfolios in such a setup.

Work in progress

"Solving Dynamic Models using Hybrid Methods"

Last revised on 23 April 2014.