Oslo Summer School in Comparative Social Science Studies 2011

Estimation and Evaluation of Dynamic Stochastic General Equilibrium (DSGE) Models

Lecturer: Dr. David de Antonio Liedo

Main disciplines: Economics, Macroeconomics, Forecasting

Dates: 1 - 5 August 2011
Course credits: 10 pts (ECTS)
Limitation: 20 participants


"Hands-on" course designed to help the students program in Matlab the basic steps involved during the Bayesian estimation of dynamic stochastic general equilibrium (DSGE) models.  Several assumptions will be considered regarding the information set observed by both the econometrician and the representative agents characterized in the model. 

The course will make special emphasis on the need to validate those models. Our validation strategy consists in constructing forecasts based on DSGE models and compare them with those resulting from bayesian VARs or dynamic factor models. 


Students attending this course need to bring their own laptops with Matlab installed on it. Those students who have never used Matlab before, are required to at least take the following tutorial:


Essential texts needed for the course

  • Tommaso Mancini Griffoli, Dynare 4 User Guide (Public Beta Version): An introduction to the solution & estimation of DSGE models.
  • Edward Greenberg, Introduction to Bayesian Econometrics, Cambridge University Press, 2008.

Secondary texts

  • James D. Hamilton, Time Series Analysis, Princeton University Press, 1994.
  • Fabio Canova, Methods for Applied Macroeconomic Research, Princeton University Press, 2007



Lecture 1: Business Cycles under the lenses of DSGE models

What are DSGE models?. What are their limitations and what are their strengths for Business Cycle analysis? What will this course cover and not cover?


  • Fernández-Villaverde (2010) “The Econometrics of DSGE Models.” SERIES: Journal of the Spanish Economic Association 1, 3-49
  • Frank Smets & Raf Wouters (2004). "Forecasting with a Bayesian DSGE Model: An Application to the Euro Area" Journal of Common Market Studies, 42, 841-867,
  • Frank Smets & Kai Christoffel & Günter Coenen & Roberto Motto & Massimo Rostagno, 2010. "DSGE models and their use at the ECB," SERIEs, Spanish Economic Association, vol. 1(1), pages 51-65, March

Complementary reading:

  • Evans, G.W. and S. Honkapohja (2001), “Learning and Expectations in Macroeconomics" (Chapter 1)


Lecture 2: Solving linearized models

In this session, we will learn to characterize the equilibrium dynamics of a model under the rational expectations assumption.

Computing Session: Using DYNARE to solve our models and understanding the typical error messages found in the solution procedure.


  • Blanchard, O. and Kahn, C. (1980). The Solution of Linear Difference Models under Rational Expectations. Econometrica, 48, 1305–1313.
  • King RG, Plosser CI, Rebelo ST (2002) Production, growth and business cycles: technical appendix. Comput Econ 20:87–116
  • Tommaso Mancini Griffoli, Dynare 4 User Guide (Public Beta Version): An introduction to the solution & estimation of DSGE models.

Complementary readings:

  • Christiano LJ (1990) Linear-quadratic approximation and value-function iteration: a comparison. J Bus Econ Stat 8: 99–113
  • Christiano LJ (1990) Solving the Stochastic Growth Model by Linear-Quadratic Approximation and by Value Function Iteration. J Bus Econ Stat 8: 23–26
  • Lubik, T. and F. Schorfheide (2004) Testing for Undeterminacy: An Application to U.S. Monetary Policy. American Economic Review, 94, 190-217.


Lecture 3: The Kalman Filter

Using the Kalman Filter to evaluate the likelihood of a model. Why the likelihood cannot be evaluated when the number of structure shocks is smaller than the number of observables?

Computing Session: Use Matlab to write the DSGE model solution in state-space form and use the Kalman Filter to obtain the innovations that we will plug in the likelihood function.


  • James D. Hamilton, Time Series Analysis, Princeton University Press, 1994.


Lecture 4: The econometrician and the representative inhabitant of the DSGE World

Estimating the model requires some assumptions. The literature has proposed several options:

- They observe all the variables defined in the model (unbounded rationality)
- Agents make signal extraction errors (Pearlman et al. 1989)

- The econometrician observes (at least) a few of them (e.g. Smets and Wouters, 2007)
- The econometrician observes (at least) a few noisy measurements of those variables (e.g. Altug, 1989)
- The econometrician observes (at least) a few “rational” estimates of those variable (e.g. Sargent, 1989)


  • Sargent, T. J. (1989), "Two Models of Measurement and the Investment Accelerator," Journal of Political Economy, 97, 251-287
  • Pearlman, J., Currie, D., and Levine, P. (1986). Rational Expectations Models with Private Information. Economic Modelling, 3(2), 90–105.
  • Collard et al. (2009). “Imperfect Information and the Business Cycle” Journal of Monetary Economics, 56, 605-631

Complementary readings:

  • Altug, S. (1989), "Time-to-Build and Aggregate Fluctuations". International Economic Review, 30, 889-920
  • Watson, M. (1993). “Measures of Fit for Calibrated Models”. The Journal of Political Economy, 101, pp. 1011-1041
  • Orphanides, Athanasios, "Monetary policy evaluation with noisy information", Journal of Monetary Economics, April 2003, 50(3), 605-631
  • Townsend, R. M. (1983). Forecasting the forecasts of others. Journal of Political Economy, 91, 546–588.
  • Lui. Y. and Young, E. R. (2009). “Rational Inattention and Aggregate Fluctuations”. The BE Journal of Macroeconomics, 9
  • Sims, Christopher A. (2003). “Implications of Rational Inattention.” Journal of Monetary Economics, 50(3): 665–690.


Lecture 5: Estimation in practice 1/2

Likelihood based estimation of DSGE models.

Controlling the tightness of the priors.

Estimation of the posterior distribution with the MH algorithm.

Computing Session:Opening the Metropolis-Hastings algorithm black box. Using heat-maps to visualize the multivariate posterior distribution.


  • Ireland, Peter N., 2004. "A method for taking models to the data" Journal of Economic Dynamics and Control, Elsevier, vol. 28(6), pages 1205-1226
  • Sungbae An & Frank Schorfheide, 2007. " ,"Econometric Reviews", Taylor and Francis Journals, vol. 26(2-4), pages 113-172.
  • Frank Smets & Rafael Wouters, 2007. "Shocks and Frictions in US Business Cycles: A Bayesian DSGE Approach". American Economic Review, American Economic Association, vol. 97(3), pages 586-606
  • Edward Greenberg, Introduction to Bayesian Econometrics, Cambridge University Press, 2008.


Lecture 6: Estimation in Practice 2/2

Assigment for ECTS credits (Part I): Estimation of a model under several competing assumptions regarding the information set available to the econometrician and to the representative agent (see Lecture 4). The student, in groups, will estimate the model they choose under one of the four scenarios. Each member of the group will use a different starting value in the optimization algorithm. The leader of each group will decide how to organize the work, with the help of the instructor, so that the empirical results are briefly documented and posted in an intranet by the end of the week.


  • Siddhartha Chib and Edward Greenberg: "Understanding the Metropolis–Hastings Algorithm". American Statistician, 49(4), 327–335, 1995
  • Tommaso Mancini Griffoli, Dynare 4 User Guide (Public Beta Version): An introduction to the solution & estimation of DSGE models.


Lecture 7: VAR-based Validation 1/2

Using a BVAR to represent the main statistical features of the data.


  • Robertson and Tallman (1999). Vector autoregressions: Forecasting and reality, Economic Review, Federal Reserve Bank of Atlanta (1999).

Complementary readings:

  • Doan, T.,  R.  Litterman.  and  C.  Sims,  1984,  Forecasting  and  conditional  projection  using  realistic  prior  distributions,  Econometric  Reviews  3.  1 -100
  • Litterman.  R.B.,  1984,  Specifying  vector  autoregressions  for  macroeconomic  forecasting,  in:  P. Goel. ed..  Bayesian  inference  and decision techniques  with  applications:  Essays  in  honor  of  Bruno  de Finetti  (North-Holland,  Amsterdam).


Lecture 8: VAR-based Validation 2/2

Interpreting the DSGE model as a BVAR with strong cross-equation restrictions. How good are those restrictions? Model validation along several dimensions:

a) Impulse Response Functions: DSGE vs BVAR
b) Spectral Density: DSGE vs BVAR
c) Out-of-sample Point Forecasting/ Density Forecasting

Computing Session: Illustration of Ingram and Whiteman’s idea in Matlab. Using DYNARE to reproduce Del Negro & Frank Schorfheide’s methodology.


  • Ingram, B F and C H Whiteman (1994), Supplanting the `Minnesota' prior: Forecasting macroeconomic time series using real business cycle model priors," Journal of Monetary Economics, 34, 497-510.
  • Del Negro, Marco & Schorfheide, Frank & Smets, Frank & Wouters, Rafael, 2007. "On the Fit of New Keynesian Models,"Journal of Business & Economic Statistics, American Statistical Association, vol. 25, pages 123-143, April
  • Marco Del Negro & Frank Schorfheide, 2004. "Priors from General Equilibrium Models for VARS" International Economic Review, vol. 45(2), pages 643-673, 05


Lecture 9: DFM-based Validation

Using a Dynamic Factor Model to represent the main statistical features of the data.
Is there a mapping between DSGE models and dynamic factor models?

Model validation along several dimensions:

a) Impulse Response Functions: DSGE vs DFM
b) Spectral Density: DSGE vs DFM
c) Out-of-sample Point Forecasting/ Density Forecasting


  • Giannone et al. (2006) VARs, Factor Models and the Empirical Validation of Equilibrium Business Cycle Models”, Journal of Econometrics, Volume 132, 257-279.

Complementary readings:

  • De Antonio Liedo (2010) “General Equilibrium Restrictions for Dynamic Factor Models”, Banco de España Working Papers -1012
  • Giannone et al. (2004) “Monetary Policy in Real Time” In Mark Gertler and Kenneth Rogoff, editors, NBER Macroeconomics Annual 2004, Pages 161-200.MIT Press.
  • Marc P. Giannoni & Jean Boivin, 2005. "DSGE Models in a Data-Rich Environment, 431, Society for Computational Economics.


Lecture 10: Group Assignment for ECTS credits

Assigment for ECTS credits (Part II): By aggregating the estimation results by all students, we will be able obtain a very good characterization of the DSGE model posterior distribution. A necessary (not sufficient) condition for convergence in a given model is that all students report the same results.

The students that want to obtain ECTS credits for the course will be able to use such material in order to produce a 5 pages report (plus graphs) discussing the estimation method under some of the alternative assumptions discussed in Lecture 4. In addition, the report should illustrate why the different assumptions yield significantly different (or not) posterior distributions for the model parameters.


  • De Antonio Liedo (2010) “What Are Shocks Capturing in DSGE Modeling?”, Mimeo 


The lecturer
David de Antonio Liedo has been working as an economist in institutions such as the Bank of Spain and the European Central Bank, where he has become familiar with the use and interpretation of country models for the analysis of the Spanish economy, the euro area and the US. Dr. de Antonio's main area of expertise is the development and implementation of large scale models for short term macroeconomic forecasting and business cycle analysis. Dr. de Antonio has obtained his PhD. in Economics at the Université libre de Bruxelles (ECARES) and he is currently a Visiting Research Fellow at the London Business School.

Back to Main Page

Published Aug. 24, 2011 2:20 PM - Last modified Aug. 24, 2011 2:35 PM