Discussion Paper
No. 2008-30 | October 09, 2008
Camilo E. Tovar
DSGE Models and Central Banks


Over the past 15 years there has been remarkable progress in the specification and estimation of dynamic stochastic general equilibrium (DSGE) models. Central banks in developed and emerging market economies have become increasingly interested in their usefulness for policy analysis and forecasting. This paper reviews some issues and challenges surrounding the use of these models at central banks. It recognises that they offer coherent frameworks for structuring policy discussions. Nonetheless, they are not ready to accomplish all that is being asked of them. First, they still need to incorporate relevant transmission mechanisms or sectors of the economy. Second, issues remain on how to empirically validate them; and finally, challenges remain on how to effectively communicate their features and implications to policy makers and to the public. Overall, at their current stage DSGE models have important limitations. How much of a problem this is will depend on their specific use at central banks. Submitted as Survey and Overview

JEL Classification:

B4, C5, E0, E32, E37, E50, E52, E58, F37, F41, F47


Cite As

[Please cite the corresponding journal article] Camilo E. Tovar (2008). DSGE Models and Central Banks. Economics Discussion Papers, No 2008-30, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2008-30

Comments and Questions

Anonymous - Referee Report
November 12, 2008 - 11:24
See attached file

Camilo Tovar - Reply to referee 1
November 15, 2008 - 20:25
“DSGE models and central banks” Response to the referee report No. 1 I would like to begin my response by thanking the referee for his comments. Secondly, it is welcomed and stimulating for me to find a report in which the referee states that “[r]eviews and summaries like this one are much needed since it is important to carefully appraise what has actually been achieved in the science to avoid mistakes that occurred in the past”. Equally, encouraging is that the report concludes that “[a]ll in all, this paper provides a useful summary of [the] problems […]”, and that ”[t]he review offers valid points and the issues and challenges discussed have to be on the agenda of policy institutions which are interested in the usefulness of DSGEs for policy analysis and forecasting”. These comments suggest that the review and the challenges and issues raised in the paper have achieved the goal of identifying a topic that is of central interest for academics and the central banking community. Furthermore, I fully agree with the referee when he states that we need to discuss what it means to develop these models and what do they really offer in terms of policy analysis. In this respect, my paper is a contribution in that direction. Said this, I would like to comment/respond to some of the remarks made in the referee report. Comment 1: “[…] I fear that the paper only provides a useful study for a more limited audience than the title might suggest. Under the assumption that the intended reader is someone looking for an introduction an overview of the academic issues, the presentation is a bit too detailed without much value added compared to the sources cited. On the other hand, if the intended reader is someone looking for an overview of the policy issues in general and the interaction of DSGE models in the policy process (ie practical implementation issues) the review comes too short and without much guidance for the reader”. Answer: It is difficult to find the appropriate balance for all audiences. The aim was to provide a comprehensive overview raising not just technical matters that are of relevance when thinking about using these models in policy making, but also in considering some issues that appear to be relevant when employing these models in the policy process. With this in mind, and in light of the comments, some issues are worth considering: First, the paper was submitted as a “survey and overview”, which according to the “Aims and scope” of the journal is to address a general audience interested in economic issues. Second, readers should keep in mind what the paper aims for. As stated in page 5 (and also in the abstract) the paper examines three broad issues: i) the structure of DSGE models; ii) the empirical validation of these models; and, iii) for a successful implementation of these models for policy analysis, how to communicate effectively the features and the implications of the model for policy makers and the public. Therefore, “[…] without attempting to be an exhaustive review of the literature the article highlights, in a non-technical manner, some of the issues and challenges arising from these three questions”. Third, the referee report states that the paper falls short in the overview of the policy issues and the interaction of DSGE models in the policy process. It would be relevant to have a more specific idea of what the report meant by this, in particular, given that the paper offers: a) A discussion of different views about how should DSGE models be employed for policy analysis and forecasting (page 5).b) Section 2 discusses the main elements of the benchmark DSGE model employed for policy analysis, together with a discussion of elements missing in that benchmark DSGE model and why they are important. This includes a section devoted to the challenges of employing these models in emerging market economies (EMEs).c) A discussion of why DSGE models are attractive and why they are not yet in widespread use. In this respect, it is highlighted that the attractiveness of these models is the possibility of taking to the data a model that is fully derived from first principles. However, the fact that they have been taking to the data does not mean that no challenges remain. That is why considerable attention is devoted to discuss not just the requirements that these models impose in terms of the information required, but also the challenges and weakness arising from employing different estimation methods.d) A discussion of the challenges arising in policy evaluation. In particular, the paper examines the achievements made in terms of forecasting with DSGE models and discusses whether parameter estimates are truly structural or not.e) A section review challenges arising from the communication of DSGE models. Finally, given that the referee report expresses a concern about the title being too comprehensive it is always possible to modify it. I would be willing to consider an alternative option, nonetheless, I believe that the overview clearly relates to “DSGE models and central banks”, in this respect the contents does justice to its title. Comment 2: “The paper mainly discusses issues related to academic oriented problems”. “[t]his may be a natural division since many of the research problems need to be addressed before policy questions can come into question. Nonetheless, it is an unfortunate imbalance between the sections since central banks use DSGE models in two equally important areas: research and policy analysis (the ECB has e.g. one division dealing with research namely DG Economic Research and one division dealing with more direct policy issues namely DG Economics”. Answer: As highlighted above, in writing an overview it is always difficult to achieve the best balance about its contents. In fact, those who read might have different interests making them focus on different aspects of the discussion. For instance, policy makers might look for a quick overview of the academic material, which they could otherwise find difficult to review directly from the literature. In this respect, I do not see why the referee report considers the academic emphasis an “imbalance”, and much less why he considers this as “unfortunate”. In fact, as the referee report correctly indicates many research questions often need to be addressed before policy questions can be addressed. In any case, as stressed in the previous answer, the paper has an emphasis that goes beyond purely academic issues. Also the referee seems to fail to notice that the central banking community is wide and heterogeneous: there are (few) large central banks with resources that allow them to put in place separate research teams and policy departments. However, there are also a (large) number of central banks, where resources are more limited, including the availability of a large mass of human capital. In this respect, the paper is not just concerned about the large central banks, but also about central banks in emerging market or developing countries. In this respect, I believe that this is a valuable overview for researchers at central banks who do not have the benefit of time for reading all the available academic literature on DSGE modeling. It also can also provide a useful overview to decide whether it is worth devoting resources to building their own DSGE model. Comment 3: “It is difficult to understand the point of reference or benchmark when discussing modeling problems. This is particularly relevant in section 2 and 3 where modeling challenges and data issues are discussed. It is hard to understand whether the issues are specific to DSGE models or if the issue at hand also applies more broadly”. Answer: There are two angles for a reply to this comment. The first relates to whether the paper specifies or not what the benchmark DSGE model is. In this respect, in the first part of Section 2, page 6, the paper provides a full description of the elements contained in the benchmark DSGE model, which is mainly associated with the Christiano et al (2005) and Smets and Wouters (2005) models (for references see the paper). In this respect, the paper does provide a description of what is considered the benchmark DSGE model. Furthermore, as stated in page 7, it is this benchmark model that is challenged by an incomplete structure as reflected by the omission of financial frictions, the Ricardian equivalence or the lack of modeling features of relevance for EME’s among others. The second, is related to section 3, and in particular, to section 3.1 where I discuss some issues related to data sets. Certainly, it is possible to adjust the discussion and warn the reader that some issues do apply more generally to other macroeconomic models. Nonetheless, the reason why this discussion is here, is because, the attractiveness of DSGE models for policy analysis is that fully micro-founded general equilibrium models can now be taken directly to the data. Since other models employed at central banks are not fully micro-founded in a general equilibrium framework, the data requirements and their implications are different. In other words, what is at hand here is whether the full implications of the DSGE models can be mapped into the data, and the discussion implies that we are not quite there yet. Overall, I can make the appropriate adjustment to this comment in a revised version of the paper to clarify the concern. Comment 4: “If the purpose of using a DSGE model is to be able to do counterfactual policy analysis, the fact that it forecasts worse than e.g. a vector autoregression with Bayesian techniques (BVAR) is not very interesting. The benchmark for comparison in this case – the BVAR- does not provide an alternative framework for the task at hand.” Answer: I completely agree with the referee report that if one is interested in a counterfactual analysis, whether a DSGE is able to forecast or not is of little relevance. The issue here is that academics and central banks have in mind different possible uses for DSGE models. As quoted in page 5 “[…] it remains an open issue how should DSGE models be employed for policy analysis and forecasting at central banks”. Doing counterfactual analysis is just one of the possible uses, forecasting is another. Therefore since forecasting is a possible use for DSGE models at central banks, then it is relevant to discuss whether it performs better or not than other available forecasting tools available at central banks, such as BVARs. It also makes sense to discuss were do DSGE models fit among all the forecasting tools available. Comment 5: “[t]he review comes out as far too pessimistic and critical when it comes to the use of DSGE models for policy analysis […]. If one instead would take on a weaker or minimal econometric interpretation of models in general and DSGEs in particular the conclusion may looked something like this: “The rapid progress made in recent years has made, at their current stage of development, these models ready to accomplish all what is being asked from them (but not more). Like all models they still fail to have a sufficiently articulated description of the economy, which force them to highlight the most important interactions in the economy for the questions at hand.” ”. Answer: Readers have different views and it is always difficult to accommodate all of them. Some people might take a negative view while others may be more positive. The references and the arguments are in the paper for readers to think and delve further into the topic. Nonetheless, I do not agree that the paper is “far too pessimistic”. In fact, in page 4 it reads “[…], it must be recognized that a lot of progress has been made with DSGE models. Even at their current stage of development these models have already proven to be useful for central banks. In fact a number of institutions across the world have employed these models to analyze relevant policy issues”. The paper then goes on to provide specific examples, both in performing counterfactual analysis or in forecasting at central banks in advanced economies or in EMEs. In addition, I should add that Section 6 opens with the following statement: “DSGE models are powerful tools that provide a coherent framework for policy discussion and analysis. In principle, they can help to identify sources of fluctuation; answer questions about structural changes; forecast and predict the effect of policy challenges, and perform counterfactual experiments”. I hardly find this to be “too pessimistic”. Overall, it is a responsibility of any researcher to think beyond and ask whether the tools employed: i) offer what they are said to be offering; and ii) whether we can improve the use of the tool or not. Finding an answer to these questions were elements that motivated me to write the paper. To finalize, the referee report suggested I should add a reference by Faust (2008). I have read the paper and would like to quote the following from this article: “Excitement over these incredibly important advances should not blur our vision over what of practical relevance has actually been achieved. While we have generated a family of models that broadly matches business cycle features, existing implementations of these models are little better, and may be worse, than the models of the 1970s from the standpoint of the major critiques of those older models. We have a promising jumping off point for assaulting those problems, but considerable work remains to be done”. I wonder, if my conclusions are so pessimistic, why does the referee report suggest adding a reference to the paper that is, to say the least, not as optimistic as the referee report suggests? In this respect, it appears that my conclusions are shared by many: “[…] these models are not fully ready to accomplish all what is being asked from them.”

Anonymous - Referee Report
November 14, 2008 - 12:03
See attached file

Camilo Tovar - Response to Referee Report No. 2
December 09, 2008 - 05:45
Response to Referee No. 2 I want to thank the referee for his report and, particularly, for pointing me to some interesting references. DSGE modeling is an active area of research and it is normal that new interesting and important work appears within a short span of time. In this respect, it is also welcome to confirm that research is moving into some of the directions discussed in the paper. In what follows, I would like to comment on some issues raised in the referee report: 1.- The referee report suggests adding some references to complement the discussion in some parts of the paper. I will certainly do so in a revised version of the paper. 2.- I agree that the fact that DSGE models are misspecified should not limit their use and that what this suggests is a need for further analysis and research. Notice that the perspective of the paper is to highlight that further work is required in different areas, and that people should be aware of the current strengths and limitations of DSGE models, particularly when applied to policy analysis and forecasting. In this sense, I believe that the discussion in the current version of the paper can be fine tuned to reflect this view better. 3.- I also agree that measurement errors and informative priors can help limit the effects of misspecification on parameter estimates. Notice, for instance, that the paper does recognise that misspecification can be handled in specific manners eg adding measurement errors (see page 21) or using informative priors in a VAR-DSGE framework (page 19). Nonetheless, as I also mention in the paper, these approaches are also subject to criticism. 3.- The referee has an interesting suggestion about testing the robustness of the model’s implications to different priors. This is worth making explicit in a revised version of the paper. 4.- I think that the discussion in page 22 can be adjusted to incorporate more explicitly the comment that identification of parameters in DSGE models is weak and that the policy prescription might be strongly dependent on the researcher’s prior.