Skip to main content
European Commission logo print header

Developing a Bayesian technique to evaluate the validity of economic models: two applications to the Euro Area

Periodic Reporting for period 1 - BMC-SVAR (Developing a Bayesian technique to evaluate the validity of economic models: two applications to the Euro Area)

Okres sprawozdawczy: 2017-09-01 do 2019-08-31

When economists take research questions to the data, some type of prior views on the results is both inevitable and necessary. In fact, the data do not uniquely speak with one voice about what drives the data. It is then useful to approach the data with some views, called in Bayesian technical jargon, “prior beliefs”. The researcher combines such views with the data in order to form “posterior beliefs”. These can then be used as a basis to answer the economic question of interest. A preliminary condition for this to happen, though, is that there exist economic tools of analysis that can flexibly combine data with beliefs, and can do so in a technically non-demanding way.
Before the work of the Fellowship, the tools available for economic analysis were relative restricted. On the one hand, one needs tools that allow for the technical analysis not to be computationally demanding. In fact, if the methodology used to combine data and beliefs requires a very long time to run on the computer or if it requires knowledge beyond what taught in graduate programmes, then the ability to address economic questions is considerably impaired. On the other hand, though, the researcher requires the flexibility to form his or her own beliefs about the phenomenon studied in the analysis. Before the work of this Fellowship, the literature frequently constrained the flexibility allowed for in the formation of prior beliefs for the benefit of making the analysis computationally more tractable. In other words, only special cases of prior beliefs could be handled, since any other one would make the analysis too technical and too costly to run. Unfortunately, this was the case irrespectively of whether the special prior beliefs allowed for were too restrictive and at odds with the views of the researcher.
The key objective of the Fellowship is to offer one step forward to overcome the above challenge. The analysis now offers a new methodological framework in which the researcher can now entertain a wider class of prior beliefs. Importantly, doing so now does not come at a computational cost. This enriches the tools of analysis available to the research community at large. The analysis was carried out in the interest of academic and policy making researchers, who apply economic tools to a variety of economic questions. Research questions that can be addressed with the new methodology cover different types of Applied Economics, ranging from the study of monetary policy to the effects of fiscal stimulus, and to other sources of shocks that hit the economy.
The Fellowship started with a period of research and training aimed at understanding the literature in greater details. It required new training in the understanding of Bayesian Econometrics, and further deepening of the knowledge of the trade-off faced by researchers between flexible prior beliefs and tractable analysis of the results. The main achievement of the Fellowship has been to propose a new methodology that increases the flexibility available to researcher in the way it forms prior beliefs, but does so without a cost in terms of the computational ability to explore the results. The key intuition is developed using simulations and an application to US data. A second application uses the new methodology to show that the European Central Bank can exert quantitatively sizeable effects on inflation. This reduces the risk that deflationary spirals remain unchallenged by the monetary interventions of the European monetary authority.
The achievements from the Fellowship were disseminated through
• presentations at universities and at central banks;
• promotion of the new economic tool at workshops organized under the support of the Fellowship;
• publication of work-in-progress versions of the work, and
• distribution of publicly available teaching material that aims to make the analysis approachable to applied researchers without strong expertise in Bayesian Econometrics.
The analysis carried out during the Fellowship opens the door for a new approach to Applied Economics, and will be explored further in the years to come.
The analysis of the Fellowship pushes the frontier of research further by providing a new tool available for economic analysis. The state of the art before the Fellowship faced a trade-off between the required flexibility in the formation of prior beliefs, and the computational tractability of the analysis. After the Fellowship, the state of the art can now use a tool that makes this trade-off disappear.
Impact is expected on the community of applied researchers, with a particular focus on central banks. Since central banks frequently need to take research question to the data, enriching the tools of analysis that is available to them increases the ability to address a variety of research questions, from how developments in uncertainty affect the economy, to how successful central banks can be in achieving their targets. In addition, the Fellowship offered support to begin a new Workshop in Structural VAR models, which aims to provide a forum for discussion of recent developments in Applied Economics. The workshop is in its third edition, scheduled for June 2020, and is now an established annual meeting in the economic community working on structural analysis in time series.
Participants to the second edition of the Workshop in Structural VARs, May 2019