Stuart McIntyre (p), Gary Koop , James Mitchell

Data on Gross Value Added (GVA) are currently only available at the annual frequency for the UK regions and are released with significant delay. Regional policymakers would benefit from more frequent and timely data. The goal of this paper is to provide these. We use a mixed frequency Vector Autoregression (VAR) to provide, each quarter, nowcasts (i.e. forecasts of current GVA which is as yet unknown due to release delays) of annual GVA growth for the UK regions. The information we use to update our regional nowcasts comes from GVA growth for the UK as a whole as this is released in a more timely and frequent (quarterly) fashion. To improve our nowcasts we use entropic tilting methods to exploit the restriction that UK GVA growth is a weighted average of GVA growth for the UK regions. In this paper, we develop the econometric methodology and test it in the context of a real time nowcasting exercise.

Saket Sarraf (p), Arnab Chakraborty

Models can be evaluated based on their underlying theory, data, estimation procedures, and outcomes. Ideally, model assessment should be based on a combination of all these components.

Integrated models are often created using a combination of theory and data driven procedures to capture a range of dynamics. They often foray into areas which would demand a grand unified theory, where none exists. It is almost next to impossible to validate such models for real life applications due to limited theory, use of large amount of data from varied sources, and their attempt to model complex causal processes which offer limited direct observation.

This makes the task of establishing credibility among the end-user community very challenging. They may not understand the underlying theory, the estimation processes or interact with range of complex data. What’s most appealing to them is the robustness of model outcomes and its ability to present convincing narratives supported by a causal chain of events. Thus, there is a strong need among the end-user community for a set of procedures to ascertain the model performance purely based on its outcomes.

This paper explores different methods to assess model performance in terms of its outcomes from an end-users’ perspective. It uses the newly developed E3 India model for exposition of these methods. E3 India is a multi-sectoral multi-regional economic model based on coupled input-output econometric approach. It is envisioned as an off the shelf, end-user tool for energy related policy analysis. It is the first model of its kind in the Indian context that explores the linkage between energy, economics and environment.
This work in progress paper aims to make methodological contribution to the field of model assessment in the context of integrated models. The application of these methods to E3 India model will help end-users ascertain its credibility, and importantly provide feedback to the modeling team to improve the model performance.