A Toolkit for Endogenous Regime-Switching model Estimation

点击翻译按钮获取中文摘要

This proposal aims at developing a Toolkit for efficient estimation and use of a broad class ofstate-of-the-art time-series models for policy analysis and forecasting. Models that allowrelationships between variables and over time to evolve in a state-dependent fashion areessential for analysing time series that are subject to occasionally binding constraints. Theircomplexity and computational intensity create a significant barrier to entry for most potentialusers, including academics and policy makers. The main problem is that the estimation of (theunknown parameters of) these models currently relies on vast amounts of Monte Carlosimulations that are difficult to parallelize efficiently, resulting in exceedingly long computationaltimes. It is, therefore, unsurprising that most researchers and policy analysts still rely onscientifically unjustified solutions, leading to potentially biased forecasts and misleading policyprescriptions. We propose to address this problem by providing a novel method for theaccurate and efficient estimation of those models and by developing a Toolkit that implementsthe new method in a user-friendly environment. This method is both statistically andcomputationally more efficient than existing alternatives: it can be readily parallelized andprovides accurate estimates with orders of magnitude shorter computational times. Theproposal also aims to leverage the availability and accessibility of cloud computing at a scalethat has not been done before in this area. To initiate the commercialisation of the idea, we willbuild an advanced prototype and work with early adopters to test the prototype and prepare itfor the broader market. We will also perform a range of business-related activities to identifythe best route for commercialization. The project outcomes will be the advanced prototype, ananalysis of the competitive aspects of our technology, and a strategy regarding its route tomarket.