It is important to understand the regional evolution of economic activity and how outcomes in different regions are linked to one another. This helps policymakers estimate the impact of planned fiscal stimulus programmes and it improves the ability of financial institutions to predict the spread of shocks.
In this project, the regional network structure of the economy will be determined based on business cycle co-movements between areas. Business cycles being the fluctuation in economic activity that an economy experiences over a period of time, and co-movement the tendency of variables such as price, productivity, and investment, to move together in a predictable way over this period.
It will then be tested as to whether business cycle co-movement is better explained by geographic proximity or by supply networks, with the outcome of this used to simulate how shocks spread across the economy.
This work will use interesting techniques from machine learning to reduce the size and complexity of the data, including mixed membership stochastic block-models. Initially the work will be based on US data but the plan is to use transaction data for the UK once it becomes available.