Chapter 2 - Mapping GDP with Deep Feedforward Networks
A relevant task in applied economics is finding metrics that represent the current state of the economy. There is vast information out there, but it is not usually straightforward to isolate the signal from the noise. The gross domestic product (GDP) is the most-watched current state indicator of the economy, but due to the intrinsic complexity of its calculation, the official GDP data are usually published quarterly and with lag that can take months. As a result, many coincident indicators with a wide range of techniques have been developed in economics. In this work, we adopted deep feedforward networks, known as universal approximators, as an approach for mapping the GDP. With the end-to-end strategy allowed by this technique, we propose to map a set of high-frequency variables to build indicators that fit the GDP on a cross-sectional approach. The performance of this strategy with Brazilian data indicates that this approach must belong to the economic toolset.