The predictive power of environmental factors such as CO2 levels and Brent oil price was evaluated on the stock market (S&P 500) and foreign exchange (FX) rates for market currency pairs (EUR/USD) and emerging market currency pairs (MXN/USD) with recurrent neural networks (RNN). The S&P 500 is a collection of well-established companies in the United States representing the macroeconomic health of the nation and the global economy. The FX market is a global decentralized over-the-counter (OTC) market used to determine the spot price of currency pairs. A highly leveraged market usually trades within a specific price range. Since stocks and FX pairs are highly correlated to macroeconomic factors, it was hypothesized that environmental factors such as CO2 levels and oil prices also have predictive power due to their close causal relationship with anthropological economic activities. To verify the predictive power, an RNN model was built, and a bi-directional neural network with an internal state was used to process data sequences. The performance of RNN was quantified by measuring the residual prediction from the true value. Although the study at its current state might need further statistical rigor, it concluded that environmental factors increased the predictive power for the S&P 500 while decreasing it for the DM FX pair and the EM FX pair showed mixed results.
Read full abstract