Predictability v/s Accuracy: An empirical evaluation of the Chilean Exchange Rate with the Chilean Survey of Professional Forecasters

Abstract

We evaluate the predictability and efficiency of the Survey of Professional Forecasters (SPF) of the Chilean exchange rate relative to the US dollar (CLP). We evaluate the forecast performance through four approaches: (1) a comparison with a naïve benchmark under using a quadratic loss function, (2) Mean Directional Accuracy relative to a pure luck benchmark, (3) the correlation between SPF and the target variable, and (4) a range analysis of the respondents. Our analysis spans from April 2012 to April 2024, with a specific focus on the period from June 2018 to April 2024. Out-of-sample results reveals that SPF is consistently outperformed by the Driftless Random Walk (DRW) in terms of Mean Squared Prediction Error at several forecasting horizons. Nevertheless, SPF correlations are strong and statistically significant at short, mid, and long horizons. Therefore, CLP is predictable but inaccurate, with the bias playing an important role. To address this, we propose a bias adjusted forecast. This adjusted forecast outperforms the more competitive benchmark, the DRW+, at mid and long horizons.