The Informational Value of Technical Indicators in Forecasting Quoted Prices of a Government Bond: A Machine Learning Approach

In the context of increasing electronification and transparency in the secondary market for euro area government bonds, this study examines the informational value of technical indicators in predicting intraday quoted prices of the 10-year German government bond. Using a feedforward neural network based on 13 input datasets and two output variables, the paper aims to identify the technical indicators that contribute most to the model’s predictive accuracy and efficiency under less-volatile market conditions. The results show that trend indicators, such as moving average convergence/divergence (MACD) and weighted moving average (WMA), and relative strength index (RSI), as the only momentum indicator among the remaining observed ones, allow the models to achieve lower forecast errors than those obtained with the input variable set that does not include technical indicators. On the other hand, Bollinger Bands (BB), as a volatility indicator, result in the worst predictive performance. The findings of the study extend the use of technical analysis beyond stock markets, where these indicators were initially developed, and enable the development of more efficient forecasting models in the government bond market.