How Inflation and 10-year Bond Yield Forecasting with Computer Science (AI) Could Support Pension Funds

Inflation and bond yields forecasting helps to regularly communicate with pension fund membership about future benefits compared to the future costs of living. Inflation as a Liability Risk has a strong impact on liabilities and benefit levels of active membership and pensioners as well as on interest rates. Additionally, the direction of interest rates has a big influence on the pricing of assets (esp. bonds, equities, and real estate). It means inflation has an impact on the pension fund’s Total Risk. In Switzerland the inflation level after the finance crisis 2008 had been stable (on average over the period 2009-2021: the inflation rate was zero, 0%). Therefore, its current development has a strong impact on the "real" benefits of pension funds for active members and for pensions in payment. It would be very useful to investigate how active members and pensioners can be treated equally in terms of inflation and to make forecasts for these potential costs based on the prediction of inflation and 10-year government bond yields in advance. Scientific publications showed that forecasts for inflation, exchange rates, spot interest rates and other yields using Artificial Neural Network methods, provide very good predictions. Our forecasting results are based on Neural Network Autoregressive (NNAR) forecasting model (introduced by Prof. Hyndman and Prof. Athanasopoulos in 2018). The NNAR model is best suited for financial forecasting compared to the Autoregressive Integrated Moving Average (ARIMA) model.