Model Risk in Bond Portfolio Hedging

Empirically testing a bond portfolio hedging model is usually carried out when proposing a new model or to compare several existing models using real data. However, there are many methodological choices to be made during such exercise, which are usually made either implicitly or without sufficient discussion. We review the empirical literature and highlight differences in testing methodology. We then carry out a massive numerical experiment to assess how each of these differences influences the testing outcome thus quantifying the model risk associated with these model features. The model risk measure values we report are easily interpretable and offer insight on the discrepancies in the results of existing empirical studies and ways to avoid such discrepancies in future work