Dynamic Allocation or Diversification: A Regime-Based Approach to Multiple Assets

Feature Selection in Jump Models

“…Both seem to be in good agreement with the average annualized volatility, although we have no way to measure which state sequence is more accurate as the true states are unknown. With a more inhomogeneous data set, for example returns for different asset classes (Nystrup, Hansen et al, 2017), the resulting state sequences could be quite different.…”

Greedy Online Classification of Persistent Market States Using Realized Intraday Volatility Features

“…The HMM detects changes equally quickly as the jump model with the realized volatility features; however, this comes at the cost of many spurious and short-lived state changes. If the estimated states were used as the basis of a regime-based asset allocation strategy, then the lower persistence would lead to a substantially higher turnover, which could easily make the difference between whether the strategy is profitable or not (Nystrup et al, 2015a(Nystrup et al, , 2017a(Nystrup et al, , 2018. The results from this application suggest that using data sampled at higher frequencies can help improve the classification of persistent states in financial time series.…”

A Mixed Bond and Equity Fund Model for the Valuation of Variable Annuities

“…Our motivation is twofold. First, bond fund returns are usually modeled in the same way as equity returns in the econometric literature (e.g., Guidolin and Timmermann, 2006;Nystrup et al, 2017) and there is therefore a need for more specialized models that reflect the intrinsic relationship between fixed income fund returns and term structure dynamics. In other words, treating a bond fund return as a risky asset return that is correlated with other risky asset returns misses a key feature of this asset class.…”