Dynamic Linear Models for Forecasting Equity Risk Premia
Summary
The document asks whether continuous-state space or dynamic linear models can forecast security-level equity risk premia over medium-term horizons. It contrasts these models with discrete-state Markov and hidden Markov models, suggesting that continuous latent states may represent evolving market conditions and cross-sectional dependence more flexibly. It also raises the possibility that long-memory dynamics could better reflect recurring, nonidentical patterns.
One response points to research applying a dynamic linear model to equity risk-premium forecasts. Another argues that performance depends heavily on feature quality: richer models may outperform simple regression when the inputs contain useful information, while no model can recover patterns absent from the data. The discussion recommends considering Bayesian dynamic models more broadly. It offers no empirical results or detailed model specification, and leaves open whether particle filtering is suitable for the proposed horizons and security-by-security forecasts.
Key ideas
- Continuous-state dynamic models are proposed as a way to represent changing conditions in equity risk-premium forecasts.
- The question emphasizes security-level forecasting over medium-term horizons and possible cross-sectional dependence.
- A cited study applies a dynamic linear model to equity risk-premium prediction.
- Model complexity cannot compensate for features that contain little predictive information.
- The discussion does not provide comparative results or settle the appropriate forecast horizon.
Tags
Full text
# Is there any research on applying state-space or dynamic linear models to forecasting equity risk premia? # Is there any research on applying state-space or dynamic linear models to forecasting equity risk premia? Is there any research on applying state-space or dynamic linear models to forecasting equity risk premia on a security-by-security basis with a medium term horizon (say 3 month to 12 months horizon)? Markov Models are a special case of state-space models where the states are discrete. There's an avalanche of research on these MM's and HMM's (beginning with Ryden 1998) but it seems to me there are some advantages to taking an approach where the state is a continuous variable. I have seen papers by Johannes, Lopes, and Carvalho that focus on shorter-horizons and demonstrate the superiority of PL over MCMC methods. Seems to me that this approach could capture the time-series dependence and cross-sectional dependencies in a way that traditional panel models cannot. Also, it seems that since history rhymes but does not repeat, a long-memory state-space model would be better than an HMM. Before I go down this windy road I'm curious if anyone out there has already attempted this or if there is a weakness with this approach. For example, perhaps state-space or particle filtering models only work best when the forecast horizon is very short. ## Answer by Ram Ahluwalia (score 3, accepted) https://quant.stackexchange.com/a/2085 Here's a paper by Dangl, Halling, and Randl (2006) that uses a dynamic linear model to forecast the equity risk premium. ## Answer by William (score 4) https://quant.stackexchange.com/a/1959 I think, as with many machine learning approaches to investing decision support, it depends largely on the data. With a good selection of features, yes dynamic models like you're talking about will probably do better than a simple linear regression; but then again, with a good selection of features, linear regression will probably work reasonably well, too. On the other hand, with a poor selection of features, you can use any statistical model in the world and it won't work -- simply put, you can't learn from patterns that aren't there. That aside, you could probably also make gains by generalizing your search a bit from the models you describe to all Bayesian dynamic models.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.