Forecast Combination

What is the main motivation for combining forecasts?

• To improve predictive accuracy when models are likely mis-specified

• To increase computational speed

• To avoid using time series models

• To avoid using economic theory altogether

Which early paper is considered seminal in the development of forecast combination theory?

• Stock and Watson (2004)

• Diebold and Mariano (1995)

• Bates and Granger (1969)

• Box and Jenkins (1970)

In a simple average combination method, what is the main advantage?

• It eliminates all bias

• It avoids estimation error since weights are imposed rather than estimated

• It always outperforms all other methods

• It uses estimated optimal weights

The “forecast combination puzzle” refers to which finding?

• Simple averages often outperform more complex methods

• Sophisticated methods always outperform simple averages

• Complex models are always better than naive ones

• Weighted averages are useless in practice

Which combination method addresses outlier forecasts by excluding extremes?

• Simple average

• Trimmed mean

• Inverse MSE

• OLS regression

In the Inverse MSE method, how are weights determined?

• By expert judgment

• Equally for all models

• Proportional to the inverse of the model’s MSE

• By the rank of MSE values

Which method uses weights proportional to the inverse of the forecast models’ rank?

• Simple average

• Trimmed mean

• Inverse Rank

• OLS regression

Which method estimates weights by regressing actual outcomes on individual forecasts?

• Simple average

• OLS regression

• Bayesian Model Averaging

• Inverse MSE

In combining density forecasts, which constraint ensures the result is a valid probability density?

• Weights can be negative as long as they sum to 1

• Weights must be positive and sum to 1

• Weights must all be equal

• Weights must minimize MSE

The “linear opinion pool” refers to:

• Combining densities with equal weights

• Combining densities using inverse MSE

• Bayesian averaging over densities

• Using regression for density forecasts