CEA Economic Forecasting Techniques 4 — Questions and Answers
Question 1: The Hodrick-Prescott (HP) filter is commonly applied in macroeconomic forecasting to:
- Test for Granger causality between GDP and inflation
- Separate a time series into trend and cyclical components (Correct answer)
- Adjust data for seasonality using calendar effects
- Estimate dynamic factor models from large datasets
Correct answer: Separate a time series into trend and cyclical components
The HP filter minimizes deviations from a smooth trend, weighted by a smoothing parameter λ, to extract the business cycle component.
Question 2: Dynamic Factor Models (DFMs) are advantageous for macroeconomic forecasting primarily because they:
- Require only one or two input variables
- Extract a small number of common factors from a large dataset of economic indicators (Correct answer)
- Always outperform simple AR(1) models at all horizons
- Eliminate the need for stationarity transformations
Correct answer: Extract a small number of common factors from a large dataset of economic indicators
DFMs reduce hundreds of economic series to a handful of latent factors, efficiently summarizing co-movement and improving forecast accuracy.
Question 3: In a GARCH(1,1) model applied to economic forecasting, what does the model primarily capture?
- Long-run equilibrium relationships between prices
- Time-varying volatility clustering in financial or economic time series (Correct answer)
- Structural breaks in monetary policy regimes
- Seasonal adjustment of quarterly national accounts
Correct answer: Time-varying volatility clustering in financial or economic time series
GARCH models capture the empirical regularity that large shocks to an economic series tend to be followed by further large shocks (volatility clustering).
Question 4: Which evaluation metric is most appropriate when forecast errors are heteroskedastic and large outliers should not be penalized excessively?
- Root Mean Squared Error (RMSE)
- Mean Absolute Error (MAE) (Correct answer)
- Mean Absolute Percentage Error (MAPE)
- Theil's U statistic
Correct answer: Mean Absolute Error (MAE)
MAE weights all errors equally regardless of size, making it more robust to outliers than RMSE, which squares errors and amplifies large deviations.
Question 5: An economic forecaster notices that residuals from a regression model are positively autocorrelated. The most appropriate corrective action is to:
- Use heteroskedasticity-consistent (White) standard errors only
- Add lagged dependent or independent variables to capture omitted dynamics (Correct answer)
- Remove the trend from the dependent variable only
- Switch from OLS to GLS with a fixed autocorrelation structure
Correct answer: Add lagged dependent or independent variables to capture omitted dynamics
Positive serial correlation in residuals typically signals missing lag structure; adding lags removes the autocorrelation and improves forecast accuracy.
Question 6: Scenario analysis in economic forecasting differs from sensitivity analysis in that scenario analysis:
- Varies one parameter at a time while holding others fixed
- Defines internally consistent sets of assumptions across multiple variables simultaneously (Correct answer)
- Only applies to long-run structural models
- Requires Monte Carlo simulation techniques
Correct answer: Defines internally consistent sets of assumptions across multiple variables simultaneously
Scenario analysis creates coherent narratives (e.g., a trade war scenario) by jointly changing multiple variables, unlike sensitivity analysis which varies one factor at a time.
Question 7: The concept of 'forecast encompassing' implies that:
- A forecast cannot be improved by combining it with any competing forecast (Correct answer)
- All forecasts must be equally weighted in a combination
- The forecast with the lowest RMSE automatically encompasses rivals
- Bayesian forecasts always encompass classical forecasts
Correct answer: A forecast cannot be improved by combining it with any competing forecast
A forecast encompasses a rival if adding the rival's forecast to it provides no additional predictive information, meaning all useful information is already captured.
CEA Economic Forecasting Techniques 4
The Hodrick-Prescott (HP) filter is commonly applied in macroeconomic forecasting to: