CEA - Certified Economic Analyst Economic Forecasting Techniques Questions and Answers — Questions and Answers
Question 1: An organization wants to forecast the long-term impact of a novel, disruptive technology on a specific industry. Since the technology is new, no historical data is available. Which forecasting technique would be most appropriate in this situation?
- Holt-Winters Exponential Smoothing
- Box-Jenkins (ARIMA) model
- The Delphi Method (Correct answer)
- Multiple Regression Analysis
Correct answer: The Delphi Method
The Delphi Method is a qualitative forecasting technique that is ideal for long-term forecasting and situations with high uncertainty or a lack of historical data. It works by gathering, refining, and achieving a consensus from a panel of anonymous experts through multiple rounds of questionnaires. The other options are all quantitative methods that require historical data to build a model.
Question 2: An economic analyst is examining 20 years of quarterly GDP data. They identify a general long-term upward movement in the data, consistent peaks during the fourth quarter of every year, and a multi-year period of slower growth corresponding to a past recession. These three patterns, in the order they are described, best represent which components of a time series?
- Seasonal, Trend, and Cyclical
- Trend, Cyclical, and Seasonal
- Cyclical, Trend, and Seasonal
- Trend, Seasonal, and Cyclical (Correct answer)
Correct answer: Trend, Seasonal, and Cyclical
The general long-term upward movement over two decades is the trend. The consistent peaks in the fourth quarter of every year represent a fixed, repeating pattern, which is the seasonal component. The multi-year period of slower growth that corresponds to a business cycle (recession) is the cyclical component.
Question 3: What is the primary distinction between a causal (econometric) forecasting model and a pure time-series forecasting model?
- Causal models attempt to explain the functional relationships between variables, while time-series models use a variable's own past values to predict its future. (Correct answer)
- Causal models are only suitable for short-term forecasts, whereas time-series models are better for long-term forecasts.
- Time-series models are generally more complex and require more data inputs than causal models.
- Time-series models require the data to be stationary, while causal models can be used with non-stationary data without any transformation.
Correct answer: Causal models attempt to explain the functional relationships between variables, while time-series models use a variable's own past values to predict its future.
The core difference lies in their underlying assumptions and inputs. Causal models, such as regression, are built on the premise that the variable to be forecast (the dependent variable) is determined by the values of other variables (independent variables). In contrast, pure time-series models, like exponential smoothing or ARIMA, assume that future values of a series can be predicted from its own historical patterns (e.g., trend, seasonality) without explicitly identifying external causal factors.
Question 4: An analyst is evaluating two forecasting models for a company's monthly sales. Model A has a Mean Absolute Error (MAE) of $5,000 and a Mean Squared Error (MSE) of 40,000,000. Model B has an MAE of $5,500 and an MSE of 33,000,000. Which of the following is the most reasonable interpretation of these results?
- Model A is definitively superior because its MAE is lower.
- The higher MSE for Model A relative to its MAE suggests it is prone to producing occasional, very large forecast errors. (Correct answer)
- Both models are equally unreliable, and the choice between them is arbitrary.
- The MAE is always a more reliable metric than the MSE, so Model A should be chosen.
Correct answer: The higher MSE for Model A relative to its MAE suggests it is prone to producing occasional, very large forecast errors.
The Mean Squared Error (MSE) penalizes larger errors more heavily than smaller ones because it squares each error before averaging. While Model A has a slightly better average error (lower MAE), its significantly higher MSE suggests that it generates larger, more impactful outlier errors compared to Model B. These large errors are squared, inflating the MSE. Model B, despite a slightly higher average error, is more consistent and avoids these costly, large mistakes.
Question 5: In the context of the Box-Jenkins (ARIMA) methodology, the autoregressive (AR) component specifies that the current value of the time series is a linear function of which of the following?
- A set of independent explanatory variables.
- Past values of the forecast errors.
- The differenced values of the series used to achieve stationarity.
- Previous values of the time series itself. (Correct answer)
Correct answer: Previous values of the time series itself.
An autoregressive (AR) model, denoted as AR(p), specifies that the current value of the time series depends linearly on its own 'p' previous (or lagged) values. The "auto" prefix refers to this self-regression. Past forecast errors are related to the Moving Average (MA) component, independent variables are used in causal models, and differencing is the "Integrated" (I) part of an ARIMA model.
Question 6: An economic analyst is tasked with forecasting a turning point in the business cycle, specifically the beginning of a recovery after a recession. Which of the following economic indicators would be most useful for this purpose?
- Average duration of unemployment
- New orders for consumer goods and materials (Correct answer)
- Industrial Production Index
- Change in the Consumer Price Index (CPI)
Correct answer: New orders for consumer goods and materials
Leading indicators are economic variables that tend to change before the overall economy changes, making them useful for predicting turning points. New orders for consumer goods and materials are a classic leading indicator because companies increase orders in anticipation of future economic activity and consumer demand. The Industrial Production Index is a coincident indicator (moves with the economy), while the average duration of unemployment and changes in CPI are generally considered lagging indicators (change after the economy has already turned).
CEA - Certified Economic Analyst Economic Forecasting Techniques Questions and Answers
An organization wants to forecast the long-term impact of a novel, disruptive technology on a specific industry.
Since the technology is new, no historical data is available.
Which forecasting technique would be most appropriate in this situation?