Quantitative Forecasting Methods in Economics

A stationary process is characterized by constant mean, variance, and autocorrelation over time, making it a fundamental concept in time-series analysis. This stability allows for reliable modeling and forecasting, as the statistical properties do not change, enabling analysts to predict future values based on historical data without adjustments for trends or seasonality.

In an ARIMA(p,d,q) model, the 'd' parameter signifies the degree of differencing applied to the time series data. This process helps to stabilize the mean of the series by removing trends or seasonality, allowing for a more accurate modeling of the underlying patterns in the data.

Mean Absolute Percentage Error (MAPE) is scale-independent because it expresses forecast accuracy as a percentage, allowing for direct comparison across different datasets with varying units and scales. This characteristic makes MAPE particularly useful in evaluating the performance of forecasting models, as it standardizes errors relative to the actual values, facilitating meaningful comparisons.

Exponential smoothing is a forecasting technique that prioritizes recent data by applying exponentially decreasing weights to past observations. This means that the most recent data points have a greater influence on the forecast, making it responsive to changes and trends in the data, which is particularly useful in dynamic environments.

Multicollinearity occurs when independent variables in a regression model are highly correlated, leading to difficulties in estimating the individual effect of each predictor. This results in unstable coefficient estimates, as small changes in the data can lead to large variations in the estimated coefficients, undermining the reliability of the model's predictions.

The Augmented Dickey-Fuller test is a statistical test used to check for the presence of a unit root in a time series. A unit root indicates that the time series is non-stationary, meaning its statistical properties change over time. Identifying a unit root helps in understanding the underlying behavior and trends of the data.

R-squared measures the proportion of variance in the dependent variable that can be explained by the independent variables in a model. It provides insight into the goodness of fit, indicating how well the model captures the underlying data patterns. A higher R-squared value suggests a better explanatory power of the model.

In vector autoregression (VAR) models, each variable's current value is influenced not only by its past values (own lags) but also by the past values of other variables in the system. This allows for capturing the dynamic interrelationships among multiple time series, making "lagged" the appropriate term to describe these past values.

A forecast with negative bias consistently underpredicts the actual values, not overpredicts. Negative bias indicates a tendency to underestimate outcomes, leading to forecasts that fall short of the true values. Therefore, the statement is false, as it mischaracterizes the nature of negative bias in forecasting.

Classical decomposition is a forecasting method that separates a time series into three distinct components: the trend, which shows long-term movement; the seasonal component, reflecting periodic fluctuations; and the irregular component, capturing random variations. This approach allows for better understanding and prediction of future values by analyzing these individual elements.

The Granger causality test assesses the predictive relationship between two time series. It determines if past values of one variable contain information that can improve the prediction of another variable, implying a directional influence. This statistical method is widely used in economics and finance to understand dynamic relationships between variables.

The Diebold-Mariano test evaluates the accuracy of two forecasts by analyzing their forecast errors, which are the differences between predicted and actual values. This statistical test determines whether one forecast consistently outperforms the other, providing insights into which model may be more reliable for future predictions.