When the forecasting procedure is based exclusively on past values of the variable being forecasted, it is termed what?

The correct answer is indeed the time series method. This approach is based solely on the historical data of the variable being forecasted, utilizing patterns or trends inherent in that past data to predict future values.

Time series methods analyze time-ordered data points to uncover seasonal trends, cyclical movements, and other temporal patterns that help in predicting future outcomes. For example, if you were forecasting sales for a retail store, you would look at sales figures from previous years to identify trends, such as a typical increase in sales during the holiday season.

In contrast, exponential smoothing employs weighted averages of past observations, giving more importance to recent data points, but it still relies on historical data trends. The cross-sectional method looks at data across different subjects at a single point in time rather than over time, making it unsuitable for forecasting based on past values. The qualitative method, on the other hand, relies on subjective judgment and theories rather than historical data, which does not match the criteria of being based exclusively on past values.

Thus, time series methods are specifically tailored for forecasting from historical data, making them the most appropriate choice in this context.