Forecasting (5): Dynamic versus static forecast

Summary

This video explains the differences between static and dynamic forecasting approaches. A static forecast uses real historical data to predict one period ahead, recalculating with every new data point. Dynamic forecasting, however, incorporates previously forecasted values as input for future predictions, making it suitable for long-term forecasting. Static methods are best for short-term predictions, while dynamic methods are used for long-term strategic planning. The video also discusses how each method works and their respective advantages in different forecasting contexts, such as for operational versus strategic planning.

Takeaways

• 😀 Static forecasting models use real data to forecast one period ahead at a time, updating the forecast with each new data point.
• 😀 Dynamic forecasting models use previously forecasted values to forecast future points, which introduces the possibility of compounding errors over time.
• 😀 In a static forecast approach, the model always uses the most recent available data, while dynamic forecasting uses both real and forecasted data as new points are added.
• 😀 Static forecasting is more suitable for short-term predictions, like daily or weekly forecasts, and operational planning.
• 😀 Dynamic forecasting is more applicable for long-term planning, such as strategic forecasts for 20 to 30 years ahead.
• 😀 Long-term forecasts are less accurate due to the unpredictable impact of factors like technological changes or global events (e.g., the COVID-19 pandemic).
• 😀 When making short-term forecasts, static models tend to perform better, as they rely on real data and are more adaptable to recent changes.
• 😀 Dynamic forecasting involves forecasting multiple steps ahead at once, but it still forecasts one step at a time using previously forecasted data.
• 😀 The main difference between static and dynamic forecasts is how they treat the use of forecasted versus real data for predictions.
• 😀 Static forecast models are typically more reliable for out-of-sample periods and shorter time frames, while dynamic forecasts are used for longer time frames but are more prone to error.

Q & A

What is the main difference between static and dynamic forecasting models?

The main difference is that static forecasting models always use real data to forecast the next period, while dynamic forecasting models use previously forecasted values as input to forecast future periods.

In a static forecasting model, how are forecasts updated?

In a static forecasting model, forecasts are updated each time new data becomes available. The model recalculates the forecast for the next period using all the real data up to that point.

How does dynamic forecasting work when predicting multiple periods ahead?

In dynamic forecasting, the model forecasts one period at a time, using previously forecasted values as inputs for subsequent predictions, even though it starts with real data for the first forecast.

What type of data is used in static forecasting versus dynamic forecasting?

Static forecasting uses only real, historical data to make predictions, while dynamic forecasting uses both real data and forecasted values for future predictions.

What is an example of when to use a static forecast?

A static forecast would be used for short-term forecasts, such as predicting daily or weekly data, where the goal is to make predictions using the most recent available data.

What is an example of when to use a dynamic forecast?

A dynamic forecast is ideal for long-term planning, such as predicting the next 5-10 years, where using forecasted values for future periods is necessary due to the long time horizon.

What happens if a dynamic forecasting model is used for short-term forecasts?

Using a dynamic forecasting model for short-term forecasts might not yield accurate results, as it relies on forecasted values, which can introduce error over short time periods.

How do forecasted values in dynamic forecasting impact the results?

Forecasted values in dynamic forecasting introduce cumulative error, as each forecast depends on the accuracy of the previous forecast. This can make long-term forecasts less reliable.

What is a common application for static forecasting models?

Static forecasting models are commonly used for operational planning and short-term forecasting, such as daily sales predictions or weekly inventory planning.

Why do dynamic forecast models perform poorly for long-term predictions?

Dynamic forecast models perform poorly for long-term predictions because the forecast errors accumulate as forecasted values are used in subsequent predictions, and external factors like economic changes can significantly impact long-term accuracy.