FORECASTING METHODS

FORECASTING METHODS

The Methodology Tree for Forecasting classifies all possible types of forecasting methods into categories and shows how they relate to one another. Dotted lines represent possible relationships.





Knowledge source: When reliable objective data are available, they should be used. Still, one might benefit from also using subjective methods.


Judgmental: Available data are inadequate for quantitative analysis or qualitative information is likely to increase accuracy, relevance, or acceptability of forecasts.

Statistical: Relevant numerical data are available.

Others: Knowledge exists about the expected behaviour of other people or organizations.

Self: People have valid intentions or expectations about their behavior. Both are most useful when (1) responses can be obtained from a representative sample, (2) responses are based on good knowledge, (3) there are no reasons to lie, (4) new information is unlikely to change the behavior. Intentions are more limited than expectations in that they are most useful when (5) the event is important, (6) the behavior is planned, and (7) the respondent can fulfill the plan (so, for example, the behavior is not dependent on the agreement of other people.

Unstructured: The information is used in an informal manner.

Structured: Formal methods are used to analyze the information. This means that the rules for analysis are written in advance and they are rigorously adhered to. Records should be kept of how the procedures were administered.

Unaided judgment: Experts think about a situation and predict how people will behave. They might have access to data and advice, but their forecasts are not aided by formal forecasting methods. This is the most commonly used method. It is fast, inexpensive when only a few forecasts are needed, and can be used in cases where small changes are expected. It is most likely to be useful when the forecaster gets good feedback about the accuracy of his forecasts (e.g., weather forecasting, betting on sports, and bidding in bridge games.)

Expert Forecasting refers to forecasts obtained in a structured way from two or more experts. The most appropriate method depends on the conditions (e.g., time constraints, dispersal of knowledge, access to experts, expert motivation, need for confidentiality). The nominal group technique (NGT) can be conducted as a simple one-round survey for situations in which experts possess similar information [1]. If experts are expected to possess different information, discussion among participants in an "estimate-talk-estimate" is appropriate. Where group pressures are a concern, Delphi or prediction markets may be best [2]. The Delphi technique involves at least two rounds with results of the previous round summarized for participants (Delphi Software). Prediction markets are incentive-based arrangements that use markets to aggregate, in the form of prices, information that is dispersed among participants. For all methods, diverse experts should be recruited, questions should be chosen carefully and tested, and procedures for combining across experts (e.g., the use of medians) should be specified in advance.

Structured analogies: An expert lists analogies to a target, describes similarities and differences, rates similarity, and matches each analogy's decision (or outcome) with a potential target situation decision (or outcome). The outcome implied by the top-rated analogy is used as a forecast..

Game theory: An attempt to explain, model, and predict behaviour in the social world. To do these things, game theorists seek to identify the rules of the situation including the utility to each party of possible outcomes. While game theory can provide ex post analysis that appears insightful, there is no evidence that the method can provide useful forecasts.

Decomposition: The problem is addressed in parts. The parts may either be multiplicative (e.g., to forecast a brand's sales, one could estimate total market sales and market share) or additive (estimates could be made for each type of item when forecasting new product sales for a division).

Judgmental bootstrapping: Derive a model from knowledge of experts’ forecasts and the factors they used to make their forecasts using regression analysis change in the historical data (such as where trying to estimate a price elasticity using time series data with little variation in price).

Expert systems: Rules for forecasting are derived from the reasoning experts use when making forecasts. Obtain knowledge from diverse sources such as surveys, interviews, protocol analysis, and research papers.

Role: People's roles influence their behaviour and there is knowledge about these roles.
Role playing/Simulated interaction: In role playing, people are expected to think in ways consistent with the role and situation described to them. If this involves interacting with people with different roles for the purpose of predicting the behavior of actual protagonists, we call it simulated interaction. That is, people act out prospective interactions in a realistic manner. The role-players' decisions are used as forecasts of the actual decision.
No role: Roles are not expected to influence behaviour, or knowledge about the roles is lacking, or there are many actors with different roles.
Intentions/expectations: Survey people about their intentions or expectations regarding their future behaviour or those of their organization. Analyze the survey data to derive forecasts.
Conjoint analysis: Elicit preferences from consumers (or other actors) for various offerings (e.g. for alternative computer designs or for different political platforms) by using combinations of features (e.g. power and weight for a laptop computer.) Regression-like analyses are then used to predict the most desirable design
Univariate: Historical data are available on the behaviour that is to be predicted (e.g., data on automobile sales from 1940-2002).

Extrapolation: Use time-series data, or similar cross-sectional data, to predict. For example, exponential smoothing is used to extrapolate over time (see extrapolation.), diffusion models are used for innovations.
Quantitative analogies: Experts identify analogous situations for which time-series or cross-sectional data are available, and rate the similarity of each analogy to the data-poor target situation. These inputs are used to derive a forecast; for example, to forecast demand for cinema seats in a new suburb, average data from cinemas in suburbs identified by experts as similar to the target could be used.
Rule-based forecasting: Expert domain knowledge and statistical techniques are combined using an expert system to extrapolate time series. Most series features are identified by automated analysis, but experts identify some factors. In particular they identify the causal forces acting on trends.
Neural network: Definition to be added.
Multivariate: Data are available on variables that might affect the behaviour of interest.
Data-based: Experience and prior research are not available and so one must try to infer relationships from the data
Theory-based: Experience and prior research provide useful information about relationships relevant to the forecast.
Data mining: Letting the data speak for themselves. In general, theory is not considered. Despite its widespread use and many claims of accuracy, we have been unable to find evidence that data mining provides forecasts that are more accurate than those from alternative methods
Linear: The problem can be modeled as linear in the parameters. For example, in economics, the common assumption of constant elasticities for key variables such as price can be modeled by taking logs of the dependent and independent variables.
Causal models: Theory, prior research and expert domain knowledge are used to specify relationships between a variable to be forecast and explanatory variables. In the case of econometric methods, regression analysis is commonly used to estimate model coefficients such that they are consistent with prior knowledge. System dynamics models relationships using stocks and flows, often with an emphasis on feedback loops. Causal models aided by the use of econometrics have been found to improve accuracy. The use of system dynamics has not.
Classification: If the problem is composed of groups that act in different ways in response to a change, one can study each group separately, then add across segments. For example, in the airline industry, price has different effects on the business and pleasure markets.
Segmentation: When segments are independent, a tree structure is appropriate. When information is available on relationships between segments, input-output analysis, system dynamics, and cluster analysis can be used. Of the dependent segmentation techniques, only input-output analysis has been found to improve accuracy.