Judgmental Forecasting Principles and Methods
What is Judgmental Forecasting?
Definition and Necessity: Judgmental forecasting is an approach used when quantitative or statistical data is unavailable or insufficient. It relies on the intuition, expertise, and subjective judgment of individuals or groups.
Specific Scenarios for Use:
Complete lack of historical data: Often occurs during the initial stages of a project or organization.
New Product Launch: When there is no sales history for a specific item.
Entry of a New Competitor: When market dynamics shift suddenly due to external actors.
Unique Market Conditions: During unprecedented events or crises where previous trends no longer apply.
Incomplete or Delayed Data: When information exists but is either missing key components or is not provided in a timely enough manner for statistical modeling.
Usage Settings for Judgmental Forecasting
Type 1: Only Feasible Approach: There are no available data, rendering statistical methods inapplicable. In these cases, judgmental forecasting is the sole option.
Type 2: Judgmental Adjustments: Statistical data are available and used to generate a baseline forecast, which is then refined or adjusted based on human judgment.
Type 3: Independent Combination: Statistical and judgmental forecasts are developed independently of one another and then combined to form a final projection.
Accuracy and Subjective Limitations
Improving Accuracy:
Accuracy increases when the forecaster possesses significant domain knowledge and access to up-to-date, timely information.
Judgmental methods are particularly adept at adjusting quickly to sudden changes or specific events.
The quality of results is enhanced by implementing well-structured and systematic approaches.
Subjective Limitations:
Inconsistency: Unlike mathematical formulas, human judgment can vary over time even when presented with the same data.
Cognitive Limits: Predictions are limited by human cognitive processing and memory capacity.
Bias and Agendas: Personal or political motivations can cloud judgment. Examples include wishful thinking or excessive optimism regarding project outcomes.
Anchoring Effect: This refers to the tendency to over-rely on an initial piece of information (the "anchor"). This can lead to conservatism and the undervaluing of more current information, creating a systematic bias.
Key Principles for Effective Judgmental Forecasting
Clear Task Specification: The forecasting task must be set concisely and clearly. Definitions should be explicit, complete, and devoid of ambiguity. Emotional language and extraneous materials should be avoided to prevent distracting the forecaster. A preliminary round of information collection may be beneficial before forecasting begins.
Implementation of a Systematic Approach: Using checklists and structured criteria improves both accuracy and consistency. This involves identifying key variables—such as price, competition quality, economic climate, or target population—and determining how to weight them.
Documentation and Justification: Formalizing and recording the criteria and assumptions used allows for the process to be repeated and audited. Requiring forecasters to defend their projections in writing increases personal responsibility and reduces individual bias.
Systematic Evaluation: Forecasts must be monitored against actual outcomes as they become available. This feedback loop allows forecasters to assess their decision rules and adjust for environment changes.
Segregation of Forecasters and Users: To maintain objectivity, those who generate the forecast should be independent of those who use the forecast to create plans or meet financial goals. This prevents "reality checks" from being overridden by management’s desire to meet specific targets.
Case Study and Discussion: National Health Insurance Scheme (NHIS) Ghana
Financial Scope: The NHIS budget is approximately , which represents roughly of the national GDP.
Drug Subsidy Structure:
NHIS Card Holders (Concession): Pay a copayment of per drug until they reach the safety net of . After this threshold, medicines are free for the remainder of the year.
General Patients: Pay a copayment of per drug until the safety net of is reached. Subsequently, they pay a reduced rate of per drug for the rest of the financial year.
Forecasting Methodology: The scheme utilizes groups of drugs across the four categories to anticipate volume and expenditure, resulting in a total of time series.
Forecasting New Listings: Companies requesting to add drugs must provide estimates for patient numbers, market share, and substitution impacts. The National Health Insurance Board mandates a highly organized, documented, and methodical approach to reduce self-serving bias. Estimates are evaluated by a subcommittee twice: before and after the drug is added. Outcomes are formally compared to projections at the and month marks.
Policy Impact Forecasts: Currently unorganized and often reliant on a single individual without formal scrutiny or documentation. Recommended changes include standardizing the process, requiring documentation of assumptions, and peer review.
Types of Judgmental Forecasting
Delphi Method: A structured, iterative group process aimed at reaching a consensus among experts.
Executive Judgment (Jury of Executive Opinion): A small group of high-level managers combines experience with models. It is fast but prone to group-think.
Sales Force Composite: Individual salespeople project their expected sales, which are then aggregated. While salespeople are close to customers, they are often overly optimistic.
Consumer Market Survey: Directly asking customers about their intentions. This can be difficult to answer, and actual behavior often differs from stated intent.
Forecast by Analogy: Using past experiences from similar situations to predict the future.
Scenario-based Forecasting: Generating forecasts based on various plausible future states of the world.
The Delphi Method: Detailed Process and Roles
Origin: Invented by Olaf Helmer and Norman Dalkey of the Rand Corporation in the s for military applications.
Core Assumption: Group forecasts are generally more accurate than individual ones.
The Facilitator's Role: The facilitator is crucial for managing the process, maintaining expert anonymity, providing feedback, and aggregating the final projections.
Process Stages:
Assemble a panel of experts.
Distribute tasks and challenges.
Experts return initial forecasts and justifications.
The facilitator compiles and summarizes these for feedback.
Experts review their forecasts based on the feedback (this may repeat iterate).
Final forecasts are constructed by aggregating expert data (usually weighted equally).
Challenges and Limitations:
Identification of Experts: Finding the right contributors is the first hurdle.
Feedback Control: Facilitators must provide summary statistics and qualitative justifications without biasing the movement toward consensus.
Iteration and Attrition: The process takes time. If it lasts too long, the panel may lose interest.
Extreme Values: Facilitators must be wary of outliers that may skew the aggregate.
Forecast by Analogy and Structured Analogy
Concept: Predicting the outcome of a situation by looking at similar historical "analogues."
Curriculum Design Example: Group members estimated a task would take to months. An expert looked at actual data for similar projects where of groups never finished and the rest took to years. The actual project took years.
Structured Analogy (Green \& Armstrong, 2007):
A panel focuses on identifying as many comparisons as possible.
Experts describe similarities and differences between the parallel case and the target scenario, scoring likeness on a scale.
Anonymity helps prevent suppressing creativity, though Green and Armstrong found no specific benefit from expert collaboration/interaction.
Accuracy is highest in experts with more than two analogies and direct experience.
Scenario-Based Forecasting
Objective: Generate forecasts based on multiple plausible scenarios, even if each individual scenario has a low probability of occurring.
Development: Scenarios are built by considering all potential drivers, their relative impacts, and their interactions.
Participant Engagement: Decision-makers often participate in scenario generation. While this can introduce bias, it improves communication and the understanding of final results.
Guidelines for Judgmental Adjustments to Statistical Forecasts
Use Sparingly: Adjustments should not be used to fix patterns (like trends or seasonality) that the statistical model already handles. Statistical models are superior for accounting for data trends.
Conditions for Adjustment: Only adjust when there is significant extra information or strong evidence not included in the mathematical model.
Structured Approach for Adjustments: Require justifications for all changes to prevent unwarranted overwriting of statistical estimates.
Fatigue Management: When using a panel to adjust many items (e.g., product lists), start with the most significant markets. As meetings progress, the quality and frequency of adjustments typically decrease due to member exhaustion.