Elasticity, Demand Estimation and Forecasting

Introduction

Managers make decisions under uncertainty. Human resource managers forecast staffing requirements, training demand, compensation budgets, and the effects of economic conditions on recruitment and retention. These decisions are influenced by demand: demand for products affects an organisation’s labour requirements, while demand for labour is itself affected by wages, productivity, technology, and the prices of related inputs.

Elasticity provides a systematic way to measure how responsive one variable is to another. Demand estimation uses economic and statistical evidence to identify the relationship between demand and its determinants. Demand forecasting then uses that relationship, together with historical patterns and assumptions about the future, to support planning.

This chapter integrates economic theory with practical quantitative methods. It distinguishes between measuring responsiveness, estimating relationships, and predicting future values—three activities that are related but not identical.

Learning Objectives

By the end of this chapter, you should be able to:

1. Explain the meaning and managerial significance of elasticity.
2. Calculate and interpret price, income, and cross elasticities of demand.
3. Distinguish point elasticity from arc elasticity.
4. Formulate and interpret demand functions.
5. Explain how regression analysis can estimate demand relationships.
6. Apply time-series and qualitative forecasting methods.
7. Use scenario analysis to support planning under uncertainty.
8. Evaluate the limitations, ethical issues, and governance requirements of demand forecasting.

1. Elasticity as a Measure of Responsiveness

A percentage change is more informative than an absolute change when comparing variables expressed in different units. Elasticity measures the percentage change in one variable resulting from a one-percent change in another variable.

In general form:

Elasticity is unit-free. This makes it useful for comparing responsiveness across products, markets, departments, or time periods.

1.1 Price Elasticity of Demand

Price elasticity of demand measures the responsiveness of quantity demanded to a change in the product’s own price. It is commonly denoted by E_d.

where Q_d is quantity demanded and P is price. For a conventional downward-sloping demand curve, the elasticity is negative because price and quantity demanded move in opposite directions. In discussion, economists often refer to the absolute value, |E_d|, when classifying responsiveness.

Price elasticity is relevant to pricing, revenue, capacity planning, recruitment services, and workforce budgets. If demand for a firm’s output is highly elastic, a price increase may cause a substantial loss of sales and therefore reduce derived demand for labour.

1.2 Point Elasticity

Point elasticity measures responsiveness at a specific point on a demand curve. If quantity is a differentiable function of price, the formula is:

The derivative dQ_d ÷ dP gives the local change in quantity associated with a small change in price. Point elasticity is appropriate for very small changes or when a continuous demand function is available.

For the linear demand function Q_d = a − bP, where a > 0 and b > 0,

Therefore:

A linear demand curve has a changing elasticity: it is generally more elastic at high prices and low quantities, and less elastic at low prices and high quantities.

1.3 Arc Elasticity

Arc elasticity measures responsiveness over a finite interval. It avoids obtaining different answers depending on whether the initial or final value is used as the base.

Here, Q₁ and Q₂ are the initial and final quantities, while P₁ and P₂ are the initial and final prices. The midpoint method is especially appropriate when the price change is not small.

Worked Example 1: Arc Price Elasticity

A training provider reduces its course fee from $500 to $450. Enrolment rises from 200 to 230 participants.

Percentage change in quantity:

Percentage change in price:

Thus:

The absolute elasticity is approximately 1.32. Demand is elastic: enrolment responds more than proportionately to the fee change.

1.4 Elasticity and Total Revenue

Total revenue is price multiplied by quantity sold:

When demand is elastic, a price reduction tends to increase total revenue because the percentage increase in quantity exceeds the percentage decrease in price. When demand is inelastic, a price reduction tends to reduce total revenue. This relationship assumes other relevant conditions remain sufficiently stable.

Revenue analysis should not be confused with profit analysis. Profit also depends on costs, capacity, taxes, and fixed commitments.

2. Income and Cross Elasticities

2.1 Income Elasticity of Demand

Income elasticity measures the responsiveness of demand to a change in consumer income.

where Y represents income or purchasing power.

Interpretation depends on the sign and magnitude:

• E_Y > 1: demand is income-elastic; the product is often treated as a luxury.
• 0 < E_Y < 1: demand is income-inelastic; the product is often treated as a normal necessity.
• E_Y < 0: demand falls as income rises; the product is inferior in the relevant income range.

These labels are contextual, not permanent characteristics. A product can be a necessity for one income group and a luxury for another.

For HR planning, income elasticity can affect demand for premium services, business travel, hospitality, and private education. Rising incomes may increase demand for high-skilled labour in some sectors, but automation or substitution can weaken that relationship in others.

2.2 Cross Elasticity of Demand

Cross elasticity measures the response of demand for one good to a change in the price of another good.

where Q_x is demand for product X and P_y is the price of product Y.

• E_xy > 0 usually indicates substitutes: an increase in the price of Y raises demand for X.
• E_xy < 0 usually indicates complements: an increase in the price of Y lowers demand for X.
• E_xy ≈ 0 suggests weak or no measurable relationship, subject to statistical uncertainty.

The strength of the relationship depends on the relevant market definition and time period. Two products may appear unrelated in the short run but become substitutes over a longer period as consumers adjust.

Worked Example 2: Cross Elasticity

The price of a competing online platform rises by 10%, and demand for a firm’s platform rises by 4%.

The positive value indicates substitute goods. The magnitude suggests a relatively modest response.

3. Demand Functions and Estimation

A demand function expresses quantity demanded as a function of its determinants. A simple linear specification is:

where:

Economic theory commonly predicts b < 0. The expected sign of c is positive for a normal good, while d is positive for a substitute and negative for a complement. These are hypotheses, not guarantees; empirical evidence is required.

The term “partial effect” is important. In a multiple-variable model, coefficient b measures the estimated change in quantity associated with a one-unit change in price while the other included variables are held constant.

3.1 Linear, Logarithmic, and Semi-log Models

A linear model is easy to interpret in original units but assumes a constant absolute effect. Alternative specifications include:

In a log-log model, coefficients are elasticities directly, provided the variables are positive and the specification is appropriate. For example, β estimates price elasticity.

A semi-log model may be written as:

Here, β is approximately the proportional change in quantity for a one-unit change in price, with exact interpretation depending on the model and units.

Model choice should be guided by theory, data quality, diagnostic evidence, and interpretability—not by the desire to maximise a single statistical measure.

4. Regression Analysis for Demand Estimation

Regression analysis estimates the parameters of a demand function from observed data. In ordinary least squares (OLS), coefficients are selected to minimise the sum of squared residuals:

where Q_i is observed quantity and Q̂_i is predicted quantity for observation i.

A basic regression model is:

The OLS estimates can be interpreted causally only under strong conditions. Important assumptions include:

1. The model is correctly specified enough for the purpose.
2. Relevant explanatory variables are measured adequately.
3. The error has zero conditional mean: E(u_i | X_i) = 0.
4. Observations are sufficiently independent, or dependence is appropriately modelled.
5. Severe multicollinearity does not prevent useful identification.
6. The sample is relevant to the population and decision context.

The coefficient of determination, R², measures the proportion of sample variation in the dependent variable explained by the fitted model. It does not prove causation and does not guarantee accurate forecasts outside the sample.

Statistical significance should be considered alongside economic significance. A very small effect may be statistically significant in a large sample but irrelevant to a managerial decision. Conversely, a practically important estimate may be imprecise in a small sample.

4.1 Endogeneity and Simultaneity

Demand estimation is difficult because price may respond to demand conditions. For example, a retailer may reduce prices when it anticipates weak demand. A simple regression may then show a misleading relationship. This is an endogeneity problem: price is correlated with the error term.

Possible responses include natural experiments, instrumental variables, controlled experiments, panel-data methods, fixed effects, and structural models. These methods require careful assumptions and should not be applied mechanically.

4.2 Forecast Evaluation Metrics

For forecasts F_t and actual values A_t, common error measures include:

Mean absolute error:

Root mean squared error:

Mean absolute percentage error:

MAE retains the original units. RMSE penalises large errors more heavily. MAPE is intuitive as a percentage but is problematic when actual values are zero or close to zero. Forecast evaluation should use a holdout sample or rolling-origin testing rather than assessing accuracy only on the data used to fit the model.

5. Time-Series Demand Forecasting

Time-series methods use the historical sequence of a variable. A series may contain:

• Trend: long-term movement.
• Seasonality: systematic variation within a year, week, day, or other fixed cycle.
• Cycles: longer fluctuations associated with business conditions.
• Irregular variation: unpredictable disturbances.

5.1 Moving Averages

A simple moving average of order k is:

where A_t is the actual observation at time t. It smooths random variation but may lag when a trend is present.

5.2 Exponential Smoothing

Simple exponential smoothing is:

where 0 < α < 1 is the smoothing constant. A high α responds more quickly to recent changes; a low value produces greater smoothing. This method is suitable when the series has no strong trend or seasonality. Trend-adjusted and seasonal methods are needed when those features are important.

5.3 Seasonal Indices

A seasonal index compares a period’s typical value with an appropriate average or trend level. An index above 1 indicates above-average seasonal demand; an index below 1 indicates below-average demand. Seasonal adjustment must be based on a sufficiently long and stable history, and analysts should check whether institutional changes have altered the pattern.

6. Qualitative Forecasting and Scenario Analysis

Quantitative methods are not always sufficient. New products, reorganisations, technological disruptions, or policy changes may have little historical precedent. Useful qualitative methods include expert panels, the Delphi method, structured management judgment, customer research, and market intelligence.

Judgment should be documented rather than hidden. A sound process records assumptions, evidence, confidence levels, disagreements, and responsibility for approval.

Scenario analysis develops internally coherent combinations of assumptions. For example:

Scenarios are not probabilities unless probabilities are explicitly assigned. They are structured “what-if” narratives that help managers test resilience.

A simple planning identity is:

This assumes stable productivity and clearly defined units. If productivity changes, labour hours per unit must be adjusted. Absence, turnover, training time, legal limits, and skill constraints should also be incorporated.

7. Illustrative Case: Forecasting Demand for a Regional Training Provider

Background

This illustrative case concerns SkillBridge, a fictional regional provider of professional data-analysis courses. Its HR team must decide how many instructors, coordinators, and support employees to schedule for the next quarter.

Problem

Recent fees and enrolments are:

Management is considering a fee of $440. It also expects household incomes in the target market to rise by 3% and a competing platform’s fee to rise by 5%.

Analysis

The most recent arc price elasticity was approximately −1.32. Suppose market research provides an income elasticity of 0.8 and a cross elasticity of 0.4. A rough scenario estimate can combine percentage effects:

The price change is:

Estimated effects are:

• Own-price effect: (−1.32)(−2.22%) ≈ +2.93%.
• Income effect: (0.8)(3%) = +2.40%.
• Competitor-price effect: (0.4)(5%) = +2.00%.

The combined approximate increase is 7.33%. Starting from 230 enrolments:

The planning estimate is therefore about 247 enrolments. This is not a guaranteed prediction. It assumes elasticities remain stable, effects are additive in percentage terms, and no major capacity or reputational shock occurs.

If each instructor can effectively support 25 enrolments during the quarter, the instructor requirement is:

Since partial instructors are not feasible for scheduling purposes, SkillBridge would plan for 10 instructor-equivalents, subject to availability, leave, preparation time, and class-size regulations.

Approach and Solution

The HR team should combine the scenario estimate with a regression model using longer historical data, a seasonal review, and consultation with programme managers. It should create a base case, an upside case, and a downside case rather than relying on one number.

Outcome

The recommended plan is to schedule nine instructors initially, secure a qualified tenth instructor on a flexible contract, and review enrolments after the first four weeks. This balances service capacity with the risk of overstaffing.

Learning Points

1. Elasticities translate economic assumptions into approximate demand effects.
2. Forecasts should disclose assumptions and uncertainty.
3. HR capacity planning must convert demand into hours, skills, and feasible staffing.
4. A flexible response can be superior to committing immediately to a single forecast.

8. Limitations, Ethics, and Governance

Forecasts are conditional statements, not facts. Common limitations include structural breaks, omitted variables, changing consumer preferences, poor measurement, small samples, model misspecification, and data revisions. Historical relationships may fail when technology, regulation, competition, or organisational strategy changes.

Ethical forecasting is especially important when forecasts affect people. HR leaders should avoid using opaque models to justify discriminatory recruitment, dismissal, scheduling, or pay decisions. Data should be collected lawfully, used for specified purposes, protected against unauthorised access, and reviewed for disparate impact. Sensitive personal data should not be included merely because it improves predictive accuracy.

AI-enabled forecasting can process large datasets and detect nonlinear patterns, but it can also reproduce historical bias, conceal assumptions, and create false confidence. Governance should include human review, documented model purposes, validation, audit trails, access controls, explainability appropriate to the decision, and a mechanism for correcting errors. Forecasts used in high-impact employment decisions require stronger oversight than forecasts used for low-risk inventory planning.

Best practice is to:

• begin with a clearly defined decision;
• state the forecast horizon and units;
• separate training, validation, and test data where appropriate;
• compare the model with a simple benchmark;
• inspect residuals and unusual observations;
• report uncertainty ranges, not only point estimates;
• update forecasts as new evidence arrives;
• monitor actual outcomes and document model changes;
• combine statistical evidence with informed professional judgment.

Chapter Summary

Elasticity measures percentage responsiveness. Price elasticity indicates how quantity demanded responds to own price; income elasticity measures response to income; and cross elasticity measures response to another product’s price. Elasticity may be calculated at a point or across an interval using the midpoint method. Its interpretation supports pricing, revenue analysis, market assessment, and workforce planning.

Demand functions express quantity as a function of price, income, related prices, advertising, and other determinants. Regression estimates these relationships but requires attention to assumptions, endogeneity, multicollinearity, statistical uncertainty, and economic significance. Time-series methods such as moving averages and exponential smoothing are useful when historical patterns persist, while qualitative methods and scenario analysis are essential when structural change or limited data makes history unreliable.

A professional forecast is transparent about assumptions, tested against alternatives, evaluated with appropriate error measures, and connected to an actionable decision. In HRM, the final objective is not merely to predict demand but to translate it into staffing levels, skills, schedules, training, and resilient workforce plans.

Key Terms

Elasticity: Percentage responsiveness of one variable to a percentage change in another.

Price elasticity of demand: Responsiveness of quantity demanded to own price.

Income elasticity: Responsiveness of demand to consumer income.

Cross elasticity: Responsiveness of demand for one product to another product’s price.

Point elasticity: Elasticity measured at a particular point on a demand curve.

Arc elasticity: Elasticity measured over a finite interval using average values.

Total revenue: Price multiplied by quantity sold.

Demand function: A relationship showing demand as a function of its determinants.

Intercept: Estimated value of the dependent variable when explanatory variables equal zero.

Coefficient: Estimated effect associated with an explanatory variable.

Error term: Unobserved or random influences in a statistical model.

Regression: Statistical method for estimating relationships among variables.

Endogeneity: Correlation between an explanatory variable and the model error.

Time series: Observations recorded sequentially over time.

Seasonality: Regular within-cycle variation in a time series.

Moving average: Average of recent observations used for smoothing or forecasting.

Exponential smoothing: Forecasting method that weights recent observations more heavily.

Scenario analysis: Evaluation of outcomes under alternative assumptions.

MAE: Mean absolute error, an average of absolute forecast errors.

RMSE: Root mean squared error, which gives greater weight to large errors.

Review Questions

Short Answer Questions

1. What does an elasticity of −0.6 imply about price responsiveness?
2. Distinguish between point elasticity and arc elasticity.
3. What does a positive cross elasticity usually indicate?
4. Why does a high R² not prove causation?
5. State one advantage and one limitation of exponential smoothing.

Long Answer Questions

1. Explain the economic meaning of price, income, and cross elasticities, using suitable examples.
2. Describe the main assumptions and limitations of OLS regression in demand estimation.
3. Compare moving averages, exponential smoothing, regression, and qualitative forecasting.
4. Explain how elasticity information can support HR workforce planning.
5. Discuss ethical and governance requirements for AI-supported demand and workforce forecasting.

Analytical / Application Questions

1. A product’s price rises from $80 to $88 while quantity demanded falls from 1,000 to 900 units. Calculate the arc price elasticity and classify demand.
2. A firm estimates E_d = −1.4, E_Y = 0.6, and E_xy = 0.3. Estimate the approximate percentage change in demand if own price falls by 5%, income rises by 4%, and the substitute’s price rises by 6%.
3. Design a three-scenario workforce plan for an organisation whose forecast output may grow by 2%, 6%, or 10%, identifying the staffing assumptions and risks that should be monitored.

Multiple Choice Questions with Answers

Question 1

If price rises by 10% and quantity demanded falls by 5%, the price elasticity of demand is:

A. −2.0

B. −0.5

C. 0.5

D. 2.0

Correct Answer: B. −0.5

Question 2

A positive cross elasticity of demand generally indicates that two goods are:

A. Complements

B. Inferior goods

C. Substitutes

D. Unrelated by definition

Correct Answer: C. Substitutes

Question 3

If income elasticity is 1.4, demand is most commonly described as:

A. Income-elastic

B. Income-inelastic

C. Perfectly inelastic

D. Negative-income

Correct Answer: A. Income-elastic

Question 4

Which formula correctly represents total revenue?

A. TR = P ÷ Q

B. TR = P + Q

C. TR = P × Q

D. TR = Q − P

Correct Answer: C. TR = P × Q

Question 5

In a log-log demand model, a coefficient on log price is commonly interpreted as:

A. A price elasticity

B. The price intercept

C. A quantity unit

D. A forecast error

Correct Answer: A. A price elasticity

Question 6

Which forecast measure retains the original units of the variable?

A. MAE

B. MAPE

C. A percentage error only

D. An elasticity

Correct Answer: A. MAE

Question 7

In simple exponential smoothing, a higher smoothing constant generally makes the forecast:

A. Less responsive to recent observations

B. More responsive to recent observations

C. Independent of actual data

D. Exactly equal to the long-run average

Correct Answer: B. More responsive to recent observations

Question 8

A major problem occurs when price is correlated with the regression error term. This problem is called:

A. Seasonality

B. Endogeneity

C. Smoothing

D. Deflation

Correct Answer: B. Endogeneity

Question 9

If demand is elastic, a price reduction will generally, other things equal:

A. Reduce total revenue

B. Leave total revenue unchanged in all cases

C. Increase total revenue

D. Eliminate quantity demanded

Correct Answer: C. Increase total revenue

Question 10

Scenario analysis is best understood as:

A. A guaranteed prediction of one future outcome

B. A method for evaluating coherent alternative assumptions

C. A replacement for all professional judgment

D. A calculation that requires no assumptions

Correct Answer: B. A method for evaluating coherent alternative assumptions