Empirical Methods for Demand Analysis
The supply-and-demand model helps explain how markets work, but managers often face a different challenge: how customers respond to changes made by their own firm. Companies like Apple are not simply passive participants in a competitive market. They have some ability to set prices, launch products, and influence demand. Because of this, managers care less about the overall market demand curve and more about the demand curve facing their specific product.
Apple's evolution from 99¢ song downloads to tiered pricing and eventually to Apple Music illustrates this challenge. Before changing prices, managers need to predict how customers will react. A higher price may increase revenue per sale, but it may also reduce the number of purchases. The key question is whether the gain from charging more outweighs the loss from selling fewer units.
Rather than relying on intuition, managers use empirical analysis—the practice of using real-world data to estimate economic relationships. Surveys, customer behavior data, market experiments, and historical sales records can all help reveal how demand changes when price changes.
One of the most important concepts in this process is price elasticity of demand, which measures how sensitive customers are to price changes.
Understanding elasticity helps managers predict whether a price increase will raise or lower revenue.
Managers often go beyond elasticity and estimate the entire demand function, a mathematical relationship linking sales to factors such as price, income, advertising, and competitor actions. These estimates are commonly produced using regression analysis, a statistical tool that identifies relationships in data and allows managers to make evidence-based decisions.
Ultimately, managerial economics shifts the focus from simply understanding markets to using data to make better decisions. By estimating demand, measuring customer responsiveness, and forecasting future outcomes, managers can choose prices, plan investments, design marketing campaigns, and allocate resources more effectively.
Apple's evolution from 99¢ song downloads to tiered pricing and eventually to Apple Music illustrates this challenge. Before changing prices, managers need to predict how customers will react. A higher price may increase revenue per sale, but it may also reduce the number of purchases. The key question is whether the gain from charging more outweighs the loss from selling fewer units.
Rather than relying on intuition, managers use empirical analysis—the practice of using real-world data to estimate economic relationships. Surveys, customer behavior data, market experiments, and historical sales records can all help reveal how demand changes when price changes.
One of the most important concepts in this process is price elasticity of demand, which measures how sensitive customers are to price changes.
- If customers dramatically reduce purchases when prices rise, demand is highly elastic.
- If purchases change only slightly, demand is relatively inelastic.
Understanding elasticity helps managers predict whether a price increase will raise or lower revenue.
Managers often go beyond elasticity and estimate the entire demand function, a mathematical relationship linking sales to factors such as price, income, advertising, and competitor actions. These estimates are commonly produced using regression analysis, a statistical tool that identifies relationships in data and allows managers to make evidence-based decisions.
Ultimately, managerial economics shifts the focus from simply understanding markets to using data to make better decisions. By estimating demand, measuring customer responsiveness, and forecasting future outcomes, managers can choose prices, plan investments, design marketing campaigns, and allocate resources more effectively.
Elasticity: Measuring Customer Responsiveness
Managers care not only about whether demand rises or falls, but how much it changes when prices change. Elasticity is the tool economists use to measure this responsiveness. Specifically, the price elasticity of demand measures the percentage change in quantity demanded (Q) divided by the percentage change in price (p). That is, the price elasticity of demand (which we represent by the Greek letter epsilon) is:
Elasticity answers a critical managerial question: If we change our price, how will sales respond? A demand elasticity of -3 means a 1% increase in price causes quantity demanded to fall by 3%. An elasticity of -0.5 means the same price increase causes sales to fall by only 0.5%. The larger the magnitude of the elasticity, the more sensitive customers are to price changes.
Because managers usually observe demand at two different prices rather than an entire demand curve, they often calculate an arc elasticity. Arc elasticity uses the average of the starting and ending prices and quantities as the denominator under delta Q and delta p for calculating percentage changes. This approach avoids getting different answers depending on which point is treated as the starting value and provides a consistent measure of responsiveness over a range of prices.
- If ∣ε∣<1 = Inelastic
- If ∣ε∣>1 = Elastic
- If ∣ε∣=1 = Unitary
Because managers usually observe demand at two different prices rather than an entire demand curve, they often calculate an arc elasticity. Arc elasticity uses the average of the starting and ending prices and quantities as the denominator under delta Q and delta p for calculating percentage changes. This approach avoids getting different answers depending on which point is treated as the starting value and provides a consistent measure of responsiveness over a range of prices.
For managers, elasticity is far more than a mathematical exercise. It provides direct guidance on pricing decisions. A firm can experiment with prices in a small market, observe how sales change, and calculate an elasticity estimate. That estimate helps predict the impact of future price changes on sales, revenue, and profit.
The key insight is that elasticity transforms customer behavior into a measurable number. Rather than guessing how consumers will react to a price change, managers can use data to estimate demand sensitivity and make pricing decisions with greater confidence. In many industries, understanding elasticity is one of the most important tools for determining whether raising prices will increase revenue or drive customers away.
Point Elasticity: Responsiveness at a Single Point
The arc elasticity you just learned measures responsiveness between two distinct price–quantity combinations. But sometimes a manager wants to know what happens from a very small change in price at a specific point on the demand curve. That measure is called point elasticity.
The Idea
Imagine zooming in on a demand curve until the movement between two points becomes infinitesimally small. Instead of measuring responsiveness over a range of prices, you're measuring responsiveness at one exact price and quantity combination.
Point elasticity answers:
"If I raise price just a little bit from this exact point, how much will quantity demanded change?"
The Idea
Imagine zooming in on a demand curve until the movement between two points becomes infinitesimally small. Instead of measuring responsiveness over a range of prices, you're measuring responsiveness at one exact price and quantity combination.
Point elasticity answers:
"If I raise price just a little bit from this exact point, how much will quantity demanded change?"
|
|
Think of arc elasticity as the average speed during a trip, while point elasticity is the speed shown on your speedometer at a specific moment.
A key insight is that elasticity is not constant along a downward-sloping linear demand curve.
For a linear demand curve:
A key insight is that elasticity is not constant along a downward-sloping linear demand curve.
For a linear demand curve:
where b is constant, but the ratio p/Q changes as you move along the curve.
As a result:
As a result:
- Higher prices + lower quantities → demand becomes more elastic
- Lower prices + higher quantities → demand becomes more inelastic
|
|
Horizontal and Vertical Demand Curves
These are two extreme cases of elasticity that are important because they define the limits of how responsive demand can be.
Horizontal Demand Curve (Perfectly Elastic Demand)
A tiny percentage increase in price causes an enormous percentage decrease in quantity demanded. |
Vertical Demand Curve (Perfectly Inelastic Demand)
|
Other Types of Demand Elasticity
So far we've covered price elasticity of demand which measures how quantity demanded responds to changes in its own price.
Managers also care about two other major elasticities:
Managers also care about two other major elasticities:
- Income Elasticity of Demand
- Cross-Price Elasticity of Demand
|
|
Regression Analysis
Every business decision is ultimately built around relationships.
Managers want to know how sales change when prices rise, how demand responds to advertising, how wages affect productivity, or how costs change as production scales. The challenge is that the real world rarely provides clear answers. Countless factors influence outcomes simultaneously, making it difficult to isolate the effect of any single variable.
Regression analysis exists to solve this problem.
Rather than looking at individual observations and trying to draw conclusions by intuition alone, regression analysis uses data to estimate the underlying relationship between variables. It separates the outcome we are trying to explain—known as the dependent variable—from the factors we believe influence that outcome, known as explanatory variables.
In economics and business, this approach forms the foundation of what is often called econometrics: using statistical tools to measure and understand economic relationships.
Managers want to know how sales change when prices rise, how demand responds to advertising, how wages affect productivity, or how costs change as production scales. The challenge is that the real world rarely provides clear answers. Countless factors influence outcomes simultaneously, making it difficult to isolate the effect of any single variable.
Regression analysis exists to solve this problem.
Rather than looking at individual observations and trying to draw conclusions by intuition alone, regression analysis uses data to estimate the underlying relationship between variables. It separates the outcome we are trying to explain—known as the dependent variable—from the factors we believe influence that outcome, known as explanatory variables.
In economics and business, this approach forms the foundation of what is often called econometrics: using statistical tools to measure and understand economic relationships.
Building a Demand Curve from Data
Imagine a company wants to understand how price affects demand for its product.
At first, we simplify the problem. We assume that all other influences on demand—such as income levels, consumer preferences, and competitor actions—remain constant. This allows us to focus on a single relationship: how quantity demanded changes as price changes.
Under this simplified view, the demand curve becomes a mathematical description of consumer behavior.
A traditional demand function treats quantity demanded as the outcome and price as the explanatory factor. In other words, price changes first, and quantity responds.
However, the exact same relationship can be viewed from the opposite direction. Instead of asking how much consumers buy at a given price, we can ask what price consumers are willing to pay for a given quantity. This version is called the inverse demand function.
Both approaches describe the same market reality. They simply place different variables on the left side of the equation depending on what question is being asked.
The important lesson is that when performing regression analysis, the placement of variables matters because the variable being explained must appear as the dependent variable.
At first, we simplify the problem. We assume that all other influences on demand—such as income levels, consumer preferences, and competitor actions—remain constant. This allows us to focus on a single relationship: how quantity demanded changes as price changes.
Under this simplified view, the demand curve becomes a mathematical description of consumer behavior.
A traditional demand function treats quantity demanded as the outcome and price as the explanatory factor. In other words, price changes first, and quantity responds.
However, the exact same relationship can be viewed from the opposite direction. Instead of asking how much consumers buy at a given price, we can ask what price consumers are willing to pay for a given quantity. This version is called the inverse demand function.
Both approaches describe the same market reality. They simply place different variables on the left side of the equation depending on what question is being asked.
The important lesson is that when performing regression analysis, the placement of variables matters because the variable being explained must appear as the dependent variable.
Why Real Data Never Fits Perfectly
If economics only involved price and quantity, analysis would be easy.
In reality, however, demand is constantly being influenced by factors we cannot fully observe or control.
Consumer tastes change. News stories alter perceptions. Economic conditions fluctuate. Competitors launch new products. Data collection mistakes occur. Sometimes completely random events influence behavior.
As a result, actual observations rarely fall perfectly on the demand curve predicted by theory.
This is where the concept of random error becomes essential.
Random error represents all the influences on demand that are not explicitly included in our model. It captures the effects of omitted variables, unexpected events, measurement errors, and countless other factors that cause real-world outcomes to deviate from theoretical predictions.
Rather than expecting every observation to fit perfectly, regression acknowledges that noise exists and focuses on identifying the underlying pattern hidden beneath that noise.
The goal is not to predict every individual observation perfectly. The goal is to estimate the true relationship that exists on average.
In reality, however, demand is constantly being influenced by factors we cannot fully observe or control.
Consumer tastes change. News stories alter perceptions. Economic conditions fluctuate. Competitors launch new products. Data collection mistakes occur. Sometimes completely random events influence behavior.
As a result, actual observations rarely fall perfectly on the demand curve predicted by theory.
This is where the concept of random error becomes essential.
Random error represents all the influences on demand that are not explicitly included in our model. It captures the effects of omitted variables, unexpected events, measurement errors, and countless other factors that cause real-world outcomes to deviate from theoretical predictions.
Rather than expecting every observation to fit perfectly, regression acknowledges that noise exists and focuses on identifying the underlying pattern hidden beneath that noise.
The goal is not to predict every individual observation perfectly. The goal is to estimate the true relationship that exists on average.
The Fish Market Example
A fish market provides a useful illustration of how economists think about causality.
At the Portland Fish Exchange, fishing boats bring varying amounts of fish to market each day. Because fish spoil quickly, whatever arrives must be sold immediately. Buyers and sellers then interact through an auction process that determines the day's price.
What's interesting is that causality runs differently than in a typical demand example.
The amount of fish arriving at the market is largely determined by weather conditions, fishing success, and other supply factors. Once that quantity arrives, the market price adjusts to ensure all of the fish can be sold.
In this case, quantity is effectively determining price rather than the other way around.
Because quantity is doing the "explaining," economists estimate an inverse demand function, where price becomes the dependent variable.
The lesson is that regression is not just about plugging numbers into formulas. Before any analysis begins, managers must think carefully about which variable is driving the relationship and which variable is responding to it.
At the Portland Fish Exchange, fishing boats bring varying amounts of fish to market each day. Because fish spoil quickly, whatever arrives must be sold immediately. Buyers and sellers then interact through an auction process that determines the day's price.
What's interesting is that causality runs differently than in a typical demand example.
The amount of fish arriving at the market is largely determined by weather conditions, fishing success, and other supply factors. Once that quantity arrives, the market price adjusts to ensure all of the fish can be sold.
In this case, quantity is effectively determining price rather than the other way around.
Because quantity is doing the "explaining," economists estimate an inverse demand function, where price becomes the dependent variable.
The lesson is that regression is not just about plugging numbers into formulas. Before any analysis begins, managers must think carefully about which variable is driving the relationship and which variable is responding to it.
|
|
Why "Best Fit" Matters
This raises an important question:
If no line passes perfectly through all the data points, how does regression decide which line is best?
The answer is the principle of Ordinary Least Squares (OLS).
OLS selects the line that minimizes the total prediction error across all observations. Rather than focusing on one point at a time, it evaluates how far every observation lies from the predicted line and chooses the line that makes those deviations collectively as small as possible.
In effect, OLS asks:
"Which line does the best overall job of explaining the data?"
The result is not a perfect description of every observation, but it is the most statistically defensible estimate of the underlying relationship.
If no line passes perfectly through all the data points, how does regression decide which line is best?
The answer is the principle of Ordinary Least Squares (OLS).
OLS selects the line that minimizes the total prediction error across all observations. Rather than focusing on one point at a time, it evaluates how far every observation lies from the predicted line and chooses the line that makes those deviations collectively as small as possible.
In effect, OLS asks:
"Which line does the best overall job of explaining the data?"
The result is not a perfect description of every observation, but it is the most statistically defensible estimate of the underlying relationship.
Statistical Significance: Knowing When to Trust the Numbers
This chapter is less about learning new regression mechanics and more about answering a manager's most important question:
"How much should I trust the results?"
Anyone can run a regression and get a coefficient. The harder question is whether that coefficient reflects a real relationship or merely random chance. This chapter introduces the statistical tools economists use to distinguish meaningful findings from noise.
"How much should I trust the results?"
Anyone can run a regression and get a coefficient. The harder question is whether that coefficient reflects a real relationship or merely random chance. This chapter introduces the statistical tools economists use to distinguish meaningful findings from noise.
Imagine you conduct a survey to estimate how consumers respond to changes in price. After running a regression, you discover that every $1 increase in price reduces demand by 1.4 units.
That number sounds precise, but should you believe it?
If you surveyed a different group of customers tomorrow, would you get a similar result? Or would the estimate change dramatically?
These questions sit at the heart of statistical significance.
Regression does not simply provide estimates—it also provides a way to measure how confident we should be in those estimates. The goal is not merely to calculate a relationship, but to determine whether the relationship is reliable enough to support business decisions.
That number sounds precise, but should you believe it?
If you surveyed a different group of customers tomorrow, would you get a similar result? Or would the estimate change dramatically?
These questions sit at the heart of statistical significance.
Regression does not simply provide estimates—it also provides a way to measure how confident we should be in those estimates. The goal is not merely to calculate a relationship, but to determine whether the relationship is reliable enough to support business decisions.