Statistical Demand, Supply & Econometric Elasticity Analysis

The empirical estimation of market demand and supply relationships bridges abstract microeconomic theory and real-world commercial and macroeconomic policy. In competitive markets, resource allocation, consumer welfare, tax incidence, and monopoly markups hinge on the precise quantitative responsiveness of consumers and producers to shifts in prices and incomes.

Under Marshallian consumer optimization, an individual consumer maximizes an ordinal utility function $U(q_1, q_2, \dots, q_n)$ subject to the linear budget constraint $\sum p_i q_i \le M$, where $M$ denotes nominal money income. The resulting first-order conditions yield the Marshallian Demand Functions:

Marshallian demand functions satisfy three fundamental mathematical properties:

1. Homogeneity of Degree Zero: Multiplying all prices and money income by an arbitrary positive scalar $\lambda > 0$ leaves physical quantity demanded unchanged (absence of money illusion):

2. Engel Aggregation: The income-weighted sum of income elasticities equals unity:

where $w_i = (p_i q_i)/M$ is the expenditure budget share of good $i$.

3. Cournot Aggregation: The expenditure-weighted sum of price elasticities across goods satisfies:

In statistical economics, elasticity is defined as the dimensionless ratio of the proportional change in quantity to the proportional change in an underlying economic driver.

1. Own-Price Elasticity of Demand ($\eta_p$)

• Point Elasticity: For a continuous, differentiable demand curve $Q = D(P)$:

By the Law of Demand, $dQ/dP < 0$, so $\eta_p$ is typically negative. In empirical economic literature, analysts either report the signed value (e.g., $-1.5$) or reference its absolute magnitude $|\eta_p|$.

• $|\eta_p| > 1$: Price Elastic (proportional quantity response exceeds price change).
• $|\eta_p| = 1$: Unitary Elastic (total expenditure $P \times Q$ remains constant).
• $0 \le |\eta_p| < 1$: Price Inelastic (proportional quantity response is smaller than price change).

• Arc Elasticity (Midpoint Formula): When evaluating discrete price changes between $(P_1, Q_1)$ and $(P_2, Q_2)$, calculating elasticity relative to the initial point yields different results than calculating relative to the final point. To guarantee symmetry and path-invariance, economists use the Allen-Bowley Midpoint Arc Elasticity Formula:

2. Income Elasticity of Demand ($\eta_M$)

Income elasticity measures consumer responsiveness to changes in real household income $M$:

• Normal Goods ($\eta_M > 0$): Consumption expands as income increases.
• Necessities ($0 < \eta_M \le 1$): Expenditure increases slower than income; budget share $w_i$ contracts.
• Luxuries ($\eta_M > 1$): Expenditure expands faster than income; budget share $w_i$ rises.
• Inferior Goods ($\eta_M < 0$): Consumption contracts as income increases (e.g., coarse staple grains, low-cost transit).

3. Cross-Price Elasticity of Demand ($\eta_{ij}$)

Cross-price elasticity measures the percentage shift in quantity demanded of commodity $i$ induced by a one-percent change in the price of a separate commodity $j$:

• Substitutes ($\eta_{ij} > 0$): A price hike in good $j$ shifts demand toward good $i$ (e.g., tea and coffee, gasoline and electric vehicles).
• Complements ($\eta_{ij} < 0$): A price hike in good $j$ depresses demand for good $i$ (e.g., automobiles and motor fuel, hardware and software).
• Independent Goods ($\eta_{ij} = 0$).

An Engel Curve traces the equilibrium relationship between the quantity demanded (or expenditure) of a specific commodity and household income $M$, holding all market prices constant.

In 1857, the German statistician Ernst Engel discovered the empirical regularity known as Engel's Law: as household income rises, the proportion of income allocated to food declines, even if the absolute monetary expenditure on food increases:

To fit empirical consumer survey data, economists deploy several parametric functional specifications:

In empirical research, estimating demand and supply parameters from observational market data introduces deep econometric challenges.

The Double-Log Econometric Demand Specification

A standard Cobb-Douglas type empirical demand function is formulated as:

Taking natural logarithms linearizes the parameters:

where:

• $\beta_0 = \ln(A)$ is the intercept.
• $\beta_1 = \frac{\partial \ln(Q)}{\partial \ln(P)}$ is the constant own-price elasticity of demand.
• $\beta_2 = \frac{\partial \ln(Q)}{\partial \ln(M)}$ is the constant income elasticity of demand.
• $\beta_3 = \frac{\partial \ln(Q)}{\partial \ln(P_c)}$ is the cross-price elasticity.
• $u_t \sim \text{iid}(0, \sigma^2)$ is the random disturbance term.

Because variables are in logarithmic units, the estimated OLS regression coefficients directly yield the theoretical elasticities without requiring post-estimation conversion.

The Simultaneity and Identification Problem (Working's Theorem)

First formalized by E. J. Working in 1927, the Identification Problem warns that observable market data $(P_t, Q_t)$ represent simultaneous equilibrium intersections of shifting supply and demand curves:

If an analyst simply runs an OLS regression of $Q_t$ on $P_t$, the resulting regression line traces neither the demand curve nor the supply curve, but a meaningless hybrid path of moving intersection points.

Conditions for Identification:

1. Identification of the Demand Curve: The demand curve is statistically identified if and only if there exists an exogenous instrument or supply-shifter variable (such as rainfall, agricultural fertilizer costs, or manufacturing wage shocks) that enters the supply equation but is strictly excluded from the demand equation. Shifting the supply curve back and forth across a stable demand curve traces out the demand schedule.

2. Identification of the Supply Curve: The supply curve is identified if and only if there exists an exogenous demand-shifter (such as consumer income or demographic population size) that shifts the demand curve across a stable supply curve.

While single-equation log-log demand models provide convenient elasticity estimates, microeconomic theory dictates that multi-commodity demand functions must satisfy system-wide consumer optimization restrictions: Slutsky symmetry, homogeneity of degree zero, and budget exhaustion. In 1980, Angus Deaton and John Muellbauer revolutionized applied consumer economics by introducing the Almost Ideal Demand System (AIDS).

Mathematical Specification of the AIDS Model

Derived from the Price-Independent Generalized Logarithmic (PIGLOG) expenditure function, the AIDS model expresses the expenditure budget share $w_i = (p_i q_i) / M$ of good $i$ as a linear function of the logarithms of individual commodity prices and real household expenditure:

where:

• $w_i$ is the budget share of good $i$ ($\sum w_i = 1$).
• $p_j$ is the price of commodity $j$.
• $M$ is total nominal household expenditure.
• $P^*$ is an aggregate translog price index, commonly approximated in empirical research by the Stone Price Index:

Theoretical Restrictions on AIDS Parameters:

To conform to neoclassical utility maximization, the parameter matrices must satisfy three sets of linear restrictions:

1. Adding-Up (Budget Exhaustion):

2. Homogeneity of Degree Zero in Prices and Income:

3. Slutsky Symmetry:

Elasticity Formulas in the AIDS Model:

• Expenditure (Income) Elasticity:

• If $\beta_i > 0$, good $i$ is a Luxury ($\eta_{i,M} > 1$); its budget share expands as total expenditure rises.
• If $\beta_i < 0$, good $i$ is a Necessity ($\eta_{i,M} < 1$); its budget share contracts as total expenditure rises, conforming to Engel's Law.
• Uncompensated (Marshallian) Own-Price Elasticity:

• Cross-Price Elasticity:

The AIDS framework allows econometricians to test whether empirical consumer choices conform to utility theory while estimating rich price and income substitution matrices across large consumer baskets.

When governments introduce excise taxes, import tariffs, or environmental carbon fees, consumer prices rise and equilibrium consumption contracts. Econometric demand equations enable precise mathematical quantification of the resulting welfare impacts.

Compensating and Equivalent Variation

Let $v(p, M)$ denote the indirect utility function, and let $e(p, u)$ denote the expenditure function. When prices rise from $p^0$ to $p^1$:

1. Compensating Variation ($CV$): The additional monetary compensation a consumer requires at the new prices $p^1$ to restore their original baseline utility level $u^0$:

2. Equivalent Variation ($EV$): The monetary reduction in base income at initial prices $p^0$ that would reduce the consumer's welfare by the same amount as the price hike:

Marshallian Consumer Surplus and Harberger Deadweight Loss

For moderate price changes where income effects are small, consumer welfare loss is approximated by the change in Marshallian Consumer Surplus ($\Delta CS$):

When an ad valorem tax rate $t$ is levied on a competitive commodity with initial price $P_0$ and quantity $Q_0$:

• Price paid by buyers: $P_1 = P_0 (1 + t)$
• New quantity demanded: $Q_1 = Q_0 [1 - |\eta_p| t]$
• Government Tax Revenue:

• Deadweight Loss (Harberger Excess Burden):

The deadweight loss of taxation increases with the square of the tax rate ($t^2$) and proportionally with the price elasticity of demand ($|\eta_p|$). This foundational statistical theorem underpins optimal tax policy: to minimize societal welfare loss, governments should levy higher tax rates on highly price-inelastic necessities and lower rates on highly elastic goods (the Ramsey Tax Rule).

In many agricultural and raw commodity markets, production decisions require an extended biological gestation period: farmers decide how many acres of crops to plant or livestock to breed based on current market prices, but the resulting harvest output cannot reach the market until the following season. This temporal production lag produces dynamic price-quantity oscillations formalized by the Cobweb Theorem (Mordecai Ezekiel, 1938).

1. Mathematical Formulation

Consider a market with an instantaneous linear demand curve and a lagged linear supply curve:

In competitive market clearing at period $t$, quantity demanded must equal quantity supplied:

Rearranging yields a first-order linear difference equation in market price $P_t$:

2. Long-Run Equilibrium and Stability Analysis

The long-run static equilibrium price $P^*$ satisfies $P_t = P_{t-1} = P^*$:

Subtracting $P^*$ from the dynamic difference equation defines the deviation $p_t = P_t - P^*$:

Iterating backwards from an initial price displacement $p_0$:

Because the base ratio $(-\delta / \beta)$ is strictly negative, the price trajectory inevitably alternates signs from period to period, producing cyclical oscillations around the equilibrium.

Three Stability Regimes:

1. Convergent (Damped) Cobweb: $\delta < \beta \iff |-\delta/\beta| < 1$.

The supply curve is steeper than the demand curve ($|dQ^S/dP| < |dQ^D/dP|$). Price oscillations dampen over time, spiraling inward toward equilibrium $P^*$.

2. Divergent (Explosive) Cobweb: $\delta > \beta \iff |-\delta/\beta| > 1$.

The supply curve is flatter than the demand curve. Price fluctuations amplify progressively outward, destabilizing the market.

3. Continuous (Perpetual) Cobweb: $\delta = \beta \iff |-\delta/\beta| = 1$.

The system cycles indefinitely in a closed rectangular loop with constant amplitude.

As established by Working's identification problem, estimating market demand and supply equations via Ordinary Least Squares yields biased and inconsistent estimates due to the simultaneity of price and quantity: price is endogenous ($E[P_t u_t] \ne 0$).

To consistently estimate structural demand and supply parameters, econometricians deploy Two-Stage Least Squares (2SLS) with instrumental variables.

The Structural Demand-Supply System

Consider an agricultural commodity market:

where:

• $Q_t$ (quantity) and $P_t$ (price) are endogenous variables.
• $Y_t$ (consumer income) is an exogenous demand-shifter.
• $W_t$ (weather rainfall index or fertilizer cost) is an exogenous supply-shifter.

The Two-Stage Estimation Algorithm:

1. Stage 1: Reduced-Form Regression:

Regress the endogenous price $P_t$ on all exogenous variables in the entire system ($Y_t$ and $W_t$) using OLS:

Compute the fitted (predicted) values:

The predicted price $\hat{P}_t$ captures only the exogenous variation in price driven by weather and income, purging the correlation with the structural demand disturbance $u_t$.

2. Stage 2: Structural Estimation:

Replace the endogenous price $P_t$ with its fitted instrument $\hat{P}_t$ in the structural demand equation and estimate via OLS:

The resulting 2SLS estimator $\hat{\alpha}_1$ is asymptotically consistent, providing unbiased estimates of true market demand price elasticity.

In modern industrial economics, marketing science, and transportation planning, consumers do not choose continuous quantities of homogeneous goods along a smooth marginal utility curve. Instead, they choose a single, indivisible discrete alternative from a finite choice set (e.g., choosing whether to commute by personal automobile, subway, bus, or bicycle; or choosing which brand of smartphone to purchase).

To model these decisions, Nobel laureate Daniel McFadden (1974) developed the Random Utility Maximization (RUM) Framework.

1. Mathematical Formulation

Let consumer $n$ face a choice set of $J$ discrete alternatives: $\mathcal{C} = \{1, 2, \dots, J\}$. The utility consumer $n$ derives from choosing alternative $i$ is decomposed into a deterministic systematic component $V_{ni}$ and a random idiosyncratic stochastic component $\epsilon_{ni}$:

The systematic utility $V_{ni}$ is a linear function of observable attributes of the alternative (price $P_i$, travel time $T_i$, comfort quality $Q_i$) and socioeconomic characteristics of the consumer (income $M_n$, age, household size):

Consumer $n$ chooses alternative $i$ if and only if it provides the highest utility among all available options:

2. The Multinomial Logit (MNL) Model

McFadden proved that if the random disturbances $\epsilon_{ni}$ are independently and identically distributed according to the Type I Extreme Value (Gumbel) Distribution:

the choice probability takes the celebrated closed-form Multinomial Logit (MNL) formula:

3. The Independence of Irrelevant Alternatives (IIA) Property

A fundamental mathematical property of the standard multinomial logit model is the Independence of Irrelevant Alternatives (IIA): the ratio of the choice probabilities between any two alternatives $i$ and $k$ depends strictly on the attributes of those two options and is completely independent of the presence, absence, or attributes of any other alternative $j$:

The Red Bus / Blue Bus Paradox:

While computationally elegant, IIA imposes rigid, unrealistic cross-price substitution patterns. In McFadden's famous thought experiment, if commuters choose between driving a car ($P_{\text{car}} = 0.50$) and taking a blue bus ($P_{\text{blue}} = 0.50$), introducing a red bus that is identical to the blue bus in every respect should split the bus ridership equally, leaving car ridership at 50%. Under IIA, however, the model forces all three alternatives to equal 33.3%, erroneously predicting that the introduction of a red bus causes half of all drivers to abandon their cars.

To resolve IIA distortions, modern econometrics deploys the Nested Logit Model (grouping close substitutes into hierarchical branches) and the Random Coefficients Mixed Logit Model (BLP: Berry, Levinsohn, and Pakes 1995), which allows consumer preference parameters $\beta$ to vary continuously across the population, accommodating realistic, flexible cross-price elasticity matrices across differentiated product markets.

In applied microeconomics and antitrust regulatory proceedings, competition authorities must assess whether firms in concentrated industries (such as commercial banking, telecommunications, or airline transport) behave competitively or exploit market power to extract monopoly rents. Because proprietary marginal cost data are rarely disclosed to econometricians, Timothy Bresnahan (1982) and Lawrence Lau (1982) developed an econometric methodology to identify market conduct directly from observable aggregate market price and quantity data.

1. Structural Formulation

Consider an oligopolistic market with aggregate demand:

where $Y_t$ is consumer income, and $Z_t$ is an exogenous demand rotator (such as the price of a complementary service) that alters the slope of the demand curve.

The industry marginal revenue perceived by a representative firm is:

where $\theta \in [0, 1]$ is the unobserved Conduct Parameter:

• $\theta = 0$: Perfect Competition (Price equals Marginal Cost, $P = MC$).
• $\theta = 1/N$: Symmetric Cournot-Nash Oligopoly with $N$ firms.
• $\theta = 1$: Perfect Collusion / Joint Monopoly Profit Maximization.

Assuming a linear marginal cost curve:

where $W_t$ is an exogenous input cost (e.g., labor wages, energy prices).

In profit-maximizing supply equilibrium ($MR_t = MC_t$):

where $Q_t^* = \frac{Q_t}{\alpha_1 + \alpha_3 Z_t}$ is the rotated effective quantity variable.

2. Simultaneous Estimation and Hypothesis Testing

By estimating the demand equation and the supply relation simultaneously via Two-Stage Least Squares or Full Information Maximum Likelihood (FIML), the econometrician identifies $\alpha_1, \alpha_3$ from the demand curve, which allows the construction of $Q_t^*$. The regression coefficient on $Q_t^*$ in the supply relation directly identifies the conduct parameter $\theta$. Statistical agencies can then formally test the null hypothesis of perfect competition ($H_0: \theta = 0$) against the alternative of oligopolistic collusion ($H_1: \theta > 0$), providing court-admissible empirical evidence of market anti-competitive conduct.

In empirical microeconomics and production economics, researchers often reject the classical Cobb-Douglas production function ($Y = A K^\alpha L^\beta$) and Constant Elasticity of Substitution (CES) function because they impose restrictive a priori constraints: Cobb-Douglas forces the elasticity of substitution between capital and labor to equal unity ($\sigma = 1$), while CES forces it to remain constant across all output levels.

To model unrestricted substitution dynamics, Laurits Christensen, Dale Jorgenson, and Lawrence Lau (1973) developed the Transcendental Logarithmic (Translog) Production Frontier.

Mathematical Formulation

The translog production function represents a second-order Taylor series approximation to an arbitrary, twice continuously differentiable production function:

where:

• $Y$ is total physical output.
• $X_i, X_j$ are input factors (e.g., Capital $K$, Labor $L$, Energy $E$, Materials $M$).
• By Young's theorem on the symmetry of cross-partial derivatives:

1. Output Elasticities with Respect to Inputs

The output elasticity of factor $i$ is:

Notice that unlike Cobb-Douglas (where elasticity is the constant parameter $\alpha_i$), the translog output elasticity varies continuously with the levels of all inputs deployed. Under constant returns to scale:

2. Allen-Uzawa and Morishima Elasticities of Substitution

The elasticity of substitution between factor $i$ and factor $j$ in the translog cost dual function $C(w, Y)$ is given by the Allen-Uzawa Partial Elasticity of Substitution (AES):

where $s_i = w_i X_i / C$ is the cost share of input $i$. Because the Allen-Uzawa elasticity can produce ambiguous signs in multi-factor settings, modern econometrics prefers the Morishima Elasticity of Substitution (MES):

The Morishima elasticity measures the percentage change in the input ratio $(X_i / X_j)$ induced by a one-percent change in the price of factor $j$ ($w_j$). It is inherently asymmetric ($\sigma_{ij}^{\text{MES}} \ne \sigma_{ji}^{\text{MES}}$), accurately capturing that adjusting capital equipment in response to a wage hike is structurally different from adjusting labor hiring in response to an interest rate shock.

In environmental and urban economics, public goods such as clean atmospheric air, urban green canopy, quiet residential neighborhoods, and high-performing public schools do not trade in separate private markets. To quantify consumer valuation and willingness to pay for these non-market amenities, Sherwin Rosen (1974) formulated the Hedonic Property Pricing Method.

1. The Hedonic Price Function

Let $P_i$ denote the market sale price of residential property $i$. A house is viewed as a heterogeneous bundle of structural attributes $S_i$ (square footage, bedrooms, bathrooms, lot size), neighborhood socioeconomic characteristics $N_i$ (crime rates, school test scores), and environmental amenities $Z_i$ (ambient particulate pollution $\text{PM}_{2.5}$, proximity to parks, highway traffic noise):

The parameter $\theta = \frac{\partial \ln(P)}{\partial Z}$ represents the percentage change in property value resulting from a one-unit change in environmental quality. The Marginal Implicit Price (Marginal Willingness to Pay, MWTP) for amenity $Z$ is:

2. Capitalization and Welfare Evaluation

When a public infrastructure project or clean air regulation reduces local air pollution from $Z_0$ to $Z_1$:

• Housing values in the affected zone appreciate, a phenomenon known as Amenity Capitalization.
• Under the first-stage Rosen hedonic equilibrium, the aggregate capitalized property value increase across all $H$ affected homes provides an empirical measure of the societal welfare benefit:

By analyzing thousands of real estate mortgage and title transfer records via hedonic regression, economic statisticians supply court-tested cost-benefit valuations for environmental policy mandates and public infrastructure projects.

In natural monopoly utility markets (such as national electrical power grids, municipal water supply networks, postal networks, and passenger railway infrastructure), high fixed infrastructure costs mean that average costs decline continuously over relevant market demand (economies of scale). If regulators force the utility to price at marginal cost ($P = MC$), the firm incurs severe operational deficits and bankrupts ($P < AC$). If the utility is left unregulated, it charges monopoly prices, generating massive deadweight welfare loss.

To resolve this dilemma, Frank Ramsey (1927) and Marcel Boiteux (1956) formulated the theory of Optimal Second-Best Regulated Pricing (Ramsey-Boiteux Pricing).

Mathematical Formulation

Consider a multi-product regulated public utility producing $n$ distinct services (e.g., residential peak electricity, commercial off-peak electricity, industrial power). The regulator seeks to maximize total societal consumer surplus subject to the financial self-sufficiency constraint that the utility earns zero economic loss (total revenues cover total costs):

Formulating the Lagrangian with shadow multiplier $\lambda > 0$:

Differentiating with respect to each price $p_i$ and assuming zero cross-price substitution across independent services yields the Inverse Elasticity Rule for Optimal Taxation and Regulation:

where:

• $\frac{p_i - MC_i}{p_i}$ is the proportional markup over marginal cost.
• $|\eta_{i, p}|$ is the absolute own-price elasticity of demand for service $i$.
• $\frac{\lambda}{1 + \lambda} \in (0, 1)$ is the normalized Ramsey shadow price of the public utility's budget constraint.

Economic and Equity Implications:

The Ramsey rule establishes that to minimize societal deadweight loss while balancing the utility's budget:

• Services with highly price-inelastic demand (e.g., basic residential lighting where $|\eta_p| \approx 0.2$) must bear a high percentage markup over marginal cost.
• Services with highly price-elastic demand (e.g., industrial manufacturing power with competitive on-site alternatives where $|\eta_p| \approx 1.5$) must receive a low percentage markup close to marginal cost.

While mathematically efficient in minimizing Harberger deadweight triangles, pure Ramsey pricing can produce regressive distributional consequences by charging captive residential households higher markups than footloose corporate consumers, requiring regulators to integrate distributional equity weights into statutory rate-setting tariffs.