Endogenous Product Design: A Linear Demand Approach

1. Problem Statement

The paper addresses the challenge of modeling endogenous product design within a tractable analytical framework. Traditional models of product differentiation often rely on specific assumptions (e.g., Hotelling location models or discrete choice models) that limit their applicability to complex market structures or require heavy computational burdens for general equilibrium analysis.

The core problem is to determine how firms optimally design product characteristics (both intensity—how much of a feature to include, and orientation—which features to prioritize) to maximize profits. The author seeks a framework that:

• Allows characteristics to affect both consumer valuations and the intensity of competition (substitutability).
• Handles any finite number of goods, firms, and characteristics.
• Accommodates both vertical (quality) and horizontal (preference) differentiation.
• Remains analytically tractable under asymmetries (cost differences) and various market structures (monopoly vs. oligopoly).

2. Methodology

A. Theoretical Framework: Characteristics-Based Linear Demand

The paper builds upon a representative consumer model with quasi-linear quadratic utility:
$$U(q, Y, p) = q'\delta - \frac{1}{2}q'Mq - \phi(Y - q'p)$$
Where:

• $q$ is the consumption bundle.
• $\delta$ is the vector of initial marginal utilities.
• $M$ is the positive definite Hessian matrix representing diminishing marginal utility and substitution patterns.
• $p$ is the price vector.

Key Innovation: Orthonormalization of Attributes
To make product design endogenous, the author transforms observable product characteristics ($X$) into orthonormal attributes ($S$) using QR decomposition and spectral decomposition.

1. Decomposition: The utility vector is expressed as $\delta = X\beta + v = S b + v$, where $S$ is an orthonormal matrix of attributes and $b$ is the vector of attribute utilities.
2. Hessian Approximation: The paper proposes a novel approximation for the Hessian matrix $M$, assuming it is partially determined by observable characteristics:
   $$\hat{M} = S \Gamma S' + \rho I$$
   Here, $\Gamma$ is a diagonal matrix representing the competitive salience of each attribute (how much differentiation in that attribute affects substitution), and $\rho I$ represents baseline idiosyncratic differentiation.
3. Cost Function: Firms face quadratic costs for increasing attribute intensity, $c_n(x_n) = \frac{1}{2}x_n'x_n$, which is re-expressed in terms of the orthonormal attributes $S$.

B. Game Structure

The model utilizes a two-stage game:

1. Stage 1: Firms choose product characteristics (attributes $S$) to maximize expected profits.
2. Stage 2: Firms set prices (Bertrand competition) or quantities (Cournot, though the paper focuses on Bertrand pricing).

The analysis covers two primary market structures:

• Multi-product Monopoly: A single firm controls $N$ goods.
• Single-product Oligopoly: $N$ firms, each controlling one good.

3. Key Contributions

1. Novel Specification of Linear Demand: The paper introduces a characteristics-based specification for linear demand that links product design directly to the Hessian matrix of the utility function. This bridges the gap between hedonic pricing models and standard linear demand systems.
2. Tractability in General Settings: Unlike previous literature that often relies on cosine similarity (which lacks micro-foundations and ignores vertical differentiation), this approach uses an orthonormal basis ($S$) to handle any number of goods and characteristics while maintaining analytical solvability.
3. Distinction Between Intensity and Orientation: The framework explicitly separates the amount of a characteristic (intensity) from the direction of the characteristic vector (orientation), allowing for precise analysis of horizontal vs. vertical differentiation.
4. Entry and Asymmetry Analysis: The model extends to analyze market entry and cost asymmetries, providing insights into how new entrants influence incumbent design strategies.

4. Key Results

A. Monopoly Pricing (Multi-product Firm)

• No Horizontal Differentiation: A profit-maximizing monopolist has no incentive to horizontally differentiate its products in observed attributes.
• Attribute Alignment: The monopolist aligns all its goods toward a single "optimal direction" (the direction yielding the greatest marginal utility per unit of cost).
• Indifference to Vertical Differentiation: Once the aggregate attribute intensity is fixed, the monopolist is indifferent to how attributes are distributed across goods (selling one "super-product" vs. a range of vertically differentiated goods yields the same profit).
• Cannibalization is Weak: The fear of cannibalizing one's own sales is not strong enough to justify diversifying the product line; the firm prefers to concentrate investments.
• Asymmetry Breaks Indifference: If attribute costs differ across goods (product-level heterogeneity), the monopolist will drop all goods except the single one with the lowest cost per marginal utility gained.

B. Oligopoly Pricing (Single-product Firms)

• Unique Symmetric Equilibrium: In a symmetric oligopoly, firms reach a unique equilibrium where they do not horizontally differentiate in observed attributes.
• Balanced Loading: Unlike the monopolist, competing firms distribute their design efforts more evenly across attributes. They emphasize attributes with lower competitive salience to minimize "business stealing" by rivals.
• Strategic Tilting: As the number of competitors increases, firms tilt their attribute vectors away from the most profitable (but highly competitive) attributes toward less salient ones to differentiate themselves strategically.

C. Heterogeneity and Entry

• Cost Asymmetry:
  • In Oligopoly, firm-level cost asymmetries sustain horizontal differentiation. Firms adjust designs to weigh cost-weighted utility against the strategic designs of competitors.
  • In Monopoly, product-level cost asymmetry leads to a "winner-take-all" outcome where the firm sells only the single most efficient good.
• Market Entry:
  • Greater similarity between an entrant and an incumbent lowers demand and profits for both, discouraging entry.
  • Post-Entry Design: Upon entry, the incumbent's optimal design tilts away from the entrant's design (in the exact opposite direction) and away from the pure monopoly orientation.
  • Niche Strategy: This dynamic suggests that "niche" or extreme designs may be a viable early-entry strategy to avoid direct competition, even if the long-run equilibrium involves broader appeal.

5. Significance and Implications

• Policy Relevance: The model provides a rigorous tool for regulators to evaluate minimum standard requirements (e.g., USB-C standards, right-to-repair laws). It predicts how such mandates might alter firm design incentives, potentially reducing horizontal differentiation or forcing firms to compete on different margins.
• Strategic Management: The findings challenge the intuition that firms should always diversify product lines to capture different market segments. Instead, under monopoly or symmetric oligopoly, concentration of design efforts on a single optimal vector is often superior.
• Demand Prediction: The framework offers a method to predict demand for new products based on their similarity to existing goods, suggesting that high similarity reduces the profitability of entry.
• Theoretical Advancement: By demonstrating that linear demand can be fully characterized by product attributes without losing tractability, the paper revitalizes the use of linear demand systems for studying complex endogenous design problems, moving beyond the limitations of discrete choice models.

In summary, Rodrigues demonstrates that when product characteristics are explicitly modeled as the drivers of both utility and competition, firms behave in counter-intuitive ways: monopolists avoid diversification, and oligopolists balance their attributes to minimize competitive friction, with cost asymmetries playing a critical role in determining whether differentiation occurs.