Digital tech on platforms can boost participation while compressing prices and commissions; partial adoption often benefits consumers and providers but may cut social welfare, and platforms prefer partial adoption when operating costs are low and full adoption at moderate costs.

Main Finding

Adopting digital technology on a monopolistic two-sided platform increases both supply and demand while lowering transaction prices and commission rates. However, partial (hybrid) adoption produces different strategic and welfare outcomes than full adoption: (i) in some regimes the price of traditional (non‑tech) services under partial adoption exceeds the price under no adoption; (ii) platforms do not always adopt technology when operating costs are high—there are three regimes of the platform’s optimal choice as a function of marginal operating cost c (low c → partial adoption; intermediate c → full adoption for a defined subrange; high c → no adoption); and (iii) partial adoption can raise consumer and provider surplus but may reduce total social welfare.

Key Points

• Setup and players:
  • Monopolistic two‑sided platform connecting consumers and providers.
  • Consumers and providers have heterogeneous preferences toward technology; platform can choose no adoption, partial adoption (tech and non‑tech channels coexist), or full adoption (only tech-enabled services).
  • Market‑clearing assumption: equilibrium supply equals demand; price p is determined by supply–demand interaction.
• Main comparative results:
  • Any adoption of digital technology (partial or full) tends to reduce unit prices and commission rates and to increase transaction volume (supply and demand).
  • Under partial adoption, traditional (non‑tech) services may be priced higher than in the no‑tech baseline because of competition/differentiation effects between tech and non‑tech offerings.
• Adoption decision and operating costs:
  • The platform’s optimal adoption strategy is nonmonotonic in the platform’s marginal operating cost c:
    • Low c: partial adoption is optimal.
    • Intermediate c: there exists a subrange where full adoption is optimal.
    • High c: the platform may optimally forgo technology adoption.
  • Thus higher operating cost does not automatically induce technology adoption.
• Welfare and distributional outcomes:
  • Consumer and provider surplus maxima follow the same three regime pattern as platform profit: low c → partial maximizes surplus; mid c → full maximizes surplus; high c → no adoption maximizes surplus.
  • Partial adoption can increase consumer and provider surplus relative to other regimes, yet it can reduce aggregate social welfare compared with alternatives.
• Pricing nuance:
  • Platform sets commission(s) (transaction commission α and technical commission β in extended model). Differentiated pricing between traditional and improved services is examined; tech adoption affects optimal α, β and p_T/p_I.
• Mechanisms captured:
  • Technology reduces consumer dissatisfaction (parameter ρ) and platform operating cost (parameter δ) but may impose consumer tech usage costs (f) and require investment cost C(k).
  • Network effects on both sides (consumers λ, providers φ) are included and affect equilibrium participation.

Data & Methods

• No empirical dataset: the paper is theoretical/analytical.
• Methods:
  • Game‑theoretic Stackelberg model: platform is the leader choosing commission rates and investment; consumers and providers respond by deciding participation and whether to offer/purchase tech or non‑tech services.
  • Market‑clearing supply = demand equilibrium used to determine prices.
  • Backward induction to derive equilibria under three technology regimes (no, partial, full).
  • Comparative statics analyze how equilibrium prices, participation, surpluses, and platform profit depend on parameters: marginal operating cost c, cost reduction from technology δ, consumer tech benefits ρ, tech usage cost f, network effects λ, φ, and investment cost C(k).
  • Extended model allows differentiated pricing and a technical commission β.
• Outputs: closed‑form expressions and threshold conditions (proofs in appendix) that identify the regime boundaries and welfare comparisons.

Implications for AI Economics

• Strategic adoption is context dependent: Policy‑makers and platform managers should not assume that higher operating costs automatically motivate full AI adoption. Optimal adoption can be partial, full, or none depending on cost parameters, tech effectiveness, and user acceptance.
• Distributional vs aggregate welfare tradeoffs: Partial (hybrid) AI deployment can benefit consumers and providers (higher surpluses) while reducing total social welfare — implying potential need for regulation or incentives (e.g., subsidies, standards) if social welfare is the objective.
• Pricing and market power: AI adoption tends to lower prices and commissions, affecting platform revenue composition and possibly intensifying competition for providers. Platforms must account for heterogeneous willingness to use AI when setting differentiated prices.
• Design of hybrid offerings: Because partial adoption can raise prices for remaining non‑tech services, platforms should carefully manage channel design and pricing to avoid unintended price increases for vulnerable user segments who avoid AI.
• Modeling recommendations for AI economists:
  • Include heterogeneity in user acceptance and provider adoption costs.
  • Incorporate supply–demand balance (market‑clearing) and network effects when assessing platform AI policies.
  • Consider nonmonotonic adoption incentives with respect to operating costs and investment costs—policy levers (subsidies, cost sharing) can shift regime boundaries.
• Empirical and normative follow-ups:
  • Empirical validation: measure real platforms’ outcomes across different adoption mixes (hybrid vs full) to test the model’s threshold predictions.
  • Dynamic and competitive extensions: study multi‑platform competition, dynamic learning/adoption, multi‑homing, and the role of data/network externalities in long‑run welfare and market structure.

If you want, I can extract the model’s key threshold expressions (conditions on c, δ, ρ, etc.) and summarize the algebraic boundaries that determine when each adoption regime is optimal.