Moderate data sharing maximizes welfare: a theoretical model finds that sharing big data incurs privacy costs that can slow welfare initially but, when calibrated optimally, drives long-run growth through innovation and multiplier effects. Policymakers should balance privacy protections against the long-term productivity gains of data-driven R&D.

Main Finding

The paper develops a macroeconomic model that endogenizes big data as a production factor and shows that there is an interior (optimal) degree of data sharing that maximizes individual welfare by balancing consumption gains against privacy costs. Big data can suppress welfare growth in the short run (because of adjustment/transition costs and R&D pattern shifts) but fosters sustained welfare gains in the long run via a multiplier effect that raises intermediate-goods quality and accelerates technological progress.

Key Points

• Big data is modeled as a non-rival, low-cost production factor that firms can share economy-wide; its application level enters the Cobb–Douglas production function alongside labor, intermediate goods, and technology.
• The model integrates two distinct channels by which big data influences technological progress:
  • A multiplier effect that speeds up quality-ladder improvements of intermediate goods (raising the quality of inputs).
  • A transformation of R&D patterns (changing how and where technological progress is generated).
• Individuals’ welfare is modeled as a function of consumption and a privacy cost arising from data sharing. Thus welfare reflects both economic gains from higher consumption and losses from privacy intrusion.
• There is a trade-off: more data sharing increases output and consumption (via the multiplier and tech channels) but raises privacy costs. Welfare is non-monotonic in sharing intensity; optimal sharing balances these effects.
• Dynamic outcome: short-term welfare can fall after increasing data sharing (due to adjustment and transitional costs), while long-term welfare rises as data-driven innovation accumulates.
• Contributions highlighted by authors:
• Endogenizing big data in the production function and formalizing its multiplier effect on intermediate-goods quality.
• Decomposing technological progress into production-technology improvements and intermediate-goods quality gains (beyond classic models that focus only on intermediate-goods quality).
• Introducing privacy costs into a macro welfare framework and solving for an optimal data-sharing degree.

Data & Methods

• Theoretical framework: continuous-time macro growth model grounded in the creative-destruction / quality-ladder literature (Aghion–Howitt style), extended to include big data D as an endogenous input.
• Production side: aggregate output Y is built from N symmetric final goods; each final good uses labor, a quality-adjusted intermediate input, production technology A, and the economy-wide application scale of big data Dj,t in a Cobb–Douglas form.
• Intermediate goods: quality-ladder setup where firms sequentially improve the quality κ of intermediate inputs; big data enters via a multiplier f(t) that raises the pace/size of quality improvements.
• Welfare side: individual utility depends positively on consumption and negatively on privacy costs associated with data sharing; consumption is driven by production outcomes that depend on both the multiplier effect and R&D pattern shifts.
• Analysis: characterization of the optimal degree of data sharing (interior solution), dynamic comparative statics for short- vs long-run welfare, and numerical calibration/simulation to illustrate and test theoretical predictions (parameter calibration and simulation results are reported in Section 4 of the paper).
• Key modeling assumptions: product homogeneity and symmetry across goods, current-period data only (past data loses effectiveness), non-durability of intermediate inputs, perfect competition in final goods markets.

Implications for AI Economics

• Policy design: Regulators should aim for an optimal, not maximal, level of data sharing. Policies that ignore privacy costs and push for unfettered data openness may reduce welfare in the short run and could be welfare-suboptimal overall.
• Privacy regulation and institutional design: Effective privacy protection, data-governance frameworks, or compensated opt-in mechanisms can shift the welfare trade-off (reduce privacy cost per unit of sharing), allowing greater beneficial sharing without the same welfare loss.
• Role of data intermediaries and market design: Because data is non-rival and yields economy-wide externalities (via the multiplier on intermediate-goods quality), there is a potential role for intermediaries, public infrastructure, or coordination mechanisms to achieve efficient sharing levels and internalize externalities.
• Long-run investment vs short-run disruption: Firms and policymakers should expect transitional costs as production and R&D patterns shift (creative destruction). Support measures (retraining, investment subsidies for new intermediate inputs, transitional assistance) can reduce short-run welfare losses and hasten long-run gains.
• AI firm strategy: AI-driven firms that harvest or aggregate data should account for consumer privacy valuation in their product and data-pricing strategies; welfare-maximizing outcomes may require explicit compensation or privacy-preserving technologies to expand beneficial data use.
• Research directions: The framework invites empirical calibration of privacy-cost functions, measurement of the multiplier effect of data on intermediate-goods quality, and testing of the model’s short- vs long-run predictions in sectoral or country-level data. Model limitations (e.g., current-data-only assumption, product homogeneity) suggest further extensions to capture data persistence, heterogenous goods, and firm heterogeneity.