Working Paper: CEPR ID: DP17842
Authors: Giacomo Calzolari; Anatole Cheysson; Riccardo Rovatti
Abstract: Recent technological developments in ICT and Artificial Intelligence allow extracting valuable information from data that machines generate with production, machine data (MD). Although possibly more valuable than personal data, the growing market for MD and its analytics may suffer from several issues, such as datasets fragmented into many small data-producers, externalities as with non-rival information and fuzzy property rights. We combine these market elements with critical properties that we derive from actual Machine Learning algorithms for analytics. We explore how and to what extent a data aggregator can operate, contracting with different data producers to share data and analytics. We identify conditions that impact the market organization for MD, such as producers' heterogeneity, their preference for anonymity, and the intensity of competition in final markets.
Keywords: Data Analytics; Machine Learning; Artificial Intelligence; Machine Generated Data; Nonpersonal Data; IoT; 5G; ICT; Enabling Technology; Market Organization; Externality; Anonymity; Property Rights; Competition
JEL Codes: L12; L13; M11; M15; D21; D43
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| fragmentation of datasets across small producers (D20) | inefficiencies in data utilization (D61) |
| data aggregator can improve market outcomes by pooling data from multiple sources (C43) | overall market efficiency (G14) |
| data producers share their MD (D20) | enhance the analytics produced (C55) |
| enhance the analytics produced (C55) | increase overall market efficiency (G14) |
| preference for anonymity among producers (D83) | constrain the value of analytics (Y10) |
| constrain the value of analytics (Y10) | market breakdowns (G10) |
| intensity of competition in final markets (L13) | affects data sharing decisions (D91) |
| higher competition (L13) | reduce the willingness of producers to share data (D16) |
| reduce the willingness of producers to share data (D16) | limit the effectiveness of analytics (C91) |