Institutional Determinants of Privacy Design in Central Bank Digital Currencies

Central bank digital currencies (CBDCs) make transaction privacy an explicit institutional design choice. We ask whether political and economic conditions predict the privacy posture documented by central banks, while separating privacy vocabulary from design commitments. A frozen multilingual corpus contains 113 official or authority-attributed documents (3,963 pages) across forty-seven jurisdictions and two institutional composites. Two independently run large-language-model extractors produced 11,358 source-verified statements, deduplicated into 6,949 candidates. We aggregate them with a predeclared equal-vote ensemble in which every candidate has unit mass and provider disagreements are retained rather than averaged away. On a separate calibration set, the verified union attained precision 0.980, recall 0.766 and span fidelity 1.000; it failed the preregistered recall gate and is therefore treated as exploratory, not automated gold. In the blind probability sample, reviewer-specific design-weighted validity was 0.786 and 0.761; inclusion agreement was 62.6% (Gwet’s AC1 = 0.426; Krippendorff’s alpha = −0.069), and exact-code agreement among the 214 jointly retained candidates was 46.7%. In the production union, the privacy-family share and documented privacy posture correlate only r = 0.219. No tested institutional condition survives family-wise correction; the largest commitment-side association is with the informal economy (r = −0.286). The design demonstrates both the value and the current limits of reproducible multi-model qualitative extraction.

Keywords: central bank digital currency; privacy; institutional design; csQCA; financial infrastructure
JEL: E42, E58, C83