Managing claims, reconstructing credit: Replication data and code

This replication package supports “Managing claims, reconstructing credit: Documentary obligations and business activity in fifteenth-century London”. It contains coded observations, source locators, primary-case interpretations and executable code for reconstructing the manuscript's descriptive results from the edited Court of Common Pleas calendar published by Jonathan Mackman and Matthew Stevens on British History Online. The complete extraction comprises 6,319 calendar-entry headings from 134 term pages covering selected years between 1399 and 1500. The analytical sample contains 4,008 entries whose displayed writ field begins with Debt and whose pleading year falls between 1399 and 1468 inclusive. The four analytical cohorts are 1399–1409, 1420–1429, 1442–1450 and 1460–1468. An entry is an edited record of litigation, not a unique transaction, loan, lawsuit or economic actor. The deposited files include the complete coded entry dataset, the exact analytical subset, 8,292 direct plaintiff appearances across the complete extraction, and a twenty-entry qualitative evidence register comprising fourteen entries concerning Richard Nordon's estate and six contrasting disputes. Each entry retains its source-page identifier, actual HTML anchor and CP40 reference. A variable dictionary documents every field and coding rule; case notes distinguish alleged obligations, credited payments, representative authority, contested instruments and observed procedural dispositions. The code reproduces the manuscript's cohort table, the separate restrictions excluding mixed bond writs and retaining one non-estate plaintiff, estate-by-cohort cross-tabulations, and plaintiff-appearance totals. The first and last cohorts have bond-bearing shares of 53.8% and 69.0%. Estate entries constitute 461 of 4,008 analytical entries and contain 1,169 of the 5,322 analytical plaintiff appearances. These are record-level descriptive quantities, not measures of lending volumes, principal, realised recovery or unique creditors. The main replication command, `python -B reproduce.py`, runs offline using Python's standard library. It validates file checksums, record keys, sample membership and appearance counts, recalculates the results from the deposited data, and compares them with the reference outputs. Optional HTML-extraction code and a 134-file source checksum manifest permit exact-version source reconstruction using independently supplied local files. The publisher's full HTML, pleading narratives and complete named-person tables are not redistributed. The package retains the difference between 6,319 extracted headings and the published project total of 6,321. It does not invent the two missing entries, certify all continuations as identified, or claim a new reading of the original Latin rolls. Narrative inconsistencies and unresolved identifications are documented in the case notes. The qualitative cases are purposively selected and do not establish the prevalence of the practices they illustrate. No synthetic historical observations are included. Data coding and documentation are licensed under CC BY 4.0 to the extent of the depositor's rights; original Python code is licensed under MIT. These licences do not extend to third-party rights in the underlying edition. Full file-specific terms are in LICENSE.md. Please cite both this package and Mackman and Stevens (2010) when reusing the data.