Legal Reform or Implementation? The Traction of Governance-Related Structural Conditionality in IMF Programs after the 2018 Framework
Download PDF All ArticlesThe International Monetary Fund’s 2018 Framework for Enhanced Fund Engagement on Governance committed the institution to systematic engagement with governance vulnerabilities. Whether the resulting conditions are delivered, and whether delivery depends on what a condition asks a government to do, has not been independently established. Using the Fund’s Monitoring of Fund Arrangements database, this paper analyses 6,227 structural benchmarks attached to 89 arrangements in 59 countries approved after May 2018. Conditions are classified as governance-related by mapping the Fund’s own economic descriptors onto the six state functions named in the Framework, and as legal-adoption or operational by coded analysis of condition text. The Framework-based mapping identifies 2,775 governance conditions where the database’s anti-corruption descriptor alone identifies 382, confirming that descriptor-based identification substantially understates governance conditionality. Governance conditions are less likely to be implemented than others (odds ratio 0.74 on time, 0.70 including delay). Legal-adoption conditions are met on time in 47.1 per cent of cases against 66.9 per cent for operational conditions, but the gap closes once delayed completion is counted, and the governance-by-type interaction is null. Contrary to the capability-trap expectation, legislative conditions are not more readily delivered than administrative ones; they are delivered late.
国际货币基金组织《2018 年加强基金治理参与框架》承诺,基金组织将系统性应对治理薄弱问题。但这些框架衍生的改革条件能否落地,以及落地与否是否取决于条件本身要求一国政府采取何种行动,尚未得到独立验证。本文利用基金组织的《基金安排监测数据库》,分析了 2018 年 5 月之后获批、覆盖 59 个国家的 89 项贷款安排所附带的 6227 项结构性基准指标。研究通过将基金组织自有经济描述符匹配到该框架列明的六项国家职能,以此判定条件是否属于治理类;同时通过对条件文本进行编码分析,将条件划分为**法律采纳型**与**运行实施型**。基于该框架的匹配方法识别出 2775 项治理类条件;而仅依靠数据库的反腐败描述符只能识别 382 项,这说明依靠描述符识别会严重低估治理类条件的数量。治理类条件相较于其他条件更难落实(按时完成的比值比为 0.74;计入延期完成时比值比为 0.70)。法律采纳型条件按时达标率为 47.1%,而运行实施型条件为 66.9%;但若将延期完成情况纳入统计,二者差距将会缩小,并且治理属性与条件类型不存在显著交互效应。与能力陷阱假说的预期相反,立法类条件并不比行政类条件更容易落地,只是立法类条件往往延期完成。