Improving Overnight Loan Identification in Payments Systems
研究了Furfine算法识别场外银行间隔夜贷款的准确性，发现每日误报率上限为10%-20%，并提出改进方法将上限降至10%以下，适用于加拿大数据。
Information on the allocation and pricing of over‐the‐counter (OTC) markets is scarce. Furfine (1999) pioneered an algorithm that provides transaction‐level data on the OTC interbank lending market. The veracity of the data identified, however, is not well established. Using permutation methods, I estimate an upper bound on the daily false positive rate of this algorithm to be between 10% and 20%. I propose refinements that reduce the bound to 10% or lower with negligible power loss. The results suggest that the inferred prices and quantities of overnight loans provide viable estimates of interbank lending market activity for Canadian data.