{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,17]],"date-time":"2026-10-17T06:41:27Z","timestamp":1792219287108,"version":"4.4.0"},"posted":{"date-parts":[[2026]]},"group-title":"SSRN","reference-count":0,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2026,1,1]],"date-time":"2026-01-01T00:00:00Z","timestamp":1767225600000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/publicdomain\/zero\/1.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:p>A credit model can look smarter simply because a new feature tells it when a loan was made. Banks add outside information, such as the tone of central bank statements, to models that predict loan default, and regulators expect the added complexity to earn its place through better performance. This paper shows that the usual evidence can mislead: a feature that says nothing about borrowers can still raise a model's accuracy or rank high in an importance report, because a series that changes over time reveals the period in which a loan was made. To our knowledge, no accepted test separates real information from this timing effect. We propose a placebo test: compare the real feature with random series built to mimic it, control for the period of lending, and test at the level of each central bank statement. We apply it to European Central Bank and Federal Open Market Committee statements, using more than half a million peer-to-peer loans from Lending Club and Bondora. Without period controls, all 360 random series passed the usual check of improved accuracy; with quarterly controls, about one percent did. Once the period is controlled, no tone effect is detectable, although this pilot has limited power. We outline the same test on supervisory loan data for corporate and small business lending.<\/jats:p>","DOI":"10.2139\/ssrn.7622180","type":"posted-content","created":{"date-parts":[[2026,10,17]],"date-time":"2026-10-17T05:53:31Z","timestamp":1792216411000},"source":"Crossref","is-referenced-by-count":0,"title":["Explainable Machine Learning for Basel and IFRS 9 Credit Risk: A Placebo-Controlled Test of Narrative Features"],"prefix":"10.2139","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-3496-5746","authenticated-orcid":true,"given":"Nawaz","family":"Muhammad","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","container-title":[],"original-title":[],"deposited":{"date-parts":[[2026,10,17]],"date-time":"2026-10-17T05:53:31Z","timestamp":1792216411000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ssrn.com\/abstract=7622180"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":0,"URL":"https:\/\/doi.org\/10.2139\/ssrn.7622180","relation":{},"subject":[],"published":{"date-parts":[[2026]]},"subtype":"preprint"}}