A rise in US jobless claims gave no clear weekly Nasdaq return forecast gain in a 2015–2025 archive test

We tested whether a rise in US four-week initial unemployment insurance claims above their one-year low improves next-week Nasdaq Composite price-return forecasts. Each claims input uses its own historical ALFRED version and stays fixed in later fits. Weekly models use 260 earlier available return labels. In 550 fresh test weeks from 2015 to 2025, a fixed ridge model reduced mean squared error by 1.10% against the past mean. The paired 95% 13-week-block interval was -0.19% to +2.82%. The test therefore gave no clear forecast gain. The 2020 weeks supplied 95.65% of the net score gain. Without 2020, the reduction was 0.06%, with interval -0.42% to +0.52%. Both main models always predicted a positive return and matched the always-positive direction accuracy of 61.27%. All planned penalty, input-cap, time-period, revision, freshness, and block-length intervals included zero. The source audit preserved one invalid claims sequence and 22 stale forecast cases. A separate checker matched all forecasts and 148,720 training labels. The package preserves code, claims source records, price-source provenance, forecasts, fits, 130,000 uncertainty draws, figures, and negative findings. This is a retrospective test of one index and one claims rule on market history shared with earlier SCI studies. It does not establish reliable stock-return prediction or net trading profit. Raw Nasdaq prices and individual measured returns remain separately retained under provider terms.

stock return forecasts · initial unemployment claims · point-in-time data · ALFRED · chronological evaluation · negative result · Nasdaq Composite