Adding Nasdaq inputs raised weekly Nikkei return forecast error by 2.5% in a 2018–2025 archive test

We tested whether past Nasdaq Composite inputs improved next-week Nikkei 225 price return forecasts after past Nikkei inputs. Each forecast used dated ALFRED records and past labels. Fixed ridge models were fitted again each week. In 399 weeks with recent inputs in 2018–2025, the added Nasdaq terms raised mean squared error by 2.54%. The paired 95% eight-week block interval was a 0.88% to 4.48% increase. The larger model also had 4.48% more error than a past return mean. An earlier archive cutoff gave the same conclusion. A secondary return variation gain was uncertain. An audit found and corrected a calendar error from 48 holiday dates that had prices for only one archive day. We preserve the initial and corrected results, all forecasts, model states, source provenance, code, figures and repeat instructions. The test covers delayed weekly forecasts for two price indexes. It does not test immediate market responses or establish trading profit.

Nikkei 225 · Nasdaq Composite · return prediction · dated archives · chronological test · negative result · cross-market forecasts · data revisions