Cyclical Activity and Inflation Under Secular Stagnation: Empirical Evidence Using Data on Japan's Lost Decades

ABSTRACT This paper re‐examines whether the short‐run inflation‐activity trade‐off remains stable during prolonged secular stagnation, using Japan's ‘lost decades’ as a laboratory. Standard slack measures (e.g., unemployment gaps or detrended unemployment) can embed persistent low‐frequency movements that drift over time, obscuring the underlying Phillips‐curve relationship. To address this measurement problem, we construct a cyclical activity index that isolates business‐cycle‐frequency comovement across a broad set of labour‐market and real‐activity indicators. Using monthly Japanese data from 1983 to 2025, we estimate a short‐run Phillips curve and find a negative, economically meaningful relationship between cyclical activity and short‐term inflation that is stable across subsamples; structural break tests fail to reject the stability hypothesis. In contrast, conventional slack measures deliver weaker and less stable estimates. An instrumental variable approach that uses the cyclical activity index to instrument the unemployment gap yields a larger and more stable slope, consistent with low‐frequency contamination or measurement error in conventional slack. Results are robust to alternative index constructions, inflation measures and expected‐inflation controls, including moving‐average proxies and survey expectations at different maturities. Overall, demand conditions continue to matter for inflation under secular stagnation, but their influence can be obscured by the way slack is measured.