There’s a story about Japan’s lost decades that most people already know by heart: productivity stayed flat, and stagnant wages simply followed. It’s a tidy story, and it’s wrong, or at least badly incomplete. A Bank of Japan working paper published this month1 goes back through decades of worker-level microdata and finds that Japanese firms kept getting more productive through the deflationary period. What broke wasn’t the engine. It was the transmission — the mechanism that was supposed to carry productivity gains into paychecks.

The freeze that outlasted the reason for freezing

The paper’s account of the mechanism is almost mundane, which is what makes it convincing. During Japan’s deflationary decades, firms that expected they could never cut nominal wages later — because cutting pay is politically and socially costly even when a firm badly needs to — responded by refusing to raise wages now. Better to hold the line in a good year than to be stuck defending a raise you can’t walk back in a bad one. This is a textbook case of downward nominal wage rigidity, but the paper’s contribution is showing what that rigidity does when it collides with sustained deflation: it doesn’t just prevent wage cuts, it starts suppressing wage increases too, since any increase might eventually need to be reversed.

According to the paper, the share of workers facing this kind of upward wage rigidity rose from under 1% in the early 1990s to roughly 7% by the early 2000s. That’s a large jump for a mechanism that, in a normal inflationary environment, barely exists. Under 2% inflation, a firm can quietly erode real wages just by holding nominal pay flat — no confrontation required. Under deflation, that escape hatch closes. Nominal wages become the only lever, and firms that fear ever having to pull it downward stop touching it at all.

Why this is two problems, not one

Here’s the part I found more interesting than the headline finding. The paper doesn’t stop at “wages grew more slowly than they should have.” It argues the wage freeze did a second kind of damage: it dulled the signal that’s supposed to move workers toward more productive firms. If wages stop reflecting a firm’s actual productivity — because everyone is frozen at roughly the same nominal level regardless of how well the firm is doing — then workers lose the main piece of information they’d use to decide whether to switch jobs. The paper reports that the elasticity of job-to-job separations with respect to relative wages fell from about 1.6 in 1997 to about 1.4 by 2003, and that as of the paper’s most recent data, it still hadn’t recovered.

That second number is the one I keep coming back to. Inflation in Japan has been running near or above 2% for a few years now. The precautionary logic that caused the freeze — “don’t raise wages you might have to cut” — should be losing its grip, since deflation is no longer the backdrop that made the fear rational. And yet, per the paper’s counterfactual exercise, restoring pre-deflation wage-setting norms would still raise real wages by roughly 10% and aggregate productivity by close to 0.75%, gains that haven’t materialized despite the return of positive inflation. The price environment healed years ago. The behavioral norm it produced apparently didn’t.

Two kinds of stuck

I think what makes this paper worth sitting with is that it separates two things that are usually talked about as one: the external condition (deflation) and the internal habit it produced (treat any wage increase as a commitment you might regret). The condition can end on a schedule set by monetary policy. The habit doesn’t have a schedule. It persists because it’s now embedded in how HR departments budget, how unions negotiate, how workers themselves calibrate what a “reasonable ask” looks like. Nobody has to keep believing deflation is coming back for the behavior to keep replicating itself — it just has to be what everyone still does by default.

This is a familiar shape to anyone who’s watched an organization outgrow a policy but keep the reflex. A hiring freeze ends, and headcount requests stay conservative for another two years because that’s just how planning got done during the freeze. A safety incident triggers a review process, and the process outlives every person who remembers why it exists. The interesting question is never whether the original cause has gone away — it usually has, quietly, without anyone announcing it — but whether anything exists to force a reassessment. If wage-setting is decentralized across thousands of individual firms, each one just doing what worked last year, there’s no single moment where someone asks “wait, are we still doing this because it’s still true?”

I don’t think the paper fully answers what would force that reassessment in Japan’s case, and I’m not sure it’s supposed to — it’s a diagnostic paper, not a policy paper. But it does leave the diagnosis in an unusually sharp form: the thing that needs fixing is not a number (inflation, productivity) that policy can move directly. It’s a behavioral equilibrium that a decade of a different number happened to produce, and that a few years of a corrected number hasn’t yet been enough to unwind.

An open question

What I don’t have a settled view on is how much of this generalizes. Downward wage rigidity is well documented outside Japan too, but the “elasticity that doesn’t bounce back” finding is more specific — it implies that a sufficiently long deflationary period can leave a lasting scar on how a labor market allocates people, not just on how it prices them. If that’s right, then a central bank’s job during a deflationary episode isn’t only to get prices moving again. It’s to get prices moving again before the habits built to survive deflation calcify into something inflation alone can’t dissolve. Whether that’s a matter of years, as this paper’s data might suggest, or something firms and workers eventually correct for on their own once enough job-switchers get burned by staying too long, is a question I don’t think the data can answer yet — it just hasn’t been long enough since inflation returned to tell the difference.

  1. Bank of Japan Research Laboratory, “What Prevents Productivity Gains from Translating into Wages: Lessons from Japan’s Deflationary Period”, BOJ Working Paper No. 26-E-12 (2026-08-14). Accessed 2026-08-20. ↩