東京都区部 · veg watch
Monthly averages from a government retail survey (総務省統計局 小売物価統計調査; 2000–2026) sampling representative supermarkets across the Tokyo 23 wards. These track market trends, not individual store prices (use flyers and discount-tracking apps for current weekly deals if you're so inclined). Solid is measured data. The dashed lines are a seasonal-model projection for the rest of 2026 (not official, just an extrapolation!). Trust its shape more than its exact yen level, especially for rice, whose big moves come from supply and policy shocks that a seasonal model can't anticipate.
tokyo_prices.csv yourself with “Load CSV” below.
Rice has two separate spikes, and I believe they're unrelated. The series shows a high in 2000–2001 (around ¥5,900) that fell by more than half by 2002 and then sat near ¥2,300–2,700 for two decades, plus the recent climb from mid-2024 to ¥5,400+. The units are the same ¥/unit throughout. This is a real level, not a denominator change. The old drop appears only in rice (every other item moves normally across those years), which points to a rice-specific market or policy cause rather than a survey change. Quite likely, the driver is Japan's rice-market deregulation phasing in around then, though I haven't confirmed that against the survey's own change-logs. Maybe I'll do that when I have a bit of time. Either way, rice's big moves are supply/policy shocks a seasonal model can't foresee, so trust the projection's direction for rice, not its exact level.
Sudden "steps" shared by several items are usually survey changes, not real jumps. When a clean level-shift lands in the same year across unrelated items, that's the fingerprint of a methodology revision: the government survey periodically changes the exact product it prices for an item (package size, grade, unit), moving the measured level overnight with no real market change. Treat a same-year step shared by several items (green peppers and strawberries are the main ones I can see by eye) as a data artifact; treat a break in one item alone (like rice's) as more likely real.