How Much Demand Response Can a Real Tariff Deliver? Evidence from High-Price Days under the French TEMPO Scheme

Time-varying electricity pricing is a central lever for managing peak demand as grids decarbonise, yet most credible estimates of how much residential demand responds come from small, shortlived pilots. We provide evidence from a tariff that is neither - the French TEMPO scheme, an operational national tariff that for nearly three decades has priced peak hours four to five times higher than standard on up to 22 pre-announced "Red" winter days. Using national daily consumption for the TEMPO-subscriber segment—nearly one million metering points—linked to the grid operator's assignment calendar and forecasts, we find that Red days cut total residential consumption by about 23% relative to comparable non-Red days, peak-hour consumption by roughly 30%, and off-peak consumption by about 13%. The off-peak decline shows the response is genuine curtailment, not merely load-shifting within the day. These effects, drawn from a tariff already in force at scale, sit at or above the upper end of what critical-peak-pricing experiments have recovered from far smaller samples. The central identification challenge is that the operator assigns Red days to the highest-demand days, so raw colour comparisons confound behaviour with weather. We address this by exploiting three features of the assignment rule that move colour status independently of contemporaneous household demand—renewable-driven variation in net load, a threshold paced mechanically by a fixed seasonal quota, and an end-of-season clause that forces the operator to place its remaining Red days regardless of the forecast (13 of 22 Red days in 2025–26 fell in this forced window). Matching comparable days on a forecast-based demand index and letting the binding quota widen the comparison window yields short-run peak elasticities of −0.23 to −0.27 and total-consumption elasticities of −0.28 to −0.32, stable across three estimators that lean on the rule to differing degrees. The results give tariff designers a real-world benchmark for event-based pricing, and the identification strategy—a quota-forcing regression discontinuity—applies wherever a capped allocation must be exhausted on a deadline.