Kerosene Up, Petrol Down: Demand and Distribution after Madagascar’s Thirty-Month Fuel Price Freeze

Between July 2022 and December 2024, the administered retail prices of the three main petroleum products sold in Madagascar were held constant in ariary for 30 months. In January 2025, the government resumed monthly adjustment. The resulting price shocks went in opposite directions across products: by September 2025, petrol had fallen by 13.7 percent and diesel by 7.6 percent, while kerosene, the lighting fuel of poor households, had risen by 39.1 percent. We use this configuration to estimate the response of deliveries to relative prices on a monthly product panel that we construct from successive issues of the national statistical office’s economic bulletin, covering October 2021 to September 2025, and we are explicit about how far the design can carry a causal reading. The estimates are consistent in sign and order of magnitude across specifications: the own-price elasticity is −0.84 in first differences, −0.51 in levels with product-specific trends, and −1.52 for kerosene taken alone. We test each estimate against an exact randomization distribution built by displacing the price path in time, a device that never treats the three products as interchangeable. Two of the five estimates clear that test at its attainable floor of p = 0.021; the pooled levels estimate does not, at p = 0.38. With two price events, the design has limited power, and we report failures alongside successes. A distributed-lag specification shows that part of the monthly response is undone within two months, so the monthly elasticity is an upper bound on the sustained response; on quarterly aggregates the elasticity falls to −0.45 and loses significance. Kerosene deliveries fell 17.9 percent. Household survey data collected between May 2021 and April 2022, the Enquête Permanente auprès des Ménages of 2021-2022, published in March 2024, show that 42.6 percent of the poorest quintile and 45.3 percent of the second quintile light their homes with kerosene lamps, compared with 10.7 percent of the richest quintile. That gradient would have to close by three-quarters before the poorest quintile ceased to be the more exposed group, so the distributional conclusion does not depend on the elasticity estimates. The reform raised the price of fuel for the poor and lowered the price of fuel for the rich.