Case study

Case study: the summer 2018 beer surge

England's June–August mean temperature went from 16.0°C to 17.1°C (1.4°C above the 1991–2020 normal, the second-warmest such summer in a 142-year record) while beer's price relative to the rest of the shopping basket fell from 101.2 to 99.4. Outcome: real price −1.8%, clearances +6.8% (165.4 to 176.6 thousand hectoliters of pure alcohol a month), of which the model attributes +3.8 points to price, +2.9 to retail activity, +1.2 to the heat, −3.5 to structural decline, and +2.1 to nothing it can name. Two curves, both moving, neither moving much. The flat supply line sits 1.8% lower in 2018: beer's CPI rose 0.7% while all-items CPI rose 2.5%, so beer got cheaper in real terms without any duty cut. At an estimated demand elasticity of −2.14 that slide along the curve is worth +3.8% of volume; the remaining +2.8% is the demand curve itself shifting right, mostly retail activity and the heat. HMRC clearances are national only, so each region's figure is the national temperature response applied to that region's own warming and its share of the market. No regional beer series enters it, and the regional panel does not identify a separate regional temperature response (−0.101, se 0.255). Shares are Defra Family Food purchases × population, pooled over 2017/18 and 2018/19. The heat does about a fifth of the work here, not the headline: +1.2 of the +6.6 log points, on a temperature coefficient of 1.07% per °C. That coefficient is significant (t 2.9), but because it is linear in a monthly mean it is still the wrong functional form for a heatwave, so read it as a floor rather than a point. Nor is this a tax story: beer duty was not cut, the statutory rate was frozen through 2018, and the real price fall is simply 2.5% inflation outrunning a flat duty. The price channel rests on an elasticity of −2.14 whose weak-instrument-robust interval runs from −0.93 to −3.84; at the shallow end the price contribution is +1.7 rather than +3.8, and the unexplained residual roughly doubles. These charts apply the model's estimated coefficients to a single summer, to show where its +6.6% came from. The coefficients themselves were measured on the full 1999–2025 sample, not on this episode alone.

Trust, but verify

How the estimate compares

Most published estimates of beer demand come from household surveys. This one uses the volumes HMRC released for duty.

Published UK and meta-analysis estimates of beer's own-price elasticity. Dots are point estimates, bars are reported ranges or on/off-trade splits. Huang (2003) is the only other estimate built on clearances rather than household surveys. The bar on this estimate is the weak-instrument-robust interval, wider than the conventional one. HMRC reports beer twice: as liquid, and as the pure alcohol it contains. The same model on each, over the same months, splits the response. Volume falls at −1.43, alcohol at −2.01. The 0.58 gap is weaker beer rather than less beer (average strength fell 4.53% to 4.22% ABV), which the 2023 duty reform was designed to encourage. Bars are 95% intervals; the pints response is not distinguishable from zero at its shallow end. The headline uses the alcohol basis, the only measure that continues past 2023.

Sources: HMRC (Alcohol Bulletin: duty clearances and receipts); ONS (consumer prices, retail sales index, ITL1 boundaries); Met Office (Central England Temperature and HadUK-Grid regional series); Defra (Family Food). Contains public sector information licensed under the Open Government Licence v3.0.