Speaker
Description
Keywords
Climate change, adaptation, weather, unhealthy diet, sugary drinks, retail
JEL: D12, Q54, I12
Introduction
Two out of five American adults are obese, driven by excessive calorie intake and a lack of physical activity. Extreme temperatures are becoming more frequent and intense, particularly extreme heat. They directly negatively impact health (Deschenes & Greenstone, AEJ:AE, 2011) and physical activity (Zivin & Neidell, JLE, 2014). However, amid the current syndemic of obesity and climate change, little is known about the effect of extreme temperatures on dietary behaviour, particularly consumers’ beverage choices. Americans are among the highest consumers of sugary drinks globally (Lara-Castor et al., Nature Com., 2023). Physiologically, there is no rationale for consuming sugary drinks to fulfil the body’s increased hydration needs in hot temperatures. High sugar intake contributes to weight gain and chronic diseases and may exacerbate dehydration.
Objectives
This study aims to improve our understanding of how climate change affects dietary behaviour by (1) examining the impact of extreme temperatures on soft drink purchases and (2) investigating potential drivers and the mitigating effect of adaptation to climate. Historically, the economic literature has focused on the effects of weather on economic productivity rather than consumption (Burke et al., Nature, 2015). This paper contributes to recent works studying the impacts on investment decisions and retail sales (Liao et al., JAERE, 2020; Roth Tran, MS, 2023), including the strand investigating psychological mechanisms (Busse et al., QJE, 2015). While the agricultural economics literature has demonstrated the negative impact of climate change on crop yield and food security (Costinot et al., JPE, 2016; Wheeler & von Braun, Science, 2013) and the public health literature has found an association between extreme heat and highly processed food purchases (Lopez-Olmedo et al., FN, 2021), this study provides the first evidence of a causal effect on dietary behaviour using detailed longitudinal data and explores potential drivers. It also contributes to a large body of work investigating adaptation to climate, including on household spending (Lai et al., Nature HB, 2022), agricultural outcomes (Burke & Emerick, AEJ:EP, 2016), and health outcomes (Carleton et al., QJE, 2022).
Methods
The analysis is based on household purchase data from the nationally representative NielsenIQ Consumer Panel 2011-2019 matched with meteorological data from the National Oceanic and Atmospheric Administration. Zip code-level daily inverse-distance weighted average of the (up to) five closest weather stations within a maximum radius of 300km to the centroid geolocation are computed for each weather variable (Barreca et al., JPE, 2016). The sample size is 3,933,041 household-month observations, including data from 94,738 unique households in 11,954 zip codes across the continental US.
The main specification consists of Poisson pseudo maximum likelihood household-level multi-way panel fixed effect regressions. We exploit the randomness of weather shocks after controlling for seasonality and location (Dell, Jones & Olken, JEL, 2014). The dependent variable is the monthly per capita volume purchased and the main independent variables consist of a series of 16 five-degree daily maximum temperature bins ranging from below 25°F (-4°C) to over 95°F (35°C). Household, county x month-of-the-year, and year x quarter-of-the-year fixed effects are included, as well as additional controls for other weather characteristics and time-varying household characteristics.
Various potential drivers and modifiers are investigated. Further, we test for decreasing marginal effects with short-run repeated exposure to extreme temperatures using fourth-order polynomials of the independent variables. Finally, we investigate long-run climate adaptation by historical experience by interacting the occurrence of extreme temperature days with their zip code-specific historical probability (based on 1980-2010) as a proxy for unobserved adaptation investments (Roth Tran, MS, 2023).
Findings
Volume sold rises non-linearly with temperatures with positive effects above 80°F (27°C). An extra day with a maximum temperature above 95°F (35°C) increases the average monthly per capita purchased volume of sugary drinks by 0.3% and bottled water by 0.6%. Temperatures do not affect diet drink purchases (containing non-sugar sweeteners). Extreme cold days have minor and non-statistically significant effects. The immediate effect of extreme heat is persistent and not offset by inter-temporal shifts.
Extreme heat days have a minor impact on the monthly frequency of shopping trips (-0.05%). Effects on volume are driven by trips to convenience stores, which tend to be closer to household locations and display more unhealthy beverage options. We do not evidence inter-channel substitutions with the on-trade sector (e.g., bars). Using the NielsenIQ Retail Scanner dataset 2011-2012 and Fisher price indices for 32,742 stores across the country, we observe limited price adjustments to temperatures by retailers (-0.06% to 0.05%), in line with Gagnon and López-Salido (JEEA, 2020). Results marginally support salience as a psychological mechanism based on temperature shocks relative to zip code-specific historical weather.
Rural households’ sugary drink purchases are more sensitive to extreme heat, but we find no differences by income level. Working in an outdoor occupation amplifies the effect, likely due to higher exposure and lower adaptation capacity. Other potential unobserved drivers may include heat-induced mood change or stress and advertising adjustments.
We observe constant marginal effects for repeated extreme heat exposure in the same month for sugary drinks but decreasing marginal effects for bottled water. While the human body can acclimate to repeated heat exposure (Sexton et al., JAERE, 2022), our results may highlight persistent heat-induced craving for sugar. In line with Roth Tran (MS, 2023), we find evidence of long-run adaptation with historical cold exposure but not heat exposure, except for bottled water.
Applying these results to downscaled daily climate predictions from NASA Earth Exchange, we plan to estimate the distributional impact of climate change on sugary drink intake up to 2100 by US climate regions.
Conclusion
Future increased extreme heat frequency is expected to stimulate sugary drink intake, with negative implications for public health. Households with higher exposure to extreme heat are more vulnerable. Acclimatization and long-run adaption by historical experience have only a limited moderating effect. These findings can inform policymaking to promote healthier diets under climate change, particularly in settings grappling with an obesity epidemic.
