Winter Binds, Summer Does Not: The Heat-or-Eat Trade-off across the Income Distribution in South Korea

Technical Summary: Winter Binds, Summer Does Not: The Heat-or-Eat Trade-off across the Income Distribution in South Korea

Problem Statement
This study addresses the unequal distribution of economic pressure arising from climate variability, specifically examining how households reallocate budgets between energy and food in response to seasonal temperature changes. While existing literature acknowledges that rising energy costs can force trade-offs with other necessities, there is a gap in understanding how these trade-offs vary across the income distribution and between seasons. Prior research often relies on annual averages or focuses solely on mean temperature levels, potentially missing the non-linear effects of temperature variability and the specific "heat-or-eat" dilemma faced by low-income households during winter. The paper investigates whether a "cool-or-eat" trade-off exists in summer to provide a symmetric contrast, and whether the interaction between temperature levels and variability creates threshold effects that linear models fail to capture.

Methodology
The authors utilize nationally representative panel data from the Korean Labor and Income Panel Study (KLIPS) covering the years 2017 to 2021. The dataset includes approximately 5,000 urban households, with expenditure data categorized into home energy (electricity, gas, water) and food (at-home and dining out).

• Temperature Variables: Daily mean temperature data from the Korea Meteorological Administration (KMA) are aggregated at the district (si-gun-gu) level. The study constructs two moments for both summer (June–September) and winter (November–February): the mean and the variance. Following Coibion et al. (2024), these moments are calculated using a triangular weighting scheme that places greater emphasis on the tails of the distribution (extreme heatwaves and cold spells). Winter mean temperatures are multiplied by −1 to facilitate interpretation (where larger positive values indicate colder conditions).
• Empirical Model: The authors employ a survey-weighted fractional probit model (quasi-binomial GLM with a probit link) to estimate the effects of temperature on the budget share of energy and food expenditures. The dependent variable is the share of total household expenditure allocated to a specific category, bounded between 0 and 1.
• Specification: The model includes interaction terms between the mean and variance of temperature for both seasons. It controls for household income, price indices for energy and food (using first differences to capture short-term changes), sociodemographic factors (age, household size, education, tenure status), and year and region fixed effects.
• Inference: Standard errors are clustered at the district level (si-gun-gu), as temperature regressors vary only at this level. Robustness checks include a within-household specification (with household and year fixed effects) and a wild cluster bootstrap at the provincial (si-do) level.

Key Findings
The analysis reveals a robust, income-differentiated, and seasonally asymmetric pattern of expenditure reallocation:

1. Winter "Heat-or-Eat" Trade-off:

   • Low-Income Households: As winter temperatures drop (and variability increases), lower-income households significantly increase their energy budget share while simultaneously cutting their food budget share. Specifically, a one-unit increase in the sign-reversed winter mean temperature leads to a 0.039 increase in the energy share and a 0.020 decrease in the food share for households below the Government Poverty Line (GPL).
   • High-Income Households: Higher-income households also increase energy spending in colder winters but show no statistically significant reduction in food spending. This indicates that economic resources allow them to absorb heating costs without compromising nutrition.
   • Threshold Effect: The interaction between winter mean and variance reveals a non-linear threshold. For low-income households, as winter conditions become extremely severe (high variability and low mean), the ability to further increase heating spending diminishes, and the reduction in food spending halts. This suggests a "floor" under minimal nutrition is preserved even when heating needs are not fully met.

2. Summer Asymmetry:

   • In contrast to winter, no comparable "cool-or-eat" trade-off is observed in summer. For both income groups, higher summer temperatures do not force a statistically significant reallocation away from food to fund cooling. The summer results serve as a control, confirming that the winter trade-off is a specific response to heating constraints rather than a generic reaction to temperature stress.

3. Combined Effects:

   • Under the most extreme winter conditions, low-income households exhibit a significant energy increase but a food reduction that shrinks toward zero, reinforcing the existence of a threshold where further nutritional sacrifice is no longer feasible. High-income households maintain stable food spending regardless of winter severity.

Key Contributions
The paper makes four specific contributions to the literature:

1. Explicit Trade-off Analysis: It treats the energy–food margin as the primary object of analysis, demonstrating that the "heat-or-eat" dilemma is real but strictly confined to the winter season in the South Korean context.
2. Joint Modeling of Moments: By modeling the mean and variance of seasonal temperature jointly, the study captures the role of temperature variability, which previous literature often overlooked or treated separately.
3. Non-Linear Thresholds: The estimation of mean-by-variance interactions uncovers a threshold effect where extreme winters cap the heating response and halt further food cuts among the poor—a nuance missed by linear, mean-only models.
4. Consumption Inequality Perspective: The study frames the energy–food trade-off as a dimension of consumption inequality, showing that identical weather conditions produce binding trade-offs for poor households but not for richer ones.

Significance and Policy Implications
The authors conclude that the "heat-or-eat" squeeze is a recurring, climate-driven form of consumption inequality that annual-average welfare measures fail to capture. The findings suggest that policy interventions should be season-specific, prioritizing direct winter heating and food assistance for lower-income households rather than applying uniform, year-round approaches.

Furthermore, the paper argues that targeting based solely on income thresholds may be insufficient, as housing energy efficiency plays a critical role in the heating burden. The authors propose that energy and food assistance programs (such as Korea's Energy Voucher and Agri-food Voucher) should be more explicitly coordinated to address the linked nature of these expenditures under cold stress. While the study is modest regarding its generalizability, it suggests that these patterns of income-differentiated winter trade-offs may be relevant to other high-income societies with cold winters and widening consumption inequality.

The study acknowledges limitations, including the reliance on expenditure shares rather than direct health or nutritional outcomes, the weak signal of summer effects, and the focus on temperature rather than other climate risks. Future research is encouraged to integrate health outcomes and evaluate the effectiveness of integrated welfare programs in mitigating these trade-offs.