Social and behavioral determinants of indoor temperatures in air-conditioned homes

The causes and consequences of indoor heat exposure are receiving growing attention as global temperature rises and people seek respite from the heat in indoor spaces. In this study, we measured indoor temperatures of 46 air-conditioned residences in Phoenix, Arizona, United States. Temperatures were collected concurrently at 5-min intervals from August 21 to September 19, 2016. Indoor temperatures exhibited significant heterogeneity across all residences, ranging from 16.5 to 37.2 °C with a mean (SD) of 26.4 °C (2.2 °C). On average, the 5-min indoor temperatures were moderately correlated with outdoor temperature (r = 0.421), although individual household correlations were highly variable, ranging from r = -0.244 to r = 0.924. Households were grouped into six clusters using K-means based on 19 temperature metrics. We tested for differences in demographic, behavior, and infrastructure indicators between those six clusters based on responses to a social survey. Nearly half the variance in preferred thermostat setting was explained by cluster (R² = 0.455, p < .001). For the most part, measures of air-conditioning use, limitations on air-conditioning use, and household resources (e.g., income) did not vary significantly by cluster. The same was true for heat-related health and comfort outcomes. Two households that did not pay their own electric bill were by far the coldest homes (average temperature of 20.0 °C). We conclude that indoor temperature preference may supersede concerns related to the cost of using air-conditioning and that resource-constrained households may be sacrificing other necessities to keep their homes comfortable.