Tempers Rising: The Effect of Heat on Spite
The relationship between heat and harmful outcomes is well documented, with research connecting various adverse economic outcomes to the climate. In the presence of increasing global warming and climate change, understanding why the climate leads to negative economic outcomes is essential for forming peaceful institutions of the future. We study how behavioral economic outcomes change in the presence of heat through a lab experiment involving 1,110 observations conducted in five different countries. This paper specifically focuses on the social preference outcome of spite. We find that increased time exposure to the treatment effect of heat is required to elicit an individual’s spiteful behavior. Our results also suggest heterogeneity in this effect with a particular difference along gender and income consistency. We deploy novel methods to analyze heterogeneity using a machine-learning causal forest and Sorted Group Average Treatment Effect (GATES).
1 Introduction
Climate and economic outcomes have been at the forefront of research and policy concerns in recent decades as global warming predictions materialize (Bathiany et al. 2018). Compounding current-day climate volatility with scientists’ predictions that temperatures will continue to rise in decades to come makes understanding climate and its effect on human behavior a crucial concern for policymakers when designing future institutions. Although climate concerns have recently elevated their importance in modern science, the role that climate plays in shaping societal outcomes has had a long history of sparking researchers’ interest (Almås et al. 2019).
The large body of research connecting climate with behavioral outcomes overwhelmingly shows a link between increased temperatures and detrimental socioeconomic outcomes (Hsiang, Burke, and Miguel 2013). Research has frequently shown evidence of the relationship between hot climates and increased group-level and individual-level conflict (Burke, Hsiang, and Miguel 2015). Heat’s linkage to increased conflict and aggression is supported through substantial research mapping heat as a causal force driving harmful societal outcomes. Recent studies have mapped heat to increased crime rates, hostility towards others, civil conflict, political unrest, and more (Hsiang, Burke, and Miguel 2013). Temperatures’ connection to harmful outcomes is well established, yet the mechanisms behind how warmer climates are causing detrimental outcomes are still unknown.
Climate can affect societal outcomes in three ways (Falk et al. 2018). The first effect the literature defines is known as the direct effect. The direct effect describes the process humans directly undergo when the temperature rises; people may get uncomfortable, more irritable, or lethargic (Bushman and Anderson 2020). The direct effect is often difficult to pin down because detrimental outcomes correlated with hot climates are endogenous with many other variables that the climate affects. For example, hot temperatures are also responsible for droughts, volatile crop yields, and extreme weather events (Arnell et al. 2019). These outside effects of heat can then cause outcomes of conflict which are what the literature defines as indirect effects. Research frequently shows indirect effects being at least partially responsible for poverty, instability, and migration outcomes. This paper takes advantage of a lab experiment’s ability to block out the indirect effects of hot temperatures on behavioral outcomes. Our experiment will exogenous vary room temperatures to look at different behavioral outcomes. We will also introduce a competition within the experiment to examine the heterogeneous treatment effects between winners and losers.
Research suggests that the causal mechanism behind heat’s relationship to harmful outcomes may be two-fold. First, rising temperatures cause people to become irritable, uncomfortable, or upset, and when combined with a provoking event, an individual is more likely to react anti-socially. Studies have found evidence of this interaction in sporting events (Larrick et al. 2011), parking lots (Kenrick and MacFarlane), and social media (Baylis et al., 2018). This experiment pulls inspiration from a similar lab experiment from (Almås et al. 2019), where the temperature was controlled in a lab setting. They found little evidence of heats direct effect on social preferences, except for a significant treatment effect for a sub population of individuals from Kenya that identified as belonging to an ethnic group that had been politically marginalized in a recent national election. The heterogeneity found in this subpopulation in Kenya supports the theory that heat requires a provoking interaction to produce harmful outcomes. This paper will add to the literature in three ways:
We conducted a randomized control experiment across five countries, collecting over 1,100 observations and measuring various behavioral economic outcomes.
We will analyze the treatment effect on the anti-social outcome of spitefulness.
We will test out novel machine learning methods to compare there results wtih traditional methods for examining heterogeneous treatment effects.
Spitefulness will be the central behavioral outcome that this paper examines. Spite comes from the root word despite and is defined by many behavioral economists as any action that causes harm to others without having any benefit to the actor themselves (Fehr, Glätzle-Rützler, and Sutter 2013). Our experiment will measure spitefulness through four rounds of a single shot, anonymous dictator games, which are commonly used in economic literature to measure individuals’ social preferences. We find that 7% of the sample act in a way categorized as “Strong Spite” and 24% of the sample categorized as “Weak Spite” based on their dictator game answers. We look at both categories as separate outcome variables with our primary analysis on the Strong Spite outcome.
We find that the treatment effect of heat has a negligible effect on individuals’ probability of acting spitefully for both outcomes, with a coefficient less than 0.05 and an insignificant p-value, however we do find significant heterogenaity within the treatment effect with specific subpopulations displaying a statistically significant treatment effect. We find a significant positive treatment effect for individuals who played version B of the economic experiment. Version B differed from version A in two ways. First that individuals played a competition round immediately before answering the dictator games, where spiteful behavior was measured. The second difference between the two versions is that version B exposed individuals to the temperature of the room for 15 minutes before they answered the dictator games compared to 5 minutes in version A. Our results suggest that this increased exposure to high temperatures may be creating a significant treatment effect of temperature on spite. We find that for individuals playing version B and in room over 24 wet bulb Celsius, they are 9% more likely to respond spitefully with a significance at the below the 5% level.
Furthermore we find that there is heterogeneity in this later treatment effect of individuals that playued version B, among gender and income consistency. We uncover our findings through numorus methods including traditional OLS and have these findings tested through novel methods of a machine learning methods of a causal forest (Wager and Athey 2018) and generic machine learning (Chernozhukov et al. 2018). The techniques of a causal forest and generic machine learning have been leading the way in the recent research in estimating heterogeneous treatment effects. Our findings can contribute to the literature by provide insight into for whom does having increased exposure to high temperatures cause anti-social outcomes.
It is important to consider our results’ external validity in the context of the more prominent topic of climate change’s impact on human behavior. Two limitations to keep in mind from our lab experiment are one, the lack of variation in the age and occupation of our participants due to the fact that all experiments where conducted on university campuses. Secondly, our experimental environment was all indoors, where temperature variation was easy to manipulate however the effects of climate change will predominantly be experienced outdoors. Our findings most closely generalize to situations with hot indoor environments; however, these results can still provide evidence to the broader concern of how climate change will change our behaviors.