Weather and children's time allocation

Abstract

This paper presents the first causal estimates of the effect of weather on children's time allocation. It exploits exogenous variations in local weather observed during the random diary dates of two nationally representative cohorts of Australian children whose time-use diaries were surveyed biennially over 10 years. Unfavorable weather conditions, as represented by cold or hot temperature or rain, cause children to switch activities from outdoors to indoors, mainly by reducing the time allocated to active pursuits and travel and increasing the time allocated to media. Furthermore, the effects of bad weather are more pronounced on weekends and for children with asthma. Our results also provide some evidence of adaptation, as temperature tends to have greater impact not only in winter months but also in colder regions. Our findings are robust to a wide range of sensitivity checks, including controlling for individual fixed effects and using alternative model specifications. Overall, the results suggest that extreme weather conditions may diminish children's health, development and long-term achievements through their effects on children's time allocation.

Causal Impact of Physical Activity on Child Health and Development

“…As mentioned above, IV regressions may provide inaccurate estimates because they don't control for time invariant unobservable factors which may be associated with the instruments and outcomes at the same time.20 Fist-stage regression results from IV and FE-IV estimators are reported in column 1 and 2, respectively, of Appendix TableA7. The results are largely in line with those documented in other studies(Nguyen et al 2021b;Nguyen et al 2022). For instance, lower temperature or precipitation statistically significantly decreases the time spent on physical activity.…”

“…We Specifically, prior studies have shown that the allocation of children's time to physical activities is particularly sensitive to both the local daily maximum temperature and precipitation (Nguyen et al 2021a;Nguyen et al 2021b). 12 These instruments are also theoretically sound as plausibly exogenous local weather conditions directly affect individual's physical activity, but only indirectly affect their development outcomes through the physical activity channel.…”

“…The remaining 39% of TUDs were filled in roughly 1.4 days after the survey date. The difference between the dates of measured physical activity and those of measured outcomes, when viewed with evidence that physical activity is only affected by weather conditions recorded on (not before or after) the date of TUD (Nguyen et al 2021b), further alleviates any concern that our instruments may affect outcomes directly.…”

“…We first test the robustness of the results to a different functional form of the temperature instrumental variable. In particular, we include daily maximum temperature and its square, together with the daily precipitation, as instrumental variables in equation ( 2) to capture the potential non-linear impact of temperature on physically active time, as found in the literature (Nguyen et al 2021a;Nguyen et al 2021b). 23 The results, reported in Panel B1 of Appendix…”

“…For example, it turns the estimates of physical activity on In addition, we test whether unobserved time-variant factors may account for the impact of physical activity on development by controlling for some important time-variant variables which are potentially associated with our instruments and developmental outcomes at the same time. Specifically, in light of previous findings, including a statistically significant impact of local weather on children's daily time spent on media and travel activities(Nguyen et al 2021b) and some findings of beneficial impacts of media time on child development(Gentzkow & Shapiro 2008;Nieto & Suhrcke 2021), we separately include the individual's daily time spent on media and travel as additional explanatory variables.Estimation results reported in Appendix TableA9(Panels C1 to C3) show that either excluding or controlling for all of the above-mentioned time-varying variables does not change our results in any significant way. Results from these sensitivity checks, when viewed with results of a Sargan-Hansen test about the exogeneity of our instruments reported at the bottom of each panel in Appendix TableA9, suggest that our instruments are not correlated with time-variant unobservable characteristics in the outcome equations Park et al 2020),.…”