UC Berkeley

This dissertation consists of two chapters that examine the interaction between people and the environment. The choices individuals make are not only dependent on environmental conditions, but also change the environment we live in. Specifically I investigate how the weather affects our behavior, and how our transportation decisions affect air pollution. The first chapter investigates the relationship between weather and exercise, and ultimately how weather impacts habit formation. The second chapter studies the effectiveness of public transportation in reducing pollution due to personal transportation choices.

The first chapter, “Weather, Exercise and Habit Formation: Evidence from GPS Activity Tracking Data”, investigates how individuals choices are affected by the environmental conditions in which they live. Many people fail to meet physician recommendations for exercise despite the well established health benefits. While running and cycling are popular forms of exercise, they are subject to weather variations, and hence climate change has the potential to alter exercise patterns. I use a new panel data set of individual exercise records from GPS activity tracking to estimate the daily non-linear weather-exercise response function on the extensive and intensive margins of exercise for a sample of users in the United States. I estimate a semi-parametric bin estimator of weather on daily exercise controlling for region-by-month and region-by-year time dummies, day of the week, and individual fixed effects for both an indicator of exercise and the time recording exercise during the day. The results exhibit an inverted U-shape where the extremes of high and low temperatures decrease the likelihood of exercise on a given day by up to six percentage points compared to the most likely temperature that individuals will exercise. Precipitation also decreases the likelihood of exercise with an increasing effect for greater amounts, up to 3.6 percentage points lower for daily precipitation over 30mm compared to a day without rain. Substitution to indoor activities is minimal, with reductions in outdoor activity contributing the majority of the weather-exercise response. Building on these results and using temperature as an instrument, I estimate a monthly model of habit formation and find that a one-minute increase in exercise in the previous month leads to a 0.36-minute increase in the current month.

The second chapter, “The Impact of Public Transportation on Air Pollution: Evidence from a Transit Strike”, investigates how public policy can affect individual choices that have an impact on the local environment. While air pollution in the United States has improved since the 1960s, a number of locations still fail to meet environmental standards. Public transportation has been proposed as one method to reduce air pollution in cities. To test this I use a 2003 strike by mechanics and transit operators of the Los Angeles Metropolitan Transit Agency (LACMTA), which is the primary public transportation provider within the City of Los Angeles. During the strike, all buses and subways operated by the LACMTA in the urban areas of Los Angeles county stopped running for 35 days in October and November. Using data from air monitoring stations within the City of Los Angeles and in neighboring areas for the main gases emitted by road vehicles, I find that the daily maximum level of carbon monoxide increased by 0.285 ppm during the strike. While it is possible that daily maximum levels of nitrogen dioxide also increased during the strike, the results are not statistically significant. Ground-level ozone is not directly emitted by road vehicles, but instead results from atmospheric reactions of pollutants from many sources including transportation. However, there is no evidence for an increase in ozone levels during the strike. Increases in air pollution are limited to the area with reduced public transportation service during the strike, and therefore any benefits to reduced air pollution during regular do not appear to extend to neighboring areas. These results are consistent with the previous literature, which found the 2003 strike increased peak period commuting congestion, but increased congestion was limited to Los Angeles County.