The Economics of Traffic Congestion: A Quantitative Analysis

University of Minnesota Ph.D. dissertation. May 2021. Major: Economics. Advisor: Thomas Holmes. 1 computer file (PDF); xi, 116 pages.In this dissertation I explore the effect that traffic congestion has in shaping commutingchoices and the distribution of economic activity and people in cities. In the rest two chapters I set up a framework to evaluate the welfare impact of investmentsin the transportation network in a city. Chapter 1 theoretically lays out this framework and chapter 2 applies it to evaluate first, the extension of the Expo rail line in Los Angeles county and second, the value of the entire rail network in Los Angeles county. In the first chapter (1) I develop a structural model of the commuting market in a city. A cityis a collection of residential and working locations connected by a transportation network. The transportation network is composed of the road network and the public transit and street network. The crucial difference between the road network and the other two is that the first can be congested as a function of usage. Moreover, there is an exogenous demand for travel from residential to employment locations. On the demand side, commuters make mode and routing choices taking as given trip characteristics (travel time and travel money cost). On the supply side, the transportation network maps travel demand into trip characteristics. An equilibrium is reached when, given trip characteristics, commuting mode shares give rise to these trip characteristics and vice versa. To solve this problem I use methods from the civil engineering literature on traffic assignment.There is a plethora of algorithms that allow to solve the routing problem in a congested road network efficiently. I outline an algorithm that allows to solve for the commuting market equilibrium that has nested the traffic assignment problem. The second chapter (2) brings the model developed in the first part to the data and use it to evaluate the effect of the Expo rail line extension and the value of the entire rail network in Los Angeles counties. I start by estimating demand side parameters using a mix of the California Household Travel Survey augmented by requests to Google maps. With parameter estimates I nd that the value of time is $19.81 per hour which is in line with the median hourly wage in Los Angeles of $20.52 per hour. Next, I estimate the parameters of the congestion function using highly disaggregated highway ow and speed data from the California Department of Transportation. My parameter estimates differ from the ones suggested by the Bureau of Public Roads indicating that the Los Angeles county highway system gets congested for lower car volumes than previously thought. Then I assess the accuracy of the model and nd that the model capture accurately the main moments of the data: mode shares and average travel times. In the first counterfactual I evaluate the effect of the Expo rail line extension. This extension connects Santa Monica to Los Angeles and had a cost of $1.5 billion. I nd that this extension increased public transit share by 0.68 percentage points and reduced total travel time by 1,472 hours each 30 minutes. Welfare increased by 0.085% in the county and it takes 6.37 years to recover the investment. In the absence of congestion externalities the effect of the extension is more than double. In the second countefactual I nd that the value of the entire rail system in Los Angeles county is of $1.9 billion. This is in line with other studies that estimate congestion relief of the public transit system in Los Angeles county. The last chapter of this dissertation (3) investigates the relationship between populationdensity and commuting choices motivated by the fact that cities across the United States are considering changes in single-family zoning. The elimination of single-family zoning restrictions will make urban areas more dense and my conjecture is that this will impact the commuting market of these cities. First, I study the relationship between population density and commuting mode choice by means of a reduced form exercise. I nd that a 20% increase in population density is correlated with a decrease between 0.7 and 0.95 percentage points in private car commuting mode. Moreover I nd that this decrease in private car mode share is captured almost entirely by the public transit system. Next, I develop a model of internal city structure with residential and productive amenities, a fixed measure of development land, and endogenous travel costs with commuting mode choices. In short, people want to locate in areas with high amenity levels (agglomeration forces) but at the same time land prices and car travel times increase (dispersion forces). In this model, a change in zoning regulation is equivalent to increasing the amount of urban development land per location. Therefore, this model is able to simulate different zoning policies and capture the relationship between population density and commuting mode choices. Finally, in the last section of this chapter I use the model to simulate how different uniform increases in population density across all locations affect mode choice and travel times in Los Angeles county. I nd that a 20% increase in population density is associated with a decrease of 0.82 percentage points in car commuting share and that 90.3% of this decrease is captured by the public transit system. Therefore, the model is able to reproduce the empirical relationship previously documented. However, this increase in density is keeping all else constant, in particular where people live and work. Zoning policy changes that liberate residential land will not have a uniform effect across other locations and the full model will be to put to work to capture this fact. However, this is left as future work