Embedded Files
Working Papers
Working Papers
Freewheeling: A Spatial Structural Analysis of the Bike-Sharing Industry (Job Market Paper) [pdf]
This paper studies the spatial mismatch between consumers and bikes in the dockless bike-sharing industry and an externality exacerbating the problem: when a consumer uses a bike for a low and inflexible price, she both displaces another consumer's usage for a potential higher-value trip, and may ride the bike to unpopular destinations. With a trip-level dataset of a bike-sharing company in Beijing, China, I develop a spatial structural model to estimate the demand for bikes with search frictions and local matchings. Compared to the scenario that consumers always get bikes immediately, I find that local spatial mismatch between consumers and bikes reduces the total usage by 29.95%, or a net loss of 332,979 trips. Counterfactual analyses show that (1) doubling the number of bikes increases the trip volume by 28.46% while halving the number of bikes decreases the trip volume by 46.40%; (2) price-discriminating against short trips by 2% increases the total trip time by 0.22%; and (3) changing the frequency of bike reshuffling does not have a significant impact on the total usage of bikes.
The Animated Choropleth Illustrating the Spatial Mismatch Between Bikes and Consumers
The Animated Choropleth Illustrating the Spatial Mismatch Between Bikes and Consumers
The figure on the left is a bicolor choropleth illustrating the inconsistent spatio-temporal distribution of trips in Beijing, May 17, 2017. Each of the outlined sections of Beijing is one of the 134 locations indexed by i when serving as starting locations and indexed by j when serving as ending locations. As the legend shows, the color on the horizontal axis darkens with more trip starts; the color on the vertical axis darkens with more trip ends. Each frame is the observation in a half-hour period. The uneven distribution of bike trips is demonstrated by regions with stark different colors. There are three regions with dark colors: the northeast, the southeast, and the southwest. The northeast and the southwest regions have large residential areas and thus have high demand. The southeast region has large business districts. The city center has light colors. Government buildings and historical sites make up the majority of the area, hence the demand for bikes is low. The city outskirts also have light colors. The low number of trips in these regions is largely attributable to lower population density and less access to public transportation.
The figure on the right is a choropleth illustrating the distribution of bikes in Beijing, May 17, 2017. I divide the number of bikes in one tract over the total number of bikes in Beijing, and the darker the color, the more bikes are in the tract. Each frame is the observation in a half-hour period. Since the price is low and inflexible, the distribution of bikes cannot adjust to accommodate the change in demand as demonstrated by the figure on the left. The aforementioned high-demand areas still have dark colors, but the changes in the distribution of bikes are more significant among all tracts.
Work in Progress
Work in Progress
Financing and Investment by a Platform Start-up: An Analysis of ofo
This paper studies the business performance of a bike-sharing start-up company in China and shows how its financial conditions affect investment decisions. I fit a dataset of daily active users and trips in several cities with different specifications of time functions and analyze the effects of funding rounds from venture capitalists on the investment and business of the company with an event study framework. My estimates find that the firm increases its users and bikes by about 40% two weeks before receiving funds, suggesting that it invests extensively in expectation of deep pockets. The boost in business performance from capital influx will decrease as the market expands, suggesting decreasing returns in scale on capital. I also show that such boosts in business performance are short-lived: the number of trips and users often return to normal levels two weeks after the funding day.
A Simulation Analysis of the Research Productivity Model of Industry Shakeout: US Automobile Industry from 1900 - 1950
This paper studies the shakeout in the U.S. automobile industry with data retrieved from old annals of the automobile industry. I simulate a research productivity model and see if I could successfully trigger a shakeout. I find that only the cost reduction from technology advancements is not enough to trigger an industry shakeout and propose that more extreme settings are needed for further studies.
Honors Thesis in Mathematics at Davidson College
Honors Thesis in Mathematics at Davidson College
Google Sites
Report abuse