Tesla and the Future of Autonomous Driving

Abstract

An investor, Janice Zhuk, examines Tesla's quarterly safety data that includes the number of miles it took to register one accident among Tesla cars driving with autopilot versus those without autopilot. The data set is small, so the validity of any statistical analysis may be compromised. The data allow one to run a simple t-test comparison of means. Because of the small sample size, the case invites a discussion on the hypothesis of the regression model, in particular, the normality of the residuals.

Excerpt

UVA-QA-0911

Rev. Jul. 24, 2019

Tesla and the Future of Autonomous Driving

Janice Zhuk was an investor who for the past 20 years had success investing in new technology. Getting in on the ground floor meant being comfortable with uncertainty. Her newest interest—a fleet of shared robotaxis. As ideas around electric vehicles (EV) and autonomous vehicles (AV) moved into development, she closely followed just about everything in the works. With Toyota, Ford, and Tesla investing in EV and AV, Zhuk had much to follow, but Tesla was her favorite.

How could Zhuk resist the emotion the firm's flamboyant CEO, Elon Musk, ignited around AV and his claim that by 2020, Tesla would have a million self-driving robotaxis on the road? And with a billowing $ 2.35billion stock and bond boost in early May 2019, it looked like Tesla had investors' attention. Musk himself announced he would buy $ 25 million more in shares (102,880), pushing the deal to $ 2.7 billion. Three days later, the capital raise was closed—it had been oversubscribed.

Reacting with less emotion, Zhuk kept data on each step artificial intelligence made toward self-driving. She focused on autopilot and the question of safety—since Tesla was transparent around some data collection, she checked in. How was it going?

. . .

Keywords: Tesla, autopilot, safety, hypothesis test, regression, dummy variables