Tipping is a social norm in many countries and widely recognized as an anomalous behavior, in that a tip is common enough to have become expected when dealing with tipped industries (e.g., restaurants, bars, taxi trips), while at the same time defying rational-agent assumptions of economics. Such intriguing consumer behavior has led to its wide study across the world. However, most studies of tips and tipping behavior have suffered from a lack of data, relying on surveys and manually collected information. Here we analyze a dataset of 13 million taxi trips with their associated tips, in order to examine tips as they compare to the average income of the location from which the trip originated. We discovered that tipping behavior is temporally stable during either the time of the day or the day of the week. Also, when people tip, there is no statistically significant correlation between the amount tipped and the people’s income. However, passengers who do not leave any tip (i.e., the stiffers) exhibit consistent patterns both temporally and spatially in which the highest frequency of stiffers occurs around 4am, and the tendency of a passenger to stiff the taxi driver presents a strong negative correlation with income. The understanding of social behavior is important in ubiquitous computing in particular in smart-city contexts. A more complete understanding of human behavior is intrinsically linked to our ability to develop smarter cities.
IEEE Ubiquitous Intelligence and Computing, San Francisco, California, USA. 2017.
[Download preprint here]
[Link to publication]