An Experimental Investigation of Speed Choice and Information

Published: 10-03-2021| Version 1 | DOI: 10.17632/htwskf2vxd.1
Contributors:
Carina Goldbach,
Deniz Kayar,
Thomas Pitz,
Jörn Sickmann

Description

This data set provides data from a new experiment, played in the computer lab, in which participants had to make the choice between travelling fast or slow from a hypothetical A to B. Participants were grouped in groups of eight. The participants’ payoff depended on the time needed to arrive at B. If they arrived at B after choosing slow, they received 1 experimental currency unit (ECU). If participants arrived at B after choosing fast, they received a higher payoff of 2 ECU. However, every participant faced the risk of having an accident that resulted in not arriving at B and therefore generating a payoff of 0 ECU. The probability to arrive for fast and slow is, respectively: P(f)=1/f(1/((-n)/m) x+f) P(s)=1/s(1/(-n) x+s) where x is the number of fast players, n the group size, f the payoff of fast, s the payoff of slow and m a factor influencing the accident risk when speeding. The main variable is "fast" which takes the value of 1 if the participant chose to travel fast, and 0 if the participant chose slow. The experiment is designed in a way which emphasizes the fact that choosing fast does not only decrease the participant’s likelihood to arrive at B, but also the likelihood of everyone else and runs over a total of 100 rounds. We compare this baseline condition with three variations: In the first variation, treatment I, we increase the underlying likelihood of having an accident in order to test the sensitivity to the accident risk in our experimental setting. In treatment II quantitative information will be provided and participants will be informed about the total number of accidents after each round. In treatment III we vary the information slightly and only present a qualitative measure of how many accidents occurred. All four conditions consisted of six groups of eight players. A consecutive questionnaire collected information on age, gender, ownership of a driver’s license, basic risk aversion and time preference. Additionally, the Big Five personality factors were captured. The STATA dataset contains labels further explaining the different variables. Ultimately, results showed that additional information had an effect on speeding choice if it contains quantitative instead of qualitative information. However, the provided quantitative information led to increased speeding behavior and a higher number of accidents. Across all treatments it was found that individual characteristics help explain differences in speeding behavior significantly

Files