Prof. Artinger at Decision Analysis Conference
Prof. Dr. Florian Artinger, Head of Program, B.A. Digital Business & Management at Berlin International, participated in the prestigious Advancements in Decision Analysis Conference, which was held from June 22nd to June 24th, 2022 at the Darden Business School in Washington DC.
In his keynote address, he talked about the most important findings with regards to his research projects “How Do Taxi Drivers Determine How Much to Work If Earnings Are Hard to Predict?” and “Predicting Revenues with the Multiplier Heuristic”.
The participants were internationally leading scholars from the field and came from various institutions. The stunning program included over seventy presentations, a rich poster series, three impressive keynote speakers and PhD incubators with 18 dedicated talks.
The Advancements in Decision Analysis Conference aimed to bring together scientists working in decision analysis. The conference provided a platform for interdisciplinary discussions and included talks by researchers in many fields such as decision analysis, behavioral economics, judgment and decision-making, machine learning, statistics, and other related disciplines with a prescriptive focus.
About the Advancements in Decision Analysis Conference
The Advancements in Decision Analysis Conference is a two-year focal point for the society, it is the place where members exchange ideas and shape the scientific future of the discipline. This edition was even more significant, as it is the first occasion for the entire community to get together in person after over two years of distance meeting due to the COVID-19 pandemic.
About the research project
How Do Taxi Drivers Determine How Much to Work If Earnings Are Hard to Predict?
A fundamental assumption of expected utility models is that agents make predictions by formulating rational expectations. Building on this assumption, the literature has addressed to what extent neoclassical or behaviorally informed utility models best describe intertemporal substitution of labor and leisure, focusing on the taxi market. Using data from 10 million taxi trips, we find that hourly earnings are barely predictable. Under such uncertainty, satisficing models predict behavior of drivers better than utility models. These models do not require calculating expected earnings but terminate shifts when reaching an aspiration level on shift duration or earnings.
About the research project
Predicting Revenues with the Multiplier Heuristic
Forecasts by statistical and machine learning methods are usually regarded as superior over those made by experts. Nonetheless, many experts still rely on simple heuristics. Are there conditions under which expert forecasts based on heuristics can match the performance of statistical and machine learning methods? We examine the case of predicting revenues per customer in 20 data sets where experts rely on the following heuristic: multiply the revenue observed in the first t days by a constant. We find that with limited sample size and a shorter prediction horizon that the heuristic can perform on par or even outperform statistical and machine learning methods. Given unpredictable changes over time, the heuristic even performs on par given very large samples and a longer prediction horizon. The results provide insights when to rely on managerial judgment and when on statistical and machine learning methods.