Logo eprints

Essays on contest experiments and supervised learning in the pharmaceutical industry

Niederreiter, Jan (2020) Essays on contest experiments and supervised learning in the pharmaceutical industry. Advisor: Riccaboni, Prof. Massimo. pp. 238. [IMT PhD Thesis]

[img] Text (Doctoral thesis)
Niederreiter_phdthesis.pdf - Published Version
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (3MB)


The field of economics witnesses a growing interest to better understand how individuals form decisions and how these decisions can be supported with the help of sophisticated datamining tools. The first half of the thesis analyzes investment decisions in a competitive environment that is free of confounding factors by means of experimental data. The second half focuses on how supervised learning algorithms can be applied to predict the outcome and perceived success probability of pharmaceutical projects for aiding decisions of policy makers, managers and financial investors. More specifically, chapter two contrasts repeated individual expenditure decisions in different contest treatments by varying the uncertainty of the outcomes and the number of contest opponents. Contests with probabilistic outcomes show decreasing over-expenditures and a higher rate of “drop out”. If outcomes become deterministic, expenditures quickly convergence towards equilibrium predictions and a near to full participation. These results are robust to changes in the number of opponents. A learning parameter estimation using the experience-weighted attraction model suggests that subjects adopt different learning modes across different contest structures and helps to explain expenditure patterns deviating from theoretical predictions. Chapter three explores the presence of latent contestant types by applying the classifier Lasso to two versions of contest experiments, one that keeps the grouping of contestants fixed and one that randomly regroups contestants after each round. Results suggest that there exist three distinct types of players in both contest regimes. The majority of contestants in fixed groups behaves reciprocal to opponents’ previous choices. For experiments in which contestants are regrouped, the share of “reciprocators” is significantly lower. In both cases the reaction of other player types seems to differ from what is expected from a myopic best-reply. Pharmaceutical drug development can be seen as a real-world example for lottery contest. Assessing drug candidates’ odds of success often relies on methods based on historical success rates. However, machine learning offers a more data-driven approach to identify promising projects. To evaluate its usefulness, chapter four assesses the performance of several supervised learning algorithms that are trained and validated on a large database of projects. Using a sizeable list of project characteristics as input data, classification via state-of-the-art supervised learning methods is more accurate compared to more simplistic methods. The chapter aids stakeholders in the pharamceutical industry to make more informed decisions regarding stage-specific project outcomes. Chapter five extends the study of project outcomes by assessing the relationship between product innovation announcements of bio-pharmaceutical companies and their stock reactions using an event study approach. We hypothesize that financial returns that follow news on product innovation are shaped by a “probability effect”, that depends on how investors perceive the product’s likelihood of success, and a “portfolio effect” that depends on the relative importance of a product within a company’s portfolio. To test for the probability effect, project specific success probabilities are estimated via supervised learning methods. The portfolio effect is measured by the share of the product’s net present value. Market reactions are found to be higher when assosciated to projects with high portfolio importance but lower when associated to projects with high expected success probability.

Item Type: IMT PhD Thesis
Subjects: H Social Sciences > HB Economic Theory
PhD Course: Economics management and data science
Identification Number: 10.6092/imtlucca/e-theses/302
NBN Number: urn:nbn:it:imtlucca-27324
Date Deposited: 19 Mar 2020 14:12
URI: http://e-theses.imtlucca.it/id/eprint/302

Actions (login required, only for staff repository)

View Item View Item