Three Essays on Online Economic Experiments and Experimental Data Analysis
Bibliographic record
Abstract
This dissertation consists of three chapters on economic experiments and experimental data analysis. The first two chapters are online experiments and surveys, which explore the two topics of the health state valuation and the voluntary provision of public goods, respectively. The third chapter is a strategy analysis of trust behavior. \nIn the first chapter, to explore how people value the state of health and what socio-economic factors they might consider, I conducted a survey experiment to elicit individuals’ decisions under hypothetical health states. The main task for the subjects was a valuation task (standard gamble) under given health states, in which the subjects were required to make decisions on whether to take a risky medical treatment when facing various success probabilities. After this procedure, the subjects filled a survey about what factors they might have considered when making the previous decisions. The subjects were from two separate online pools of the United States (Amazon MTurk) and Canada (AskingCanadians). \nMy results show that in those who choose the risky medical treatment under the same health states, the Canadian participants are willing to accept a lower success probability. Among the socio-economic factors that are significant to this health valuation, several factors are considered by both samples such as “employer-purchased insurance plans”, “personal financial situations”, and “waiting times for treatment”. Some factors are only significant in one country’s participants. For the American sample, it is “access to health insurance”, while for the Canadian sample, it is “disturbances in everyday family life”. \nThe second chapter is an online experiment of a public goods game, which has a particular feature of polarized preference. From the 2020 U.S. election to the oil pipeline development in Canada, these types of situations may be modeled by a public goods game in which two groups of individuals have polarized preferences. The outcome of the election or debate will affect the utility of individuals in both groups but in an opposite direction. Meanwhile, individuals from each group can make costly efforts (in their favor) trying to affect the outcome. \nWe study a public goods game with polarized preferences by using a generalized voluntary contribution mechanism (GVCM). The strategy method was applied to the design of an online experiment. There are two groups of players, a majority group and a minority group, who have polarized preferences for a public good. Each player decides whether to contribute to their group's public account or keep the token in their private account. The experiment consists of a 2×2 design, which allows us to examine the effect of different MPCRs and frameworks in the conditional contribution. The subjects were recruited from Amazon MTurk and the experiment was implemented using o-Tree. \nThe main results that we found are that the MPCR effect and framework effect are mixed and only significant in some treatments. The results vary depending on the role of the participants (the majority and the minority). Overall, the individual contribution frequency in the majority group is significantly larger than in the minority group. Furthermore, players' contribution significantly increases with the contribution of others in their own group but is not dependent on the contribution from the other group. \nThe third chapter is an experimental data analysis, which seeks to reveal the strategy behind trust behavior among the encountering of strangers. The data set is from a trust game experiment reported by Duffy, Xie, and Lee (2013). If people never met again and people would not be punished for dishonesty (at least not directly from the person they cheated), the actions would be different. This scenario could be simulated in a trust game where there are two roles (Investor and Trustee) and the subjects are randomly and anonymously matched. \nThe method of finite automata is applied to infer the strategies subjects used in the experiment. In the strategy fitting procedure, I define for Investor 16 strategies and 6 strategy sets, and for the second player (Player B: Trustee) 24 strategies and 11 strategy sets. Then I match the data through a fitting procedure with these defined strategies. I report that the top three strategies in order are “grim trigger”, “systematically Send”, and “forgiving”; for Trustee, the most used strategies are “systematically Return”, “grim trigger”, and “tit-for-tat”. By taking the probability of the participant’s error into account, more observations are classified into the strategy, and the strategy pattern and proportions are still maintained.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.044 | 0.228 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.002 | 0.003 |
| Science and technology studies | 0.003 | 0.006 |
| Scholarly communication | 0.006 | 0.006 |
| Open science | 0.002 | 0.005 |
| Research integrity | 0.004 | 0.008 |
| Insufficient payload (model declined to judge) | 0.047 | 0.010 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".