Essays in applied microeconomics
OA Version
Citation
Abstract
This dissertation consists of three chapters concerning topics in applied microeconomics. In Chapter 1, I investigate the impact of the 1960s riots on an individual's choice to return to the South following the Great Migration. Using the 1980 5% IPUMS data and Carter's (1989) extraordinary data collection of the 1960s riots, I find a significantly negative correlation between riots severity and mobility for both white and blacks. Following Collins and Margo's (2007) instrumental variable strategy, I exploit exogenous variations in rainfall in April 1968, during which hundreds of riots broke out after the assassination of Martin Luther King Jr. Using rainfall as an instrumental variable, I find a negative causal effect of riot severity on the likelihood of return migration to the South.
In Chapter 2, using a unique dataset of retracted papers collected by Retraction Watch, I investigate the factors associated with the time to retraction and the reasons for retraction. I find that the number of authors positively correlates with time-to-retraction. Papers with at least one author that is affiliated with a US institution have a longer time to retract and are more likely to be revealed with clear reasons for retractions. The number of fields is negatively correlated with time-to-retraction, while the number of subjects is positively correlated with it. Compared to papers in social sciences, papers in environmental sciences and business and technology take less time, while papers in humanities take longer to retract. Papers in basic life sciences and business and technology are more likely to be retracted for "ugly" reasons instead of good reasons, while papers in health sciences are less likely to be retracted for bad or "ugly" reasons instead of good reasons, compared to papers in social sciences. Interestingly, having at least one author who is affiliated with a university decreases the probability of having a clearly stated reason for retraction. In addition, the publication year is negatively correlated with time-to-retraction, suggesting that either the speed of discovering false science or the speed of responding to mistakes in papers (or both) has increased over the years. Using a DID framework, I find retraction decreases the annual citations received by papers, and the impact is greater for papers that are retracted due to severe misconduct than those with milder mistakes or unknown reasons.
In Chapter 3, I extend the results in Chapter 2 and investigate further the relationship between retractions and citation patterns, focusing on citation sentiments. Using a unique collection of citation contexts of the citing papers to the retracted and control papers, and a state-of-art rule-based model, I show that there is a sharp increase in explicitly negative citations and a moderate increase in implicitly negative citations for retracted papers before retractions. While implicitly positive and neutral citations increase slightly for retracted papers before retraction, their explicitly positive citations drop sharply before retractions. Through a staggered DID framework, I also provide causal evidence that retractions lead to decreases in all types of citations by sentiments.