2. How We Analyze Data

1. Computational analysis of digital textual data found online: We are using sentiment analysis, topic modeling, Word2Vec, and machine learning classifiers to computationally analyze how online public discourse regarding emigration from Hong Kong has changed since 2019.

2. Propensity to Migrate Model: To explain what is associated with emigrating, we run a logistic model that regresses whether a given participate emigrated on the various individual and contextual level of traits of each survey participant.

3. Duration Models: To explain the length of time that passes between thoughts of emigrating and emigration, we are regressing the number of months between these two events on different traits of those who emigrated.

4. Discrete Time Hazard Survival Model: Once we have obtained a sufficient amount of data to have variability in the time that an individual undergoes major life course changes (e.g. marries, has a child, purchases a home, and obtains a foreign citizenship) ) assess how these individual-level covariates are associated with the time that passes before an individual emigrates out of Hong Kong.

5. Analysis of oral history interviews will seek to explain the processual sequence of events and conjunction of events that led to an individual emigrating.

6. Using a framework of grounded theory, we will inductively extract themes from the oral histories and test hypotheses deductively derived from the prior literature on the determinants of emigration.