DATA ANALYSIS / REGRESSION / INTERPRETATIONCOLLABORATIVE COURSE PROJECT · 2024
01 / THE STUDYCOUNTRY COMPARISONS · EXPLORATORY MODELS
CORRUPTION · PROTEST · EMIGRATION
When people can speak, or leave, what patterns appear in the data?
2022 INDICATORS · FINAL SAMPLE VARIES BY ANALYSIS
Shifting Shadows brings together the Corruption Perceptions Index (CPI), recorded protest events and emigration totals. The report describes an initial 50-country dataset, while some later analyses use a final sample of 35 countries. It uses descriptive statistics, confidence intervals, hypothesis tests, ANOVA and regression in RStudio.
Three measures, different scales
The variables are country-level aggregates, not measures of an individual’s views or migration decision. The project compares values and fits exploratory models to look for associations.
01 / CPI · 2022
Perceptions index
A public-sector corruption perceptions measure; higher scores generally indicate a cleaner public sector.
02 / PROTEST · 2022
Recorded events
Counts of protest events, as described by the project’s ACLED data source.
03 / EMIGRATION · 2020
Emigration totals
Country totals cited from Our World in Data, with a different reference year from the other measures.
04 / SAMPLE
50 → 35 countries
The written report describes an initial 50-country sample and a 35-country final sample for some analyses.
THE POWERPOINT · INSIDE THE PAGE
Browse the full presentation.
Use the blue slide controls or keyboard arrows. Each slide has a concise reading note below the presentation frame.
SHIFTING SHADOWS / COUNTRY STUDY01 / 17
01 / OVERVIEW
Shifting Shadows
A student study of corruption perceptions, protest events and emigration, using country-level indicators and statistical models.
The presentation reports a weak CPI-only relationship with emigration (R² around 0.10; p above .05), and a stronger simple association between protest counts and emigration totals (R² around 0.35). An ANOVA comparing protest means by subregion reports p around 0.33, so it does not find a statistically significant difference under that test.
A later multiple regression reports a higher overall fit, but the report and slides shift sample definitions and describe a mixture of predictors. The coefficient interpretation should be checked against the underlying data and script before being treated as a settled result.
Why the association is not a causal pathway
The data are observational and aggregated by country. A relationship between protest counts and emigrant totals cannot tell whether protest activity caused migration. Population size, conflict, labour markets, geography and other unmeasured differences could affect both.
The source slide uses CPI as a perception index, protest events from 2022, and emigration totals from 2020. Different years and raw totals reduce direct comparability.
