Minjung KTendency to Choose "Protection against Crime and Violence" as Predicted by the WGI
by MJ & Philipp; inspiration from Erin
We examined which of the five WGIs -- (1) rule of law, (2) political stability, (3) government effectiveness, (4) crime and corruption, and (5) voice and accountability -- contributed to the survey participants choosing "protection against crime and violence. Note that the analysis reported below was by me, MJ, and only I should be blamed for any errors.
We had a lot of trouble working with the data set, and Philipp created a cut-down version for us to use. But, in our subset, we noticed that government effectiveness and crime and corruption were duplicates. We think that the columns are duplicated in the original myworld_wgi.csv file.
Coded choosing-protection and not-choosing-protection as ones and zeros, respectively.
Took the mean for each country (i.e., can range from 0 to 1).
Ran regression, but unstable estimates? Final model: Y ~ GE_COC * VA + GE_COC + VA. All p < .05, adjusted R^2 = 8%. [Sorry, English below]
This is the scatterplot showing the tendency to choose "protection" as a function of voice and accountability.
For both plots, a knee-shaped bend around zero is obvious. This suggests that there are two groups of countries: ones with high WGIs, and which do not desire more protection with increasing WGIS, and ones with low WGIs, which do desire more protection with increasing WGIs.
To explore the hypothesis that there exist two subgroups of countries, we could code the countries could be coded by continent or region, then see if they cluster in the scatterplots. I don't know the scale for these WGIs, but assuming that low means bad and high means good, I have a suspicion that the countries with low scores on government effectiveness/crime and corruption tend to also have low scores on voice and accountability and vice versa.