In the previous post, we saw that your data correlates reasonably well with professional efforts to measure democracy – both the AI-generated data, and your adjusted version, after checking the AI answers. But you also collected some data on conflict (both domestic and external), which was a bit more subjective. In this post, we first check that your data on internal conflict/repression is comparable to other estimates, and look at some basic correlations between our data and estimates of internal peace and repression. (Another post will focus on external conflict and war).
Domestic Repression Data
We are going to look at two different measures of domestic peace/repression. First, the measure that you gathered with some AI help; and second, a latent measure of respect for basic physical integrity rights developed by Christopher Fariss Christopher J. Fariss, Kenwick, and Reuning (2020). Fariss also provides a measure of whether a country is experiencing “one-sided killings”, where a government is engaged in killing of their own citizens. Luckily, these measures are correlated in the right direction (Table 1), though only at the 0.6 level (in contrast to your data on democracy, which correlated at the 0.9 level with V-Dem). The correlation is strongest between the AI measure of domestic peace (lack of internal conflict/repression) and the Christopher J. Fariss, Kenwick, and Reuning (2020) measure, and weakest between the measures of one-sided killing (which are in any case rare events) and your measures of internal peace/lack of repression.
| term | POLS209 AI Internal Peace measure | POLS209 Student Internal Peace Measure | Fariss 2020 HR Respect | Fariss 2020 one-sided killing |
|---|---|---|---|---|
| POLS209 AI Internal Peace measure | NA | 0.82 | 0.63 | -0.24 |
| POLS209 Student Internal Peace Measure | 0.82 | NA | 0.62 | -0.22 |
| Fariss 2020 HR Respect | 0.63 | 0.62 | NA | -0.40 |
| Fariss 2020 one-sided killing | -0.24 | -0.22 | -0.40 | NA |
Despite these correlations, there may be some sources of error: sometimes scores of 10 were given to more repressive rather than more peaceful countries, for example (because of a problem with the prompt), and I’m not sure I caught all of these cases (see Table 3 below for a potential example); and in at least one case a country was given an impossible score. In any case, the top ten countries where your measure of domestic peace differs the most from the Christopher J. Fariss, Kenwick, and Reuning (2020) measure (Table 2) are small and difficult to research; it is not obvious if the error is from your research, or if the Fariss measure is wrong. They are also mostly cases where POLS209 considers these countries more peaceful domestically (with less repression) than Christopher J. Fariss, Kenwick, and Reuning (2020).
| Country | Diff. between scaled AI answer on internal peace and Fariss 2020 (0-1) | Diff. between scaled student answer on internal peace and Fariss 2020 (0-1) | AI peace (0-1, higher is more peaceful) | Student peace (0-1) | Fariss 2020 (0-1) | Min year | Max year |
|---|---|---|---|---|---|---|---|
| Maldives | 0.28 | 0.77 | 0.50 | 0.00 | 0.77 | 1978 | 1988 |
| Comoros | 0.47 | 0.67 | 0.50 | 0.30 | 0.97 | 2010 | 2024 |
| Uganda | 0.41 | 0.64 | 0.50 | 0.80 | 0.17 | 1986 | 2024 |
| Laos | 0.40 | 0.61 | 0.50 | 0.00 | 0.49 | 1975 | 2024 |
| Mali | 0.21 | 0.58 | 0.50 | 0.00 | 0.53 | 1960 | 1991 |
| Rwanda | 0.56 | 0.55 | 0.09 | 0.00 | 0.51 | 1973 | 1998 |
| Guinea-Bissau | 0.57 | 0.54 | 0.42 | 0.10 | 0.62 | 1974 | 2008 |
| Solomon Islands | 0.50 | 0.50 | 0.50 | 0.50 | 1.00 | 2006 | 2014 |
| Kenya | 0.46 | 0.48 | 0.60 | 0.60 | 0.13 | 2002 | 2008 |
| Peru | 0.43 | 0.45 | 0.50 | 0.46 | 0.34 | 1985 | 2020 |
Domestic Repression and Democracy
As we have learned in class, though democracies can be domestically repressive, they should be less repressive than non-democratic regimes due to the basic feedback mechanisms built into democratic politics. At the same time, not every non-democratic regime should be equally repressive.
All of the measures of domestic repression/respect for basic integrity rights are related in the expected direction with your measure of democracy (Figure 1), though the Fariss measure has a lot more dispersion: more democracy is correlated with less repression and more respect for physical integrity rights. We can also see that there are many non-democratic regimes (in the upper-left corner) with limited repression.
It’s also worth trying to see if your measure of accountability (rather than democracy) is associated with more or less repression (Figure 2). Here the relationship still exists but is noisier than the relationship with democracy as a whole.
Figure 2: Level of accountability vs. internal peace/repression
Interactive versions of these graphs are available here.
Outliers and Cases that don’t fit the General Pattern
We can identify the top least repressive non-democratic regimes in your dataset, in case you wish to investigate them further (ranked by your measure of internal peace, Table 3). Of these it looks to me like at least Eritrea is a data-entry error (it probably should have been 0 on internal peace), perhaps because when it was coded the prompt scales were wrongly set up.
| Country | Avg. POLS209 measure of internal peace (higher is more peaceful/less repressive) | Avg. POLS209 measure of democracy | Min year | Max year |
|---|---|---|---|---|
| Eritrea | 10.00 | 10.00 | 1991 | 1993 |
| Sierra Leone | 10.00 | 9.00 | 1971 | 1985 |
| United Arab Emirates | 9.49 | 22.91 | 1971 | 2014 |
| Hungary | 8.00 | 31.25 | 1956 | 1989 |
| Liberia | 8.00 | 37.00 | 1997 | 1999 |
| Uganda | 8.00 | 26.00 | 1986 | 2024 |
| Panama | 7.00 | 43.00 | 1968 | 1984 |
| Sao Tome and Principe | 6.00 | 19.00 | 1983 | 1986 |
These differ a bit from the top least repressive countries by the Christopher J. Fariss, Kenwick, and Reuning (2020) measure (Table 4).
| Country | Avg. Fariss 2020 measure of respect for HR (0-1, higher is more peaceful/less repressive) | Avg. POLS209 measure of democracy (0-100, higher is more democratic) | Min year | Max year |
|---|---|---|---|---|
| Cape Verde | 0.62 | 1.50 | 1975 | 1975 |
| Fiji | 0.90 | 5.00 | 2009 | 2013 |
| Sao Tome and Principe | 0.67 | 19.00 | 1983 | 1986 |
| Maldives | 0.77 | 22.00 | 1978 | 1988 |
| United Arab Emirates | 0.93 | 22.91 | 1971 | 2014 |
| Seychelles | 0.61 | 28.00 | 1981 | 1986 |
| Oman | 0.94 | 29.50 | 1996 | 2006 |
| Guinea-Bissau | 0.62 | 36.87 | 1974 | 2008 |
| Armenia | 0.66 | 44.00 | 1991 | 1996 |
| Kuwait | 0.88 | 46.00 | 1999 | 2006 |
What do these countries have in common, if anything?
We can also identify the top most repressive democracies/hybrid regimes in your dataset. What do you think they have in common, if anything?
| Country | Avg. POLS209 measure of internal peace (higher is more peaceful/less repressive) | Avg. POLS209 measure of democracy | Min year | Max year |
|---|---|---|---|---|
| United States of America | 4.89 | 72.96 | 1900 | 2024 |
| South Africa | 5.00 | 71.00 | 1994 | 2023 |
| Uruguay | 5.00 | 71.00 | 1985 | 1989 |
| Argentina | 3.21 | 70.00 | 1983 | 2012 |
| Dominican Republic | 5.00 | 67.00 | 1996 | 2004 |
| Turkey (Ottoman Empire) | 5.00 | 64.00 | 2002 | 2006 |
| Niger | 5.00 | 63.00 | 2021 | 2022 |
| Solomon Islands | 5.00 | 63.00 | 2006 | 2014 |
| Macedonia (Former Yugoslav Republic of) | 5.00 | 61.00 | 1999 | 2006 |
| Nepal | 5.00 | 61.00 | 2006 | 2018 |
No comments:
Post a Comment