Skip to content
Permalink
Branch: master
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
424 lines (318 sloc) 15.1 KB
---
title: Identifying Opportunity for Conflict
subtitle: POSC 3610 -- International Conflict
author: Steven V. Miller
institute: Department of Political Science
titlegraphic: /Dropbox/teaching/clemson-academic.png
date:
fontsize: 10pt
output:
beamer_presentation:
template: ~/Dropbox/miscelanea/svm-r-markdown-templates/svm-latex-beamer.tex
latex_engine: xelatex
dev: cairo_pdf
fig_caption: true
slide_level: 3
make149: true
mainfont: "Open Sans"
titlefont: "Titillium Web"
---
```{r setup, include=FALSE, cache=F, message=F, warning=F, results="hide"}
knitr::opts_chunk$set(cache=TRUE, warning=F)
knitr::opts_chunk$set(fig.path='figs/')
knitr::opts_chunk$set(cache.path='cache/')
knitr::opts_chunk$set(
fig.process = function(x) {
x2 = sub('-\\d+([.][a-z]+)$', '\\1', x)
if (file.rename(x, x2)) x2 else x
}
)
options(knitr.kable.NA = '')
```
```{r loadstuff, cache=T, eval=T, echo=F, message=F, error=F, warning=F}
library(tidyverse)
library(janitor)
library(knitr)
Data <- read.csv("~/Dropbox/projects/mid-project/gml-mid-data/2.02/gml-ndy-2.02.csv") %>% tbl_df()
Cont <- read.csv("~/Dropbox/data/cow/contiguity/3.2/contdird.csv") %>% tbl_df()
Majors <- read.csv("~/Dropbox/data/cow/states/majors2016.csv") %>% tbl_df()
Contcd <- read.csv("~/Dropbox/data/cow/coldepcont/contcold.csv") %>% tbl_df()
Contcd %>%
rename(ccode1 = statelno, ccode2 = statehno) %>%
select(ccode1, ccode2, year, land, sea, total) %>%
left_join(Data, .) -> Data
Cont %>%
rename(ccode1 = state1no, ccode2 = state2no) %>%
select(ccode1, ccode2, year, conttype) %>%
left_join(Data, .) -> Data
Data %>%
mutate(conttype = ifelse(is.na(conttype), 6, conttype),
land = ifelse(is.na(land), 0, land),
sea = ifelse(is.na(sea), 0, sea),
total = ifelse(is.na( total), 0, total),
contig = ifelse(conttype == 6 & total == 0, 0, 1),
# Code regions...
region1 = NA,
region1 = ifelse(ccode1 < 200, "Americas", region1),
region1 = ifelse(ccode1 >= 200 & ccode1 < 400, "Europe", region1),
region1 = ifelse(ccode1 >= 400 & ccode1 < 600, "Africa", region1),
region1 = ifelse(ccode1 >= 600 & ccode1 < 700, "ME/NA", region1),
region1 = ifelse(ccode1 >= 700 & ccode1 < 900, "Asia", region1),
region1 = ifelse(ccode1 >= 900, "Oceania", region1),
region2 = NA,
region2 = ifelse(ccode2 < 200, "Americas", region2),
region2 = ifelse(ccode2 >= 200 & ccode2 < 400, "Europe", region2),
region2 = ifelse(ccode2 >= 400 & ccode2 < 600, "Africa", region2),
region2 = ifelse(ccode2 >= 600 & ccode2 < 700, "ME/NA", region2),
region2 = ifelse(ccode2 >= 700 & ccode2 < 900, "Asia", region2),
region2 = ifelse(ccode2 >= 900, "Oceania", region2),
# Code majors...
maj1 = NA,
maj1 = ifelse(ccode1 == 2 & year >= 1898, 1, maj1),
maj1 = ifelse(ccode1 == 200, 1, maj1),
maj1 = ifelse(ccode1 == 220 & year <= 1940, 1, maj1),
maj1 = ifelse(ccode1 == 220 & year >= 1945, 1, maj1),
maj1 = ifelse(ccode1 == 255 & year <= 1918, 1, maj1),
maj1 = ifelse(ccode1 == 255 & (year >= 1925 & year <= 1945), 1, maj1),
maj1 = ifelse(ccode1 == 255 & year >= 1991, 1, maj1),
maj1 = ifelse(ccode1 == 300 & year <= 1918, 1, maj1),
maj1 = ifelse(ccode1 == 325 & (year >= 1860 & year <= 1943), 1, maj1),
maj1 = ifelse(ccode1 == 365 & year <= 1917, 1, maj1),
maj1 = ifelse(ccode1 == 365 & year >= 1922, 1, maj1),
maj1 = ifelse(ccode1 == 710 & year >= 1950, 1, maj1),
maj1 = ifelse(ccode1 == 740 & (year >= 1895 & year <= 1945), 1, maj1),
maj1 = ifelse(ccode1 == 740 & year >= 1991, 1, maj1),
maj1 = ifelse(is.na(maj1), 0, maj1),
maj2 = NA,
maj2 = ifelse(ccode2 == 2 & year >= 1898, 1, maj2),
maj2 = ifelse(ccode2 == 200, 1, maj2),
maj2 = ifelse(ccode2 == 220 & year <= 1940, 1, maj2),
maj2 = ifelse(ccode2 == 220 & year >= 1945, 1, maj2),
maj2 = ifelse(ccode2 == 255 & year <= 1918, 1, maj2),
maj2 = ifelse(ccode2 == 255 & (year >= 1925 & year <= 1945), 1, maj2),
maj2 = ifelse(ccode2 == 255 & year >= 1991, 1, maj2),
maj2 = ifelse(ccode2 == 300 & year <= 1918, 1, maj2),
maj2 = ifelse(ccode2 == 325 & (year >= 1860 & year <= 1943), 1, maj2),
maj2 = ifelse(ccode2 == 365 & year <= 1917, 1, maj2),
maj2 = ifelse(ccode2 == 365 & year >= 1922, 1, maj2),
maj2 = ifelse(ccode2 == 710 & year >= 1950, 1, maj2),
maj2 = ifelse(ccode2 == 740 & (year >= 1895 & year <= 1945), 1, maj2),
maj2 = ifelse(ccode2 == 740 & year >= 1991, 1, maj2),
maj2 = ifelse(is.na(maj2), 0, maj2)) -> Data
Data %>%
mutate(region = NA,
region = ifelse(region1 == "Americas" & region2 == "Americas", "Americas", region),
region = ifelse(region1 == "Europe" & region2 == "Europe", "Europe", region),
region = ifelse(region1 == "Africa" & region2 == "Africa", "Africa", region),
region = ifelse(region1 == "ME/NA" & region2 == "ME/NA", "ME/NA", region),
region = ifelse(region1 == "Asia" & region2 == "Asia", "Asia", region),
region = ifelse(region1 == "Oceania" & region2 == "Oceania", "Oceania", region),
prd = NA,
prd = ifelse(contig == 1 | (maj1 == 1 | maj2 == 1), 1, 0)) -> Data
```
# Introduction
### Goal for Today
*Discuss contiguity and how to identify opportunities for conflict.*
### Dangerous Dyads
| **Factor** | **Rank** | **Relationship** |
|:----------|:-------:|:-------------------:|
| Contiguity | 1 | + |
| Major power in the dyad | 2 | + |
| Shared alliance | 3 | - |
| Joint Militarization | 4 | + |
| Joint democracy | 5 | - |
| Jointly advanced economies | 6 | - |
| Power preponderance | 7 | - |
Table: Bremer's (1992) "Dangerous Dyads"
# Identifying Opportunity for Conflict
## Contiguity
### Types of Contiguity
- Direct land contiguity (e.g. USA-Canada)
- 12 miles or less of water (e.g. USA-Russia)
- 24 miles or less of water (e.g. United Kingdom-France; Russia-Japan)
- 150 miles or less of water (e.g. USA-Cuba)
- 400 miles or less of water (e.g. United Kingdom-Germany)
### The Diomede Islands in the Bering Strait
![America is on the left; Russia is on the right](diomede-islands.jpg)
### Additional Contiguous Relationships
Two states may be contiguous through colonial dependencies.
- e.g. United Kingdom-China [Hong Kong]
- further: U.S.-UK (Canada historically, now the Virgin Islands and Bermuda)
Contiguity rules are the same as before, but stretch to include colonial holdings/dependencies.
###
```{r contiguity-conflict-period, eval=T, echo=F, message=F}
Data %>%
mutate(era = NA,
era = ifelse(year < 1946, "1816-1945", era),
era = ifelse(year > 1945 & year < 2002, "1946-2001", era),
era = ifelse(year > 2001, "2002-2010", era)) %>%
filter(midonset == 1) %>%
group_by(era, contig) %>%
summarize(n = n()) %>%
mutate(freq = round((n / sum(n))*100, 2),
freq = paste0(freq, "%")) %>% select(-n) %>% ungroup() %>% spread(era, freq) -> mids
mids %>%
mutate(contig = ifelse(contig == "0", "No", contig),
contig = ifelse(contig == "1", "Yes", contig)) %>%
rename(`**Contiguous?**` = contig,
`**1816-1945**` = `1816-1945`,
`**1946-2001**` = `1946-2001`,
`**2002-2010**` = `2002-2010`) %>%
kable(.,caption="Contiguity and Militarized Conflict (GML MID Data [v. 2.02])",
align=c("l","c","c","c"))
#Data %>%
# mutate(era = NA,
# era = ifelse(year < 1946, "1816-1945", era),
# era = ifelse(year > 1945 & year < 2002, "1946-2001", era),
# era = ifelse(year > 2001, "2002-2010", era)) %>%
# filter(midonset == 1 & hostlev == 5 & (dispnum != 258 | dispnum != 257)) %>%
# group_by(era, contig) %>%
# summarize(n = n()) %>%
# mutate(freq = round((n / sum(n))*100, 2),
# freq = paste0(freq, "%")) %>% select(-n) %>% ungroup() %>% spread(era, freq) -> wars
#rbind(c("*MIDs*",NA,NA,NA), mids,c("&nbsp;",NA,NA,NA), c("*Wars*",NA,NA,NA), wars) %>%
# mutate(contig = ifelse(contig == "0", "No", contig),
# contig = ifelse(contig == "1", "Yes", contig)) %>%
# rename(`**Contiguous?**` = contig,
# `**1816-1945**` = `1816-1945`,
# `**1946-2001**` = `1946-2001`,
# `**2002-2010**` = `2002-2010`) %>%
# kable(.,caption="Contiguity and Militarized Conflict (GML MID Data [v. 2.01])",
# align=c("l","c","c","c"))
```
<!-- ### Contiguity and War
A comment about contiguity and war onset.
- A few "big ones" are going to account for this (e.g. Gulf War, Iraq War, world wars)
- Controlling for those, contiguity accounts for 61% of all pre-WWII wars and 45% of post-WWII wars.
There are still quite a few non-contiguous wars, like:
- Spanish-American War
- Vietnam War
- Italo-Turkish War
- Boxer Rebellion -->
## Contiguity as Opportunity
### Why Are Contiguous States Prone to Conflict?
1. Opportunity
2. Interactions/Willingness
3. Territory
We'll discuss the first two today.
### The Opportunity Argument
Contiguous states have more opportunity for conflict.
- Think of this as a "reach" argument.
- Also an argument of necessity.
How can we measure "opportunity?"
## Measuring Opportunity
### Measuring Opportunity
1. Regional dyads
2. Political relevant dyads
3. Politically relevant information environment (PRIE)
4. Political activity
### Irrelevant Dyads
This ultimately becomes an issue of a sampling frame with both substantive/statistical concerns.
- Substantive: why bother estimating the probability of conflict between Mongolia and Nigeria?
- Statistical: flooding analysis with "irrelevant" observation artificially deflates standard error.
- i.e. it makes miniscule/unimportant effects seem "significant."
### Irrelevant Dyads
Our goal: devise a sampling frame that includes *all* dyads that could have or have had at least one MID.
- That's how we will get "opportunity" as "necessary condition" (i.e. without it, a MID cannot happen).
### Regional Dyads
Earliest attempt at measuring opportunity came in Bueno de Mesquita's "regional dyads."
- Relied on CoW's classification system.
### Correlates of War Regions
| **Region** | **`ccode` Domain** | **Examples** |
|:-----------|:------------------:|:------------:|
| Americas | [2, 200) | USA (2), CAN (20), MEX (70), ..., URU (165) |
| Europe | [200, 400) | UKG (200), IRE (205), ..., ICE (395) |
| Africa | [400, 600) | CPV (402), STP (403), ..., SYC (591) |
| ME/NA | [600, 700) | MAR (600), ALG (615), ..., OMN (698) |
| Asia | [700, 900) | AFG (700), TKM (701), ..., TLS (860) |
| Oceania | [900, 999) | AUS (900), PNG (910), ..., SAM (990) |
###
```{r regions-mids, echo=F, eval=T}
Data %>%
mutate(region = ifelse(region1 == region2, "Same Region", "Different Region")) %>%
group_by(midonset, region) %>% summarize(n=n()) %>% group_by(region) %>%
mutate(freq = round((n / sum(n))*100, 2),
freq = paste0(freq, "%")) %>% select(-n) %>% ungroup() %>% spread(region, freq) %>%
rename(`**MID Onset**` = midonset) %>%
kable(.,caption="Regional Dyads and MIDs (GML MID Data [v. 2.02])",
align=c("l","c","c"))
```
### How to Read This Table
Goal: we want that "different region" and MID quadrant to be 0%.
- If regions are an adequate sampling frame, there would be *no* cross-region MIDs *ever*.
What we find: .19% of all dyad-years had MIDs between two combatants from different regions.
- Not bad. But...
### The Problem of Regions
Regions are arbitrary and miss a lot of detail.
- e.g. Canada and St. Lucia share a region, but the probability of a MID is zero.
- Russia-Turkey is the most war-prone dyad and are incidentally coded in different regions.
- We're going to miss just about everything interesting the U.S. has been doing since the 1900s.
### Political Relevant Dyads
Weede's "politically relevant dyads" offer a refinement, whereby "political relevance" includes:
- Contiguous dyads of any type, and/or:
- At least one "major" power in the dyad.
###
```{r list-majors, eval=T, echo=F}
Majors %>%
select(-ccode, -version) %>%
kable(.,
caption="The Major Powers, 1816-2016 [CoW, v. 2016]")
```
###
```{r prds-mids, echo=F, eval=T}
Data %>%
mutate(prd = ifelse(prd == 1, "Politically Relevant", "Politically Irrelevant")) %>%
group_by(midonset, prd) %>% summarize(n=n()) %>% group_by(prd) %>%
mutate(freq = round((n / sum(n))*100, 2),
freq = paste0(freq, "%")) %>% select(-n) %>% ungroup() %>% spread(prd, freq) %>%
rename(`**MID Onset**` = midonset) %>%
kable(.,caption="Politically Relevant Dyads and MIDs (GML MID Data [v. 2.02])",
align=c("l","c","c"))
```
### Assessing Political Relevance
Political relevance better captures opportunity than regions. Yet:
- It's still missing several prominent cases (e.g. Israel, Iraq)
- Not all "majors" have the "reach" of the U.S. (see: Austria-Hungary, Germany, Japan).
### Politically Relevant Information Environments (PRIEs)
Maoz (1996) offers a refinement of political relevance with PRIE. Namely:
- Disaggregating majors to regional/global.
- e.g. USA was almost always a regional power before Spanish-American War.
- Qualifies "major" status for several "majors."
- e.g. Japan is a "major" power, but only in Asia.
- Same for Austria-Hungary in Europe.
- Russia is a global major only for Cold War, a regional major in Europe/Asia every other time.
###
| **MID Onset** | PRIE = 0 | PRIE = 1 |
|:--------------|:--------:|:--------:|
| 0 | 99.92% | 96.55% |
| 1 | 0.07% | 3.45% |
Table: PRIEs and MIDs (Quackenbush, 2006)
PRIE does about the same as political relevance and it's still imperfect.
### Political Activity
Quackenbush introduces "politically active dyads" as a refinement of these measures. Codes activity if:
- Dyads are contiguous, directly or through colony.
- One of the dyad members is a global power.
- One of the dyad members is a regional power in the region of the other.
- One of the dyad members is allied to a global power. Or:
- One of the dyad members is allied to a regional power in region of the other.
###
| **MID Onset** | Politically Inactive | Politically Active |
|:--------------|:--------:|:--------:|
| 0 | 99.96% | 98.54% |
| 1 | 0.03% | 1.46% |
Table: Politically Active Dyads and MIDs (Quackenbush, 2006)
###
| **Presence of..** | **Necessary for...** | **K/N = e(p)** | **P_{i}** | **P_{ii}** |
|:------------------|:--------------------:|:--------------:|:---------:|:----------:|
| Political Relevance | MID | 357/3002 = .12 | .109 | .00 |
| PRIE | MID | 425/3002 = .14 | .131 | .00 |
| Regional Dyad | MID | 248/3002 = .08 | .074 | .00 |
| Political activity | MID | 150/3002 = .05 | .043 | .00 |
Table: An Empirical Test of Necessary Conditions (Quackenbush, 2006)
# Conclusion
### Conclusion
Contiguity is the most important correlate of conflict.
- More for MIDs, less for war.
- Common (mis)perception: contiguity proxies opportunity/"reach".
- Isolating opportunity may be best done through measuring political activity.
- At the least, don't flood your sampling frame with irrelevant cases (e.g. Mongolia-Nigeria).
You can’t perform that action at this time.