Permalink
Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
454 lines (341 sloc) 16.3 KB
---
title: "The Scientific Study of War"
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]+)$', '\\1', x)
if (file.rename(x, x2)) x2 else x
}
)
```
```{r loadstuff, include=FALSE}
knitr::opts_chunk$set(cache=FALSE)
library(tidyverse)
library(stevemisc)
library(knitr)
library(mirt)
library(QuickUDS)
MB <- read_csv("~/Dropbox/data/military-balance/2015/mb2015.csv")
NMC <- read_csv("~/Dropbox/data/cow/cinc/NMC_5_0.csv")
Polity <- readxl::read_xls("~/Dropbox/data/polity/p4v2016.xls")
logmin <- function(var){
if (min(var, na.rm = TRUE) == 0) {
result <- as.integer(log(var+1))
result <- result - min(result, na.rm = TRUE)
}
else {
result <- as.integer(log(var))
result <- result - min(result, na.rm = TRUE)
}
}
MB$ldefbudg <- logmin(MB$defbudg)
MB$ldefbudg <- with(MB, logmin(defbudg))
MB$lpop <- logmin(MB$pop)
MB$lact <- logmin(MB$active)
MB$lres <- logmin(MB$reserve)
MB$larmy <- logmin(MB$army)
MB$lnavy <- logmin(MB$navy)
MB$laf <- logmin(MB$af)
MB$lpara <- logmin(MB$para)
MB$larmyequip <- logmin(MB$armyequip)
MB$lnavyequip <- logmin(MB$navyequip)
MB$lafequip <- logmin(MB$afequip)
MB$otherrank <- NA
MB$otherrank[MB$otherequip == 0] <- 0
MB$otherrank[MB$otherequip == 1] <- 1
MB$otherrank[MB$otherequip >= 2 & MB$otherequip < 10] <- 2
MB$otherrank[MB$otherequip >= 10 & MB$otherequip < 50] <- 3
MB$otherrank[MB$otherequip >= 50 & MB$otherequip < 100] <- 4
MB$otherrank[MB$otherequip >= 100] <- 5
MB$nukes <- 0
MB$nukes[MB$country == "United States"] <- 6
MB$nukes[MB$country == "United Kingdom"] <- 5
MB$nukes[MB$country == "Russia"] <- 6
MB$nukes[MB$country == "Pakistan"] <- 5
MB$nukes[MB$country == "Israel"] <- 4
MB$nukes[MB$country == "India"] <- 5
MB$nukes[MB$country == "France"] <- 5
MB$nukes[MB$country == "China"] <- 5
sub <- with(MB, data.frame(country, year, ldefbudg, lpop, lact, lres, larmy, lnavy, laf, lpara, larmyequip, lnavyequip, lafequip, otherrank, nukes))
model <- mirt(sub[ , 3:ncol(sub)], model = 1,
itemtype = "graded", SE = TRUE, verbose = FALSE,
technical = list(NCYCLES = 5500))
scores <- fscores(model, full.scores = TRUE, full.scores.SE = TRUE)
trial <- cbind(sub,scores)
library(sqldf)
sqldf("select country, year, F1, SE_F1 from trial")
trial$ub <- with(trial, F1 + (1.96)*SE_F1)
trial$lb <- with(trial, F1 - (1.96)*SE_F1)
# trial <- transform(trial,LOCATION=reorder(LOCATION,int))
trial <- trial[order(trial$F1), ]
measures <- c("pmm_arat", "blm", "bmr_democracy",
"bnr_extended", "pmm_bollen", "doorenspleet",
"wgi_democracy","fh_total_reversed",
"fh_electoral", "gwf_democracy_extended",
"pmm_hadenius", "kailitz_tri",
"lexical_index", "mainwaring",
"magaloni_democracy_extended",
"pmm_munck", "pacl", "PEPS1v",
"pitf", "polity2", "reign_democracy",
"polyarchy_original_contestation",
"prc", "svolik_democracy",
"ulfelder_democracy_extended",
"utip_dichotomous_strict", "v2x_polyarchy",
"vanhanen_democratization", "wth_democ1")
# Generate Extended UDS
extended_model <- democracy_model(measures,
verbose = FALSE,
technical = list(NCYCLES = 2500))
extended_scores <- democracy_scores(model = extended_model)
```
# Introduction
### Goal for Today
*Discuss what we mean by "science" and how we can relate it to the study of war.*
# What Is Science?
### What Is Science?
It's easy to misuse the term "science", including shorthand for:
- Things that seem cool but we don't understand
- TED talks that are almost always bad
- An acquired repository of "facts"
- Technology
- An umbrella term for STEM stuff (i.e. "the hard sciences")
- Whatever it is that Neil Degrasse Tyson talks about.
Science is ultimately none of that.
### What Is Science?
Science is a *method*, not a discipline.
- Informally: test assumptions against observable evidence in a systematic manner.
- Each step is made explicit.
Other components:
- Scientific method is ultimately value-neutral.
- It seeks generalizations, not to explain specific observations.
For shorthand, we call this the empirical implications of theoretical models (EITM) or hypothetico-deductivism.
###
> The method of the book, then, is to develop the deductive implications of the four basic axioms for a given, highly specific set of conditions; review evidence indicating whether or not these implications are empirically correct; and present new evidence as necessary and possible to resolve outstanding empirical questions (Zaller 1992, 51).
## H-D and EITM
### Hypothetico-Deductivism (H-D)
Kyburg (1988, 65) outlined the H-D model.
- A hypothesis *H* is set up for testing or examination.
- We deduce an observation sentence *O* from *H* and its necessary qualifiers, boundary conditions, etc.
- We test *H* with an experiment or examination and observe either *O* or ~*O*.
If ~*O*, then ~*H*. If *O*, we fail to refute *H*.
### Hypothetico-Deductivism (H-D)
Replace *H* with theory *T* and we have the standard political science model (Hausman 1992, 304).
1. *Formulate* some hypothesis *H* from theory *T*.
2. *Deduce* prediction *P* from *H* with necessary qualifiers (e.g. "ceteris paribus").
3. *Test* *P*.
4. *Judge* whether *H* is confirmed or disconfirmed, contingent on *P* or ~*P*.
### Empirical Implications of Theoretical Models (EITM)
EITM is shorthand for this approach, and a push across political science.
1. Start with theory, informed by case study, field work, or "puzzle".
2. Outline model establishing causal linkages.
3. Stipulate deductions and hypotheses.
4. Outline measurement and research design to test deductions.
5. Collect and analyze data.
The results are inductively compared to the theory and its hypotheses.
### What Causes War?
Here is an example puzzle: war is costly and ex post inefficient. Why do states fight it?
- Intuition: wars happen because states cannot commit to pre-war solution.
- Further intuition: democracies, *can* credibly commit and thus avoid war.
- Deduction/hypothesis: if both sides in a dispute are a democracy, they will not fight a war.
- Research design: all dyadic (i.e. A vs. B) disputes.
- Dependent variable: did dispute escalate to war (yes or no)
- Independent variable: were both sides democratic (yes or no)
- Collect and analyze data: hello, regression...
## Measurement Concerns
### Elements of Scientific Theories
At the risk of spoiling a methods class, some terms to clarify:
- *Dependent variable* (DV): something we want to explain
- *Independent variable*: something we believe explains variation in the DV.
- *Conceptual definition*: outlining what a term we use is.
- Spoiler alert: *all* political science is conceptual (e.g. "democracy", "war", "allied", "power", "security")
- *Operational definition:* an actual implementation of the conceptual definition
### An Illustration of an Operational Definition
Democratic peace theorists argue "democracies" are more peaceful than other state types.
- Simple enough, but how do you measure it?
### The Polity Project
The Polity project is the longest-running data set on democracy, measuring "democracy" as:
- Competitiveness of executive recruitment
- Openness of executive recruitment
- Constraints on chief executive
- Competitiveness of political participation
This creates a 21-point scale [-10, 10] of "democracy" largely as measure of executive constraints.
###
```{r turkish-democracy-polity-1800-2016, echo=FALSE, eval=TRUE, warning=FALSE, error=FALSE,fig.width = 14, fig.height = 8.5}
Polity %>%
filter(ccode == 640) %>%
ggplot(.,aes(year, polity2)) + theme_steve_web() +
geom_line(size = 1.1) +
scale_x_continuous(breaks = seq(1800, 2020, by = 10)) +
scale_y_continuous(limits = c(-10, 10),
breaks = seq(-10, 10, by = 2)) +
geom_hline(yintercept = 6, linetype="dashed") +
xlab("Year") + ylab("Polity2 Score") +
annotate("text", family = "Open Sans", x= 1876, y = -3,
label = "Ottoman\nConstitution\n(1876)", size=3) +
annotate("text", family = "Open Sans", x= 1908, y = 0,
label = "Young Turk\nRevolution\n(1908)", size=3) +
annotate("text", family = "Open Sans", x= 1946, y = 8,
label = "Direct\nElection\n(1946)", size=3) +
annotate("text", family = "Open Sans", x= 1960, y = 3.25,
label = "Coup\n(1960)", size=3) +
annotate("text", family = "Open Sans", x= 1971, y = -2.5,
label = "Coup\n(1971)", size=3) +
annotate("text", family = "Open Sans", x= 1980, y = -5.5,
label = "Coup\n(1980)", size=3) +
annotate("text", family = "Open Sans", x= 2014, y = 10,
label = "Election of\nErdogan\n(2014)", size=3) +
annotate("text", family = "Open Sans", x= 1919, y = -9,
label = "Post-WWI\nInterregnum", size=3) +
annotate("rect", xmin=1917,
xmax=1922, ymin=-Inf, ymax=Inf, alpha=.3, fill="red") +
annotate("text", family = "Open Sans", x= 1800, hjust=0, y= 6.5, size=3,
label= "Conventional cutoff for democracy (polity2 = 6)") +
labs(title = "A History of Turkish Democracy in the Polity Data, 1800-2016",
subtitle = "The trajectory of Turkish democracy is always fledgling, replete with coups, and is in a current state of a clear regression to authoritarianism under Erdogan.",
caption = "Data: Polity Project, Center for Systemic Peace (v. 2016)")
```
###
```{r american-democracy-polity-1800-2016, echo=FALSE, eval=TRUE, warning=FALSE, error=FALSE,fig.width = 14, fig.height = 8.5}
Polity %>%
filter(ccode == 2) %>%
ggplot(.,aes(year, polity2)) + theme_steve_web() +
geom_line(size = 1.1) +
scale_x_continuous(breaks = seq(1800, 2020, by = 10)) +
scale_y_continuous(limits = c(-10, 10),
breaks = seq(-10, 10, by = 2)) +
geom_hline(yintercept = 6, linetype="dashed") +
xlab("Year") + ylab("Polity2 Score") +
annotate("text", family = "Open Sans", x= 2016, hjust=1, y= 6.5, size=3,
label= "Conventional cutoff for democracy (polity2 = 6)") +
labs(title = "A History of U.S. Democracy in the Polity Data, 1800-2016",
subtitle = "The U.S. as maximally democratic through the duration of the data belies important variation in U.S. democracy through its history.",
caption = "Data: Polity Project, Center for Systemic Peace (v. 2016)")
```
###
```{r american-democracy-quickuds-1789-2017, echo=FALSE, eval=TRUE, warning=FALSE, error=FALSE,fig.width = 14, fig.height = 8.5}
extended_scores %>%
filter(cown == 2) %>%
ggplot(.,aes(x=year, y=z1, ymin=z1_pct025, ymax=z1_pct975)) + theme_steve_web() +
geom_ribbon(alpha = I(0.2), color = "#000000", fill="#619cff") +
geom_line(size=1.1, color="#619cff") +
scale_x_continuous(breaks = seq(1790, 2020, by = 10)) +
xlab("Year") + ylab("Latent Democracy Estimate") +
labs(title = "A History of American Democracy, 1789-2017",
subtitle = "The estimates from the graded response model tell a more faithful story of American democracy that capture gradual enfranchisement, civil rights movements, and the current backsliding.",
caption = "Data: QuickUDS (via Xavier Marquez, c.f. Pemstein et al. [2010])")
```
### An Illustration of an Operational Definition
Stam (1996) argues "military quality" explains why one side wins a war.
- Simple enough, but how do you measure it?
How about the overall size of the military?
###
```{r, echo=F, eval = T}
NMC %>%
filter(year == 2012 & milper != 0 & milex != 0) %>%
mutate(`Military Quality` = milex/milper,
Country = countrycode::countrycode(ccode, "cown", "country.name"),
Country = ifelse(Country == "United States of America", "United States", Country),
Country = ifelse(ccode == 200, "United Kingdom", Country),
Country = ifelse(ccode == 816, "Vietnam", Country)) %>%
rename(Year = year,
`Military Personnel` = milper) %>%
select(Country, Year, `Military Personnel`) %>%
arrange(-`Military Personnel`) %>% head(10) %>%
kable(., align=c("l","c","c"),
caption="Largest Militaries (in Thousands) in the World (Data: CoW-NMC, 2012)")
```
Does this look right?
### Stam's Argument
Body count doesn't get at "quality."
- If it did, we would've expected Iraq to hang w/ the U.S. much better than it did in 1990.
Operationalization of "military quality": military spending divided over military personnel.
###
```{r, echo=F, eval = T}
NMC %>%
filter(year == 2012 & milper != 0 & milex != 0) %>%
mutate(`Military Quality` = milex/milper,
Country = countrycode::countrycode(ccode, "cown", "country.name"),
Country = ifelse(Country == "United States of America", "United States", Country),
Country = ifelse(ccode == 200, "United Kingdom", Country)) %>%
rename(Year = year) %>%
select(Country, Year, `Military Quality`) %>%
arrange(-`Military Quality`) %>% head(10) %>%
kable(., align=c("l","c","c"),
caption="Highest Troop Qualities in the World (Data: CoW-NMC, 2012)")
```
This isn't perfect. So how about this:
###
```{r latent-military-capabilities-2015, echo=FALSE, eval=TRUE, warning=FALSE, error=FALSE,fig.width = 14, fig.height = 8.5}
ggplot(trial[170:161,],aes(F1,reorder(country, F1),xmin=lb,xmax=ub))+
geom_errorbarh(height=0)+
geom_vline(xintercept=0,lty=2)+
geom_point()+facet_wrap(~year,scale="free_x") +ylab("Country") + xlab("Estimated Latent Military Capacity") + theme_steve_web() +
labs(title = "Steve's Toy Measure of Latent Military Capability",
subtitle = "There's a fair bit of face validity here: the U.S. is discernibly stronger than every country in the world except China.")
```
### If You're Curious About What I Did...
It's an item response model of:
- size of defense budget
- size of population
- size of active personnel
- size of reserve personnel
- size of active army personnel
- size of active navy personnel
- size of active air force personnel
- size of active paramilitary/irregulars
- number of tanks, recon equipment, <!--AIFVs, APCs, --> armored fighting vehicles, artillery
- number of subs, principal surface combatants, amphibious equipment
- number of combat-capable aircraft
- number of military satellites in space
- number of nukes (FAS)
### Reliability and Validity
This highlights the issue of reliability and validity in measurements.
- *Reliability*: does measure consistently capture what we want it to capture?
- *Validity*: does measure "capture" the concept in question without picking up unintended characteristics?
## Evaluating a Theory
### Deduction and Induction
Science is fundamentally *deductive* rather than *inductive*.
- Major emphasis: nothing can be learned from a confirming instance.
- Requires the future to be exactly like the past.
- The logic is also circular: induction is used to justify induction.
This does lead to something dissatisfying about the scientific method.
- We can't prove our theories *true*.
- At the most, we say our findings are consistent with the theory.
### Criteria for a Good Theory
How to evaluate theories:
- Logical consistency
- Accuracy of predictions
Another dissatisfying component here:
- "All models are wrong; some are useful."
# Conclusion
### Conclusion
Science is a method that we'll use to study what causes war. Some caveats:
- We seek to generalize, not explain specific events.
- This will involve a lot of regression. Get comfortable with it.
- Predictions are "consistent" and not "true."
- "All models are wrong; some are useful."
### PSA
For those new to the course:
- Go to http://posc3610.svmiller.com
- Lecture slides there as well.
- Check Canvas too. Know where the Turnitin link is.