Download presentation
Presentation is loading. Please wait.
Published byΠρίαμος Καζαντζής Modified over 6 years ago
1
Graphical Presentation for Robustness Checks AKA Pretty Pictures for Everyone!
Benjamin Graham
2
Today’s Plan Powersharing and Democratic Survival Varying the Lags
Hard Cases (aka subsetting the sample) Benjamin Graham
3
Powersharing and Democratic Survival
Three types of powersharing Inclusive, Dispersive, Constraining Which, if any, promote democratic survival? If you are a democracy in year t, what institutions cause you to still be a democracy in year t+5? Benjamin Graham
4
The Data 180 countries from 1975-2010
Index measures of each type of powersharing Observational data Benjamin Graham
5
Causal inference in this context
Benjamin Graham
6
19 page paper 41 page online appendix
7
Empirical Approach Sample = Democracies in year t
Dependent Variable: Regime type (binary) in year t+5 Logit regressions Benjamin Graham
8
Varying the Lags The choice of five years is arbitrary
So let me show you what happens with: 1 year lag 2 year lag ... 8 year lag 10 year lag Benjamin Graham
9
I could make a giant regression table
Benjamin Graham
10
Benjamin Graham
11
What do you need? Coefficients & standard errors from your variables of interest Organized into their own mini dataset Benjamin Graham
12
First set up the mini-dataset
drop if gwno == . scalar numobs = _N moreobs 10 gen lead = . replace lead = _n-numobs foreach var in constraining inclusive dispersive{ gen `var'_coef = . gen `var'_se = . } Benjamin Graham
13
Then fill in the values forvalues i=1(1)10 {
quietly logit F`i'.democracy constraining CONTROLS, robust replace constraining_coef = _b[constraining] if lead == `i' replace constraining_se = _se[constraining] if lead == `i' } Then drop everything else and save drop if lead == . keep lead constraining_coef constraining_se saveold varying_lags, replace version(13) Benjamin Graham
14
Benjamin Graham
15
Benjamin Graham
16
Hard Case Subsamples Some countries are at no risk of democratic failure Are low-risk countries driving our results? Evaluate subsamples of high risk cases Benjamin Graham
17
Hard Case Subsamples Recent civil war (3 flavors) High state fragility
New democracy Poor Recent financial crisis Ethnic divisions (2 flavors) Low predicted survival (our own models) Benjamin Graham
18
Benjamin Graham
19
Benjamin Graham
20
Benjamin Graham
21
And if you need help with any of this…
The slides and code are up on my website under “Useful Stuff” SPEC Lab stats consulting office hours And/or me: Don’t be shy! Benjamin Graham
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.