Thread by @ProfPaulPoast

1/ The recent Pakistan-India dispute should renew interest in studying deterrence.

But studying deterrence is hard, as it requires making an inference w/ "observational data" 🤔

We discussed this issue in my Quantitative Security class!

👇is a summary

2/ A good starting point for understanding the difficulty of studying deterrence is this 1984 @World_Pol piece by Russett and Huth

cambridge.org/core/journals/…

3/ A key finding of their paper pertains to allies.

They find that having an ally does NOT enhance deterrence

4/ However, they only looked as cases of "immediate deterrence": cases where a threat has already been made, such as those in this table

5/ But a state with a highly credible ability to repel an attack is likely to NOT be threatened in the first place.

This is something they acknowledge...

5/ ... and then go on to dismiss

6/ Along comes Fearon in this 1994 Journal of Conflict Resolution piece

journals.sagepub.com/doi/abs/10.117…

7/ He does not think the "selection bias" can be dismissed and that accounting for it will have key implications for who we think about deterrence

8/ Indeed, footnote #7 takes a small dig at Huth and Russett for "hand waving" the issue

9/ So how does Fearon address it?

He doesn't.

Instead, he just offers a different interpretation of Huth's existing data.

10/ Consider Fearon's finding regarding the deterrent effect of a nuclear ally.

He presents Table 4 below and says that, in the instances of a nuclear power's protege being threatened, deterrence "success" is closely associated with the defender having nukes.

11/ This is all well and good, but look very carefully at the table: it is still just cases where a challenger threatened a protege. Again, he's using the same data as Huth and Russett.

12/ While the literature grapples with this issue over the subsequent years, it's not until @BAshleyLeeds in this 2003 @AJPS_Editor piece that someone properly evaluates "general deterrence" (i.e. does a challenger make a threat in the first place?)

onlinelibrary.wiley.com/doi/pdf/10.111…

13/ Leeds looks at all "potential cases" of dispute onset, rather than looking only at crises that have begun. Leeds finds that having a defensive ally makes a state dramatically less likely to be threatened

14/ Of course, Leeds acknowledges that she did not directly evaluate both stages of a crisis: entering a crisis AND what happens once you enter the crisis

15/ Such an evaluation wouldn't happen until 2011, with this @FPA_Jrnl piece by @BAshleyLeeds & Jesse Johnson

academic.oup.com/fpa/article-ab…

16/ Their piece then sparked a debate b/w them and others in a 2017 @The_JOP Scholarly Dialogue

17/ By this point, we have all sorts of methodological approaches being applied to the data in order to account for selection: various matching algorithms; censored probits; and more.

You know: SCIENCE!!

18/ By the end of class, my students saw

(1) the literature has come a long way from the initial Russett and Huth paper,

(2) the literature still has a ways to go!

19/ Such is the nature when studying some of the key security questions: until IRB allows me to randomly assign countries to @NATO, we have to work with observational data!

(end)