I’ve blogged in the past about the challenges of both qualitatively and quantitatively measuring civil wars and “armed politics.” Right now, I’m having fun with cross-national civil war datasets as part of a project on leftist insurgencies in democracies (mixing Sri Lanka JVP micro-data with cross-national cases). Which requires figuring out which insurgencies were leftists, when they started, and which regimes were democracies in what years. I’ll bracket the latter question (anyone who’s tried to figure out what 1950s Pakistan counts as will be familiar with the vexations that can await).
I’ve been focusing on South and Southeast Asian cases, with Latin America up next.
Why can this be tricky? Well, take the Communist Party of Thailand (CPT). Much of the literature says it began in 1965 and Kalvyas and Balcells (2010 – data and appendix linked to here) code 1966 – but UCDP says 1974 was the year of onset. The Polity regime score of Thailand in 1965 is -7 (autocracy), but in 1974 it’s +3 (some kind of partial democracy or anocracy or whatever you want to call it). So the year of onset matters hugely if you are trying to figure out whether this was a revolt against a democratic-ish regime. Plus, telling a Thailand expert that the CPT revolt began in 1974 would probably be met with a incredulous look and then a dismissive eye roll about quantitative American social science.
Thomas and Wood (2017 – data here) code the CPP/NPA insurgency in the Philippines as a Marxist rebellion, but don’t list the 1946-1954 Huk rebellion because they focus on post-1979 cases. Kalyvas and Balcells (2010) agree on the CPP, but explicitly code the Huk uprising as a non-Marxist rebellion – even though many case experts would say was a Marxist insurgency. So who knows what the Huks were – do they count, or not? I’d say yes, but it’s a judgment call rather than a straightforward exercise.
UCDP says the CPP/NPA began in 1969 but Kalyvas and Balcells (2010) say 1972. Only three years off – but the Polity score of the Philippines in 1969 was +2 and in 1972 it was -9 (Marcos’ dictatorship). Either the CPP/NPA took on a semi-democratic regime – or it rebelled against a deeply authoritarian government (a similar quandary as with the CPT).
UCDP doesn’t have anything for Bangladesh in 1972-1974, when it was coded as an 8 in Polity (highly democratic). But the case literature points to the Jatiya Samajtantrik Dal (JSD) as a leftist insurgency during precisely this period. So if we put the JSD in, we have a leftist insurgency against a democracy. If we follow UCDP, there is nothing to see here. Once we add the Maoist Purba Banglar Sarbahara Party (PBSP) during the exact same period (the chaos leading up to Mujibur Rahman’s assassination in 1975), things get even wilder – not in UCDP, but definitely in case literatures. And both are leftists, but don’t appear in Kalyvas and Balcells (2010). We don’t know how many people (sources say things like “over two thousand”), or at whose hands, died during this period, so it’s little surprise it hasn’t shown up in datasets. But that doesn’t mean it didn’t happen or definitely doesn’t meet the relevant coding criteria.
India’s Naxalites are coded as getting going in 1969 in UCDP, 1967 in the case literature, and 1989 in Kalyvas and Balcells (2010, see Table A3). So either at the passionate height of Sino-Soviet split in the international communist movement – or in the final year of the Cold War when the whole Marxist thing was going out of style.
UCDP says the second JVP revolt in Sri Lanka kicked off in 1989. But the case literature and Kalyvas and Balcells (2010) say 1987. Who cares about two years? Well, Sri Lanka in 1987 was the year of the arrival of the IPKF, JR had been the president for a decade, and there hadn’t been a parliamentary election since 1977. In 1989, Premadasa was president, the IPKF was starting to back for the exits after having been around for a little while, there was an early-year parliamentary election – and the JVP’s strategy was totally different than in 1987. In 1989, it’s total war, total onslaught especially by the summer, targeting especially the security forces. In 1987, it was a much more selective focus on the UNP cadres while leaving open the possibility of some kind of deal (on JVP take-over terms) with leftist mainstream political forces. So your story about onset in 1989 would have to be very different than for 1987.
To be clear, there is also a ton of important agreement across datasets and case literatures. Everyone agrees that the CPB began its revolt in 1948 Burma, that the CPN-Maoist started its rebellion in 1996 Nepal, the Huks in 1946 Philippines, and the first JVP revolt in 1971 Sri Lanka. And many of the disagreements come down to specific, defensible coding criteria that come with many good things (UCDP knows its emphasis on verifiable battle death thresholds will lead to undercounting of particular types of phenomena, for instance – it’s a trade-off with the clear upside of comparability across cases). And I’m using all of this stuff for my more medium-/small-N-ish research project.
So this isn’t a cri de coeur against cross-national datasets, at all. But since we’re not dealing with many cases and we are grappling with very complicated questions, getting under the dataset hood is really important. Pulling stuff off the shelf to mix-and-match, by contrast, may not always be wise, since coding decisions can have inferential consequences.