Sri Lankan Army KIA, 1981-1999

by Dr. Paul Staniland,, October 10, 2016

Below I posted about a source I’d found with micro-data on SLA combat fatalities. The OCR-ing has proven more of a challenge than I had hoped, but the information on date and place is inputted for all observations, though we still need to add many of soldiers’ names and do extensive cleaning on names of places, how to aggregate them into districts, etc.  I think, combined with the qualitative research, this will also help us generate a clear order of battle for the 1980s/1990s SLA: where which units were, when, conditional on taking at least one fatality.

Preliminarily, this graph shows variation over time in SLA KIA:


The huge spikes are not surprising – the conventional phases of the LTTE-SLA war in the early and late 1990s, with dips during the IPKF and then the 1994-5 ceasefire.

Here are the places where 100 or more SLA members were killed over this period:

Place Freq. Percent Cum.
Vavuniya 954 8.93 8.93
Palaly 875 8.19 17.12
Paranthan 647 6.06 23.18
Kilinochchi 555 5.20 28.38
Welioya 550 5.15 33.53
Mankulam 509 4.77 38.29
Elephant Pass 434 4.06 42.36
Batticaloa 404 3.78 46.14
Mannar 355 3.32 49.46
Pooneryn 286 2.68 52.14
Jaffna 206 1.93 54.07
Mullaittivu 199 1.86 55.93
Vettalaikerni 191 1.79 57.72
Trincomalee 185 1.73 59.45
Vavuniya-Mankulam 143 1.34 60.79
Valachchenai 120 1.12 61.91
Pulmoddai 111 1.04 62.95
Kanagarayakulam 106 0.99 63.95
Vavuniya-Omanthei 100 0.94 64.88

Not much surprising here, either – these are the main battle zones in and around the Tamil north and east.

So what is surprising? I’m intrigued by how *few* SLA casualties we see in the 1980s, even as the country was politically coming apart. Body counts aren’t necessarily a proxy for political disorder – throughout the contemporary and historical literature, you get a sense of a country and political system on the brink, but these are not particularly substantial death counts in comparative perspective, or compared to the size of the Sri Lankan security apparatus.

Also, 1988 and 1989 are pretty light on fatalities despite the raging JVP rebellion in the south. There are 305 KIA in those 2 years combined, with only Welioya, Mannar, Vavuniya, Trincomalee, Cheddikulum, and Matara having 8 or more KIA combined during those two years.  I’m going to be doing more on the spatial breakdown of the 1981-87 period with an eye on the LTTE’s rise and then 1987-89 period with a JVP focus.

So either the Army wasn’t that involved in anti-JVP operations compared to the police (data on which I continue to be in search of!) and pro-state paramilitaries, or they were involved but were not particularly vulnerable to JVP operations.  I continue to wonder where the claims about total death counts (~40,000) in the 1987-1990 period actually come from, though it’s clear that civilians overwhelmingly bore the brunt of the violence and their experiences are unlikely to recovered from history.

fullsizerenderSerendipitous research gold

My Sri Lanka project sporadically moves along in parallel to my more regular preoccupation with Chapter 4 of my book. I made a very lucky find recently – while wandering the footnotes of a pop military history of the Eelam wars, I found a reference to a book called Sri Lanka Army: 50 Years On (1999), which looked to be an official history of the Army. But it turned out that only three libraries in WorldCat own a copy (the Marines, the Army, and the Library of Congress).

I got a copy ILL-ed. . . .and lo and behold, in addition to a history of the army and its activities, the last ~200-odd pages are a Roll of Honour that lists the name, rank, unit, date, and place of death for over 10,000 SLA combat fatalities between 1981 and 1999. I told my RA’s to scan the whole book before we had to return it. Being less than thrilled with scanning 1,000 pages, they went looking for a copy to buy. Even though Amazon showed no available results when searching by name, for some reason one came up when they searched by ISBN. A random used bookstore had a copy which we got for $7 (then refunded when delivery got messed up).

Initial scanning and OCR suggests we’ll be able to turn the combat fatality data into a dataset fairly easily, which will allow a new mapping of SLA deaths by date and place during the LTTE and JVP revolts of the 1980s (it’s less interesting post-1990, when it’s a series of conventional clashes with the Tigers). There’s also a lot of useful nitty-gritty security studies stuff on organizational structures and army combat operations.

Of course, the Sri Lankan Police have not responded to my ~12 emails to them about police data, so you can’t always get what you want, but some footnote digging and luck can open unexpected new doors.


Where were the LTTE & JVP strongest in 1988?

Districts of Sri LankaI’m in the process of trying to recreate the two JVP rebellions in Sri Lanka. One admittedly rough measure I’m using of armed group strength is a group’s ability to drive down turnout when they call for elections to be boycotted (caveats abound – many things drive turnout, but less then 30% turnout is not normal for a Sri Lankan election).

The 1988 Presidential election provides one way to get a sense of the geographic patterns of armed group strength (the all-Sri Lanka turnout average was 55.32% in 1988). Neither the LTTE nor the JVP wanted voters involved in these elections; the JVP in particular was unleashing a total war against UNP politicians and supporters during this period (only mid-way through the war did they go after the security forces). Here’s a map of Sri Lankan electoral districts, courtesy of Wikipedia:


In Tamil/partly-Tamil areas, we see Vanni electoral district down at 13.79% and Jaffna at 21.72%, with Trincomalee and Batticaloa districts coming in at 58.48% and 53.81%. The IPKF and EPRLF would be able to drive up those numbers for the 1989 parliamentary elections, but Jaffna and Vanni were clearly “abnormal” in 1988 and comparatively in 1989 as well. The more ethnically mixed composition of Trinco and Batti, the greater presence of Indian-linked Tamil paramilitaries in those areas, and initial northern social core of the LTTE are all reflected in these differences.

The Sinhalese areas are quite interesting. Hambantota, Matale, Matara, Moneragala, and Polonnaruwa are all below 30% turnout, while Colombo, Kandy/Mahanuwara, Kegalle, Ratnapura, Puttalam, Ampara/Digamadulla, Nuwara-Eliya, and Gampaha are all over 68%. The JVP’s impact on the Sinhalese heartland was spatially very heterogeneous: Kandy neighbors Matale and Ratnapura neighbors both Matara and Hambantota. We need to understand how seemingly-similar rural Sinhalese districts ended up taking such radically different political/military trajectories.

There is much more I will do with these (and other) elections to try to understand this divergence. But for now, this gives a crude but useful overview of where these two conflicts were likely the most intense in 1988 Sri Lanka’s cauldron of violence.

Another JVP Map: 1988 and 1989 elections

August 18, 2016

The JVP in Sri Lanka continues to be my “fun” project, a distraction from a book and multiple APSA papers. In future posts I’ll list a couple more sources I’ve found on this conflict. I also want to discuss the PDP in Kashmir and parties like it: “loyalist” parties on insurgent peripheries.

Here’s a map following up on my post about 1988’s election and what it might tell us about JVP (and LTTE) influence: the logic is that both groups wanted elections boycotted, so turnout levels provide a very rough proxy for where the groups were best able to exert their will.

This is district level data from 1988’s Presidential and 1989’s Parliamentary elections. In future, I’ll map out the 160 polling division results for each election to systematize substantial within-district variation. Unfortunately, polling divisions don’t line up with the main sub-district administrative unit, Division Secretary’s Divisions, so we can’t easily line up administrative data with disaggregated election data. Still, it’ll be a more useful map once I someday get around to it. I similarly want to get at the vote shares of different parties in these elections, plus go back to the 1982 presidential election.

You should play with turning the layers on and off; they are aligned right on top of one another so viewing them simultaneously doesn’t add anything (the top layer is the third layer below, % change in turnout by district from 1988 to 1989); click the weird arrow/drawer button on the top-left to toggle the layer options.

The first layer (blue) is the December 19, 1988 turnout percentages, just using the main town in a district as the location (I use Vavuniya as an arbitrary location for the Vanni electoral district). We see what I noted below: a low-turnout crescent from the deep south around through Uva province, into the north-center and north. Note the similarities in the deep south to the 1971 map of attacks I posted, but also a big difference: the Colombo-Kandy line appears much less hit by the JVP in 1988 than in 1971 (Kegalle was a hotbed of the 1971 revolt).

The second layer, in red, is 1989 parliamentary turnout, from February 15 (just 2 months after the presidential). We see overall major upticks in turnout, likely driven by a combination of the new, vigorous President Premadasa replacing JR and heightened security force operations in response to the JVP also targeting security forces more intensively. Overall turnout was up ~15% from 1988 (from 55% to 64%). If you click on the icon, you can see the raw percentages differences.

The third layer, in yellow, is the percentage change between 1988 and 1989. I’m not sure what to make of this: did the security situation really change so much in the space of a couple months? Did parliamentary elections drive turnout in different ways than presidential? A lot remains to be figured out.

Nevertheless, some striking stuff. Both the Vanni and Jaffna districts saw huge increases, though from low 1988 values. This is the IPKF buckling down and trying to get rid of the LTTE and prop up the EPRLF and its TULF coalition. However, EROS vote share – as “independents” in Vanni and Jaffna – may indicate support for the Tigers, so turnout alone doesn’t do the job here as a proxy for group influence. Eliyathamby Ratnasabapathy was a big vote-getter in Jaffna, and also happened to be a founder of the (by 1989) LTTE-aligned EROS; Eliyathamby Pararasasingam was another MP from EROS. Batticaloa and Trincomalee move upwards as well, from much higher baselines. These are where the EPRLF tried to make its political stand under the TULF banner.

There is some very interesting stuff in the primarily-Sinhalese districts. Massive increases in Matala, Moneragala, and Badulla, and substantial increases in Galle, Polonnaruwa, and Kurunegala. Something is shifting quite radically in these districts. Crudely, this seems like it might be a useful way of identifying where government control is returning, especially the districts with tripling or doubling of turnout.

Yet we see actual decreases in Hambantota, Matara, and Kandy. Matara and Hambantota are commonly identified as the heart of the JVP, so it’s no surprise to see that the situation does not seem to have improved at all in those districts: both remain ~20%. Interestingly, these are two of the lowest population density districts in Sri Lanka (and were in 1986 as well based on a Census estimate I found in Chicago’s library). Some of the literature also suggests that the JVP moved north as it was being hit elsewhere, which may help explain the the 18% drop in Kandy, but that’s speculative given increases or stasis around it.

This variation leaves a lot of big questions unanswered. But it valuably gives us places to look in the specialist qualitative literature (which is bigger and better than I initially thought, so though still small) for shifts in counterinsurgency strategy, security force deployment, political party campaigning, JVP strategy and operations, and other possible influences, rather than solely relying on broad macro-level claims about the conflict.

Posted .

Filed under Military.

Leave a Reply

Comment Guilelines Critical is fine, but if you’re rude, we’ll delete your stuff. No personal attacks.

  • (will not be published)