Leftist Insurgency in Democracies

In April 1971, revolutionary violence swept across much of Sri Lanka as the Maoist Janatha Vimukthi Peramuna (JVP) sought to seize power in a leftist uprising. Thousands of deaths later, the revolt was quashed. This war presents a puzzle: the United Front ruling government in 1971 had been supported by the JVP in the 1970 election, was ideologically socialist and moving policy to the left, and had not repressed or targeted the JVP until the revolt was about to break out. Indeed, the hotbeds of the JVP’s revolution were precisely those that had voted for the United Front less than a year prior.

The JVP’s rebellion challenges the expectations of important research like Huntington (1968) and Acemoglu and Robinson (2006): it was a leftist insurgency in a full-suffrage democracy under a center-left government, during a period of open political contestation and mobilization. The redistributive tilt of the ruling United Front (UF) government did not placate the JVP, which instead saw the UF as following a fruitless path to neutralized co-optation. Rather than democracy and political participation acting as a brake on revolutionary violence, violence surged precisely at a moment of mainstream leftist breakthrough.¹

1971 Sri Lanka is not unique: in Colombia (1964), India (1967), Burma (1948), the Philippines (1946 and 1968), Nepal (1996), Peru (1982), Laos (1958–9), and Bangladesh (1972), among others (such as Sri Lanka again in 1987), leftist insurgencies have emerged in periods of electoral democracy. These movements left a huge toll in blood and dislocation, some lasted for decades, and many had profound political consequences for the societies in which they arose.

This paper first uses medium-N data from South and Southeast Asia to show that existing claims about the relationship between democracy and leftist rebellion are open to empirical question. Rather than such rebellions primarily occurring against right-wing authoritarian governments or in response to repression, we see a number of wars that erupt under relatively open, full-suffrage democratic competition.

The paper then uses the 1971 JVP revolt in Sri Lanka as a theory-building case to outline a new mechanism of revolutionary insurgency under democracy. When internally-divided leftist movements face “incorporation windows” that force a hard choice between co-optation into mainstream electoral politics or rebelling against the political system itself, radical splinters are likely to break off in opposition to mainstream co-optation. This fragmentation can emerge from international diffusion, prior state efforts at division, and contingent organizational strategies adopted by leftist leaders. When an incorporation window emerges, radical left factions in fragmented environments both fear being frozen out by the mainstream left and see an opportunity to push pro-left momentum toward a seizure of power. This helps explain why a number of leftist revolts have been launched against democratic socialist and center-left elected government: indeed, some of the most loathed foes of communist revolutionaries have not been right-wing land-owners or oligarchs, but instead center-left parliamentary socialists, whether in 1919 Germany (McAdams, 2017), late 1960s West Bengal, or 1980s Peru (Ron, 2001). Material class alignments can collapse in the face of byzantine ideological and coalitional rivalries.

Drawing on the methodological approach pioneered in Haggard and Kaufman (2012), the paper examines the cases of leftist insurgency onset under democracy in post-1945 southern Asia to compare competing theories, allowing an assessment of which mechanisms provide insight into the actors, sequence, and political context of leftist revolts. Several of the cases provide clear support for the existing claim in the literature that democracies trigger revolt when they block political participation by the left. But a cluster of leftist insurgencies under democracy in southern Asia instead occurred during moments of leftward tilts in the polity, driven by feuding between factions of divided leftist movements over whether to accept mainstream incorporation in moments of electoral opportunity. These revolts emerged from tensions within the left over the prospects and perils of electoral politics. The research design more briefly explores two cases in which leftist revolt did not break out in democratic periods in southern Asia, and points to related dynamics in Latin America.

Why should we study these cases? First, revolutionary revolts in democratic systems directly contribute to broader debates in political science around redistribution, democracy, and revolution, since these debates often hinge on implicit or explicit theories of when redistributive revolts begin. By examining the actual conditions under which we see such insurgencies, this article shows the importance of giving greater analytical importance to revolutionary forces, in a corrective to the top-down, regime- and elite-centric accounts that currently dominate. “The masses” or “the poor” are analytical categories of limited use: instead, in many cases we see deep intra-class divisions along ideological lines.

Second, democratization and political incorporation put enormous pressure on political movements that must grapple with the perceived advantages and dangers of electoralism; the strategic dilemmas faced by leftist movements have far broader echoes, including in ethnic and separatist movements (Cunningham, 2014; Horowitz, 1985; Snyder, 2000). The findings here contribute to our understanding of when and how democracy co-opts potential insurgents, triggers their revolt, or does both simultaneously.

Finally, in a world of increasing inequality and political instability, conflicts driven by actors seeking to fundamentally reshape the foundations of political systems may remain relevant. Kalyvas (2015) has argued that Islamist rebels are the ideological successors of the Marxist-Leninist challenge to the Western order of the Cold War. The challenge of spiraling inequality confronts countries in both the developed and developing world; though unlikely to take the specific forms I study here, it is conceivable that we will see other armed revolutionary movements emerge.

Research on civil war has sidestepped leftist insurgency as a distinctive category of analysis. Many excellent case-based studies have been done of communist revolt, whether in Vietnam or Colombia, but leftist revolt has attracted little theoretical attention in recent years. While extensive work has gone into the specific roots of ethnic revolt, conflicts waged by Marxist-Leninist or Maoist armed groups are waged lumped into a broader category of center-seeking or “ideological” insurgencies (Sambanis, 2001).

Yet the political claims and coalitions of leftist insurgencies are quite different from those of many other rebels. Balcells and Kalyvas (2015) find that Marxist-Leninist rebels fought longer and bloodier wars than their contemporaries, while Wood and Thomas (2017) find that leftist groups are more likely to include female combatants. Wood (2001, 2000) shows that leftist revolts from below can have distinctive long-term effects on regime type. This category of revolt would benefit from closer study, as part of a broader process of disaggregating forms of civil wars and types of conflict onsets: as Sambanis (2001, p. 259) notes, “Not all civil wars are the same.”²

One important caveat, however, we can take from his literature involves scope conditions. Wealthy and well-institutionalized states rarely experience civil conflict. The scope condition here are countries that are structurally susceptible to civil war (Fearon and Laitin, 2003). This directs our attention to medium- and weak-capacity states as the domain of any kind of serious leftist insurgent revolt.

Hypothesis 2 provides a crucial insight. Yet the 1971 JVP revolt in Sri Lanka that opened this paper suggests a puzzle. The JVP had worked to help elect the United Front government of the left (Alles, 1976, p. 68), which pledged substantial nationalization and land reform. 1970–71 was a moment of left-wing breakthrough in Sri Lankan politics. Yet instead of being accommodated by democracy, mollified by the leftward tilt of policy, or content with lawful political mobilization, the JVP began preparing for revolt within weeks of the UF’s election. From its founding in the late 1960s, the JVP believed that “power could not be captured from the capitalist rulers by the democratic method of holding elections” (Alles, 1976, p. 51).

The JVP’s rhetoric focused heavily on the “betrayal” (p. 26) of the revolution by the parliamentary left, and its arming was intended “to capture power from whatever government was in control at the time [right UNP or leftist SLFP]” (Alles, 1976, p. 67). The group’s threats of revolt “commenced from August 1970” (Alles, 1976, p. 58)—soon after the left had swept to electoral power in the May 27 elections. Indeed, as Alles notes, these threats occurred “in spite of the fact that the United Front Government did not in any way curb the activities of the movement at that stage” (Alles, 1976, p. 64).

The 1971 JVP case suggests a different logic than H2’s focus on repression: rebellions can instead emerge during “incorporation windows” when left movements are faced with a hard choice between electoral co-optation, on the one hand, and revolutionary violence, on the other. These windows are particularly contentious when multiple organized actors exist on the left without a single dominant party or leadership. Fragmented movements are more prone to splintering along lines of ideology, personality, and organizational rivalry.

Incorporation windows can include the introduction of electoral democracy at decolonization, visible increases in leftist parties’ popularity in mainstream electoral politics, and overtures to left forces by political elites attempting to build new coalitions. These political moments can be momentary or last for years; in either case, they are new periods of potential advance within, rather than against, the electoral system.

In some cases, these breakthroughs are the outcome of intentional efforts to increase leftist leverage within mainstream politics. But often they arrive as unexpected shocks that reshuffle the array of possible options facing a movement: for instance, ruling elites may reach out to the left to make up for the breakdown of their own ruling coalition, international pressures may create incentives for governments to create more or less space for leftist political forces, and the success of one left faction may have implications for the prospects of others even if they were not a cause of that faction’s success. To avoid some of the problems of operationalization that can plague related arguments (like those on political opportunity structures), I specifically look for periods of (1) democratization, (2) leftist party growth in vote share and/or recent or new movements by left parties toward meaningful electoral engagement, and (3) international pressures for democratization and specifically openness toward the left. For instance, countries with tight alignment with the United States in contested parts of the developing world (such as Pakistan) during the Cold War seemed more likely to limit incorporation windows than in less-aligned countries (such as India).

Incorporation windows help us understand the timing of revolts, but on their own they are insufficient to explain variation. The outcomes of an incorporation window hinge further on movement structure: whether or not there is a dominant party that can control the behavior of leftist factions and groups. When a dominant party or leadership controls the left, it can direct political engagement and mobilization toward managed co-optation (Krause, 2017; Pearlman, 2008/2009; Staniland, 2014). If the movement is cohesive but uninterested in incorporation, these windows will have no impact one way or another: revolutionary “organizational weapons” (Selznick, 1952) will either try to use them as opportunities for subversion or simply ignore electoral politics while continuing anti-regime mobilization. In neither case should an incorporation window distinctly trigger conflict: either it will open space for smooth incorporation, or do nothing.

Things are more complex when a leftist movement is divided among multiple organizations and factions that possess substantive collective action capacity (i.e., there is not a dominant party plus a tiny and irrelevant fringe). Sometimes open rivalries and splinters emerge only once incorporation becomes possible, but I hypothesize that the organizational roots of factionalization should often be visible even prior to an incorporation window.

What explains this variation in movement fragmentation? There is no single cause of leftist unity or fragmentation, but we can identify a set of plausible mechanisms that can help explain variation across contexts. First, the historical legacies of past mobilization can create enduringly fragmented movements based on different social and class bases, linked to distinct factions, or divided among rivalrous leaders. Initial processes of party-building and contingency can have long-lasting effects (Shefter, 1994); this is especially important when these structures emerged well prior to the contemporary political system experiencing an incorporation window.

Second, international diffusion and pressure can induce both cohesion and fragmentation. When international backers push for unity, movements can consolidate (sometimes under serious pressure). By contrast, when a broader international movement is divided—for instance, in the aftermath of the Sino-Soviet split (Lüthi, 2008)—this can drive new divisions within domestic movements as different factions align with different international patrons and inspirations.

Finally, state repression in the past can have important implications for the structure of movements in the present (Della Porta, 1995, Finkel, 2017). Preemptive destruction of radical splinters can truncate the ideological range of leftist movements. Both through words and deeds, the state can lay down “red lines” for acceptable and unacceptable political positions that limit just how ideologically divided the left can be. State power can enforce a certain degree of ideological homogeneity, though there is no guarantee in a democracy that such efforts will work.

It is important to be transparent that incorporation windows and movement fragmentation are not always independent of one another: sometimes splintering results from incorporation, or regimes open space for incorporation when they see radical splinters emerge and the possibility for dividing and ruling thus emerging. We need to be attentive to this possibility in examining specific cases. However, these variables are often distinct from one another, and there is sufficient possibility of analytical distance to make the framework in Table 1 useful in helping us systematically think about variation.

Table 1 outlines different combinations of leftist movement structure and incorporation windows. When an incorporation window appears and the left is dominated by a cohesive and disciplined political party interested in joining mainstream politics, the movement may be able to embrace the parliamentary path without violent splits or major internal ruptures. A classic case of leftist incorporation, despite violence being a plausible option, was the Italian Communist Party’s (PCI) “Salerno turn” in the waning days of World War II and its acceptance of a mainstream electoral role. When a dominant party is uninterested in entering electoral politics, the emergence of an incorporation window will not have any discernible effect, one way or another: the party will continue anti-system mobilization.

When a cohesive left movement does not face a clear choice about incorporation, it will likely try to balance a degree of electoral participation alongside continuing non-electoral mobilization; I refer to this as “riding two horses.” In these cases, the movement is not forced to choose: it can maintain a foot in the electoral arena while also pursuing strategies outside of the electoral arena that may not be compatible with pure electoralism. This may be an unstable equilibrium over the long run if there is a tension between the two approaches, but a diverse repertoire of political activity can allow a movement to keep different options alive rather than making a decisive choice. Fragmented leftist movements without an electoral incorporation window will be divided and uncoordinated, but there will be no clear trigger for a radical revolutionary turn by any faction.

I argue that the greatest likelihood of revolt is found when an ideologically diverse and organizationally fragmented leftist movement faces an incorporation window. On the one hand, the possibility of gaining serious influence without the uncertainty and violence of revolution is deeply attractive to reformists, especially in a moment when previous constraints and limitation may be falling away. On the other hand, radicals often view electoral democracy as fundamentally unsuited to meaningful redistributions of power and property. The parliamentary path, in this view, only promises betrayal, moderation, and sterile electoralism. This debate has broader echoes within other strands of political mobilization, such as political Islam and ethno-separatism, and the literature on “spoilers” and rebel fragmentation in civil conflict.³

Differing ideological positions create dramatically different expectations about the likely fruits of parliamentary engagement. This ideological tension can fuel fratricidal spirals of fragmentation as individuals and factions turn against one another in the face of an incorporation window. Incorporation windows are key, though of course not the only, triggers for such conflict: they create a blend of fear and opportunity among the radical components of a divided movement. Radicals fear being frozen out in the new electoral dispensation, but also perceive an opportunity to drive a clearly ascendant general leftist tide further toward genuine revolution.

As in the JVP case, such radical breakaways are often unable to capture the traditional urban working classes because of these classes’ ties to the electoral arena through patronage and parties. Rather than representing the poorest of the poor, radicals are often students, intellectuals, and even, in some cases, have an elite background; Wickham-Crowley (1992) has noted the prominence of sons of elites in Latin American leftist insurgencies. Their exclusion from the incorporated bastions of the center-left forces them into marginal peripheries to try to build peasant armies (Scott, 1978/1979; Skocpol, 1982).

The existing work most similar to this theory is Ron’s (2001) important study of the Sendero Luminoso in 1980s Peru. He argues that “political liberalization pushed Sendero toward revolutionary war, not toward political acquiescence” (p. 582) because “those that refused to accept parliamentary democracy. . . would find themselves placed beyond the pale by the state and the newly legalized mainstream left” (p. 583). This paper builds on Ron’s valuable insights to offer a developed and predictive theory of the conditions under which these revolts occur, and by pursuing a systematic cross-case research design. Crucially, it goes further by arguing that leftist insurgencies under democracy are driven by systematic causes that cast doubt on accepted conventional wisdoms about revolution and regimes.

These intuitions lead to H3, which focuses more on politics within the left than on blocked leftist participation:

Leftist revolt will not occur through this pathway under several conditions summarized in Table 1. First, left movements that never face real incorporation possibilities will have little reason to splinter over this choice. This could be because a movement is so marginal or divided that it has no prospect of electoral success, or because the group is managing to balance parliamentary participation with extra-parliamentary mobilization (i.e., able to keep both options open rather than having to choose). Second, incorporation without violence is most likely when there is a dominant party interested in electoral co-optation, and less likely when it is divided or fragmented. The outcome of greatest interest here is the choice of rebellion: a systematic study of leftist tides across the full spectrum of Table 1 will have to await future research.

Revolts of any sort are rare, and particular types of insurgency are even rarer. There are serious limits to what we can infer. This is not a “silver bullet” study: instead, the goal is to establish the plausibility of the core propositions and to reinvigorate systematic research on revolutions and revolutionary civil war. Nevertheless, there are ways to improve confidence in claims that can guide future work. This paper adopts a multi-layered research design in an effort to maximize the credibility of findings, while transparently identifying inferential challenges and cases in which my core hypothesis is not supported. Given the difficulties of studying political violence, adopting an ecumenical and mechanism-oriented methodological approach seems like a reasonable move (Kalyvas, 2019).

The case sample is made up of countries in South and Southeast Asia since 1945. This is a context that has experienced large-scale political violence, but with dramatic variation in the nature and timing of revolts. The first layer of the research design examines all of the leftist insurgencies that emerged in post-World War II South and Southeast Asia. It intentionally selects onset cases to explore how many occurred in periods of democracy versus those that emerged under authoritarianism. These are all relatively poor and highly unequal societies in which the “median voter” should push for substantial redistribution under democracy. The Supplemental Appendix provides a detailed justification of the coding of the timing of rebellion onset and regime type, as well as far more extensive sourcing for the cases and greater discussion of relevant alternative explanations.

The next, more intensive, set of case comparisons examines the leftist insurgencies that did occur in democratic periods. It relies on “causal process observations” (Brady & Collier, 2010) within each case of leftist insurgency onset within a democracy. It compares H2 and H3 to see which best accounts for the timing and process of leftist insurgency, excluding the 1971 Sri Lanka revolt because of its use as a theory-building case for H3. This approach builds on the influential design of Haggard and Kaufman (2012, p. 4), who argue that “Selecting on the dependent variable is a central feature of this approach, which is designed to test a particular theory and thus rests on identification of the causal mechanism.” Examining each case to see which aligns most closely with the two specified mechanisms, or whether there is little support for either, then allows an assessment of the existence and relative frequency of contending mechanisms (Brady and Collier, 2010; George and Bennett, 2005).

Finally, the paper briefly examines 1947–54 Pakistan and 1949–65 Indonesia, in which insurgency did not emerge during periods of democracy. While the bulk of the empirical work focuses on the comparisons and causal process observations above, identifying non-onset cases lays the basis for more systematic future testing across contexts.

I identify 18 onsets of leftist insurgencies in post-1945 South Asia, shown in Table 2. This does not include groups that may have had Marxist leanings, like the Liberation Tigers of Tamil Eelam, but that were not formally communist organizations (cf. Kalyvas and Balcells, 2010; Wood and Thomas, 2017), as well as tiny groups that were never able to generate any real violence. I code leftist groups as those that explicitly articulate their goals in terms of Marxist-Leninist and/or Maoist theory, aiming to seize the state and engage in systematic transformation based on class categories. I identify whether the group launched its insurgency in a year of democracy or autocracy. The Appendix engages in detail with coding decisions and their implications.

Of these 18 onsets, I assess that 12 occurred in a democracy-spell. The evidence suggests that H1—the strong version of the claim that leftist insurgency is a straightforward product of authoritarian regimes—is not sustainable, especially after de-colonization. Whatever the precise balance of onset in democracy vs. autocracy, a substantial number of the leftist insurgencies occur under democracy. Table 2 also supports the claim in one strand of the literature that leftists only win against authoritarian regimes—by the end of the rebel victories in the dataset, they all faced authoritarian governments. Clearly, however, onset is not at all restricted to autocracies.

We can also consider how this record relates to the distribution of autocracy and democracy across the 15 countries I consider as part of southern Asia in the post-colonial period. For a rough but valuable sense of how common autocracy has been relative to democracy in this sample, I examine the data of Geddes et al. (2014), Svolik (2012), and Boix et al. (2013). They are consistent: all have roughly 65% to 70% of country-years as falling under some kind of autocracy and the reminder under democracy (see Appendix). Removing communist dictatorships, on the assumption that they are unlikely to be targeted by leftist insurgencies, does not change this pattern. Despite hugely important caveats, it is worth noting that the onset of leftist insurgency in democracies is higher than we would expect based on the simple distribution of country-years by regime type. Even more strikingly, in post-colonial independent countries, there were 15 leftist insurgencies (the MCP, CPB-Red Flag, and Viet Minh rebelled under colonial rule), of which as many as 12 (80%)—depending on coding decisions discussed in the Appendix—broke out under democracy, despite democracy accounting for only approximately 35% of non-communist-dictatorship years.

This section explores how H2 and H3 compare in explaining the onset under democracy cases. Both hypotheses are about processes and sequences. We can examine these sequences to see which mechanism—if any—matches the empirical record. To assess H2, I examine whether the insurgency was (1) preceded by a sustained program of repression against the left by the state that (2) “blocked” (a la Huntington, 1968) leftist participation and thus policy shifts toward redistribution. To assess H3, I measure (1) whether the revolt was preceded by/accompanied by intra-left splits over democratic incorporation, (2) the left-right orientation of the government against which the revolt was launched, and (3) whether there had been moments of democratization or leftist electoral breakthrough before or around the onset of the revolt.

This research designs selects on the dependent variable to explore the causal mechanism leading to onset, conditional on onset having occurred. There are clear limits to this approach (King et al., 1994), since there may be similar cases that did not turn to rebellion. All claims in the paper need to be appropriately caveated. However, as George and Bennett (2005), Brady and Collier (2010), and Mahoney (2010) argue, this approach can be valuable in weighing theories against one another by examining the process through which political phenomena unfolded. It allows us to compare how theories perform in explaining the timing, actors, and composition of revolts. Later in the paper, I briefly examine cases in which there was no leftist revolt during a period of democracy, to lay the basis for systematic future research.

Table 3 summarizes the 12 cases of leftist rebellion onset in periods of democracy and assesses whether each case is better explained by H2 or H3 (or neither). I find four cases that clearly support H2. I label these “State repression” cases. I identify seven cases that align reasonably closely with the incorporation windows argument of H3. However, one of the cases (1971 Sri Lanka) was a theory-building case, so I disregard it in this analysis. These emerged in periods in which the left was at least partially incorporated into mainstream politics, but radical splinters disagreed about the wisdom and viability of electoral contestation and chose revolt. This turn to violence occurred at moments of independence, democratization, and/or leftist electoral assertion in which the left’s eventual place in politics was wide open. Rather than a reaction to anti-democratic repression, these were ambitious revolts waged in large part against an ascendant, ruling center-left. The 1968–9 Philippines case provides partial support for both mechanisms.

This is a small set of cases and we can draw only very caveated inferences. However, it appears that there are two distinct pathways to leftist insurgency in democracies. One, found in the existing conventional wisdom, is driven by the (usually rightist) state repressing the left and thus spurring revolt driven by blocked political participation (H2). The other pathway is driven by dynamics within the left itself, with spiraling debates over democratic incorporation that turned violent (H3). Leftist insurgency is characterized by “equifinality” (George and Bennett, 2005)—multiple causal mechanisms can drive the onset of revolt. In the Appendix, I also explore other ways of categorizing these cases to see whether we can find other useful patterns to guide future research on the dynamics of redistributive revolt: for instance, I find suggestive evidence that Cold War alignment and transnational ideological diffusion affected incorporation windows and ideological heterogeneity.

There are four clear cases that support H2. In all, governments preemptively targeted left movements prior to the onset of the insurgency. In two cases, this repression emerged in large part due to Cold War pressures (1946 Philippines, 1959 Laos); in the other two (1987 Sri Lanka, 1974 Bangladesh), democratically elected leaders began repressing their rivals as part of a bid to consolidate and personalize power.

In the 1946 Philippines, leftist peasant guerrillas and peasant organizations who had fought the Japanese during World War II were aggressively targeted by a landlord-dominated government. A rebellion emerged in response, tying together localized peasant groupings with parts of the Communist Party of the Philippines (KPK) (Staniland, 2014). A genuine incorporation window did not exist, because of the class dynamics within the Philippines and emerging Cold War worries about the left. The evidence is unambiguous that a right-leaning government preemptively targeted the left and triggered the revolt (Kerkvliet, 1977, pp. 143–151; Slater, 2010, pp. 96–102).

Laos 1959, Bangladesh 1974, and Sri Lanka 1987 are somewhat trickier cases: some datasets code 1987 Sri Lanka and 1974 Bangladesh as autocracies, while the Pathet Lao insurgency in Laos was the continuation of the leftist component of an earlier revolt (the Lao Issara) rather than the initial turn to violence. However, these are precisely the ambiguous cases in which we would expect to see H2 operate, with democratic processes breaking down amidst growing repression.

In 1983 Sri Lanka, the center-right UNP government banned the JVP (which had returned to mainstream politics in the late 1970s) after the anti-Tamil riots of July 1983 (Gunaratna, 1990, pp. 187–188). Though the government claimed that the JVP had a key role in the rioting (Gunaratna, 1990, pp. 187–190), there was no evidence of this—“the government was unable to come up with any convincing proof that the JVP was in any way responsible for instigating the riots” (Chandraprema, 1991, p. 60), which instead were closely linked to thugs of the ruling party. The JVP was blocked, despite its genuine interest in electoral participation; Chandraprema argues that “Between 1977 and 1983 the JVP did everything possible to make its electoral projections a success” (Chandraprema, 1991, p. 51). The group was forced underground, and began preparing for the revolt it eventually launched in 1987.

Laos had experienced a broad-based, but fragmented, rebellion against the French in 1946 in the Lao Issara government. This movement largely disintegrated in 1949. However, the Communist component of the Lao Issara continued to fight, establishing the Pathet Lao in 1950. After the 1954 Geneva Accords, the Pathet Lao’s political party wing, the Lao Patriotic Front (LPF) agreed to join a national unity coalition government (Stuart-Fox, 1997, p. 96). Much of the Pathet Lao demobilized, but some elite units were moved into Vietnam as insurance (Stuart-Fox, 1997, pp. 99–100). A “remarkably free and fair” (Stuart-Fox, 1997, p. 102) election in May 1958 brought the LPF substantial electoral power, though not into the government. However, right-wing discontent, Thai encouragement, and, especially, American financial pressure led to a rightist tilt within the government that “stepped up its repression of the LPF” (Stuart-Fox, 1997, p. 107) over the course of 1958 and 1959. The Pathet Lao had not fully demobilized, and it decided to throw these forces back into the fight while expanding its insurgent mobilization in areas of strength (Stuart-Fox, 1997, pp. 103–109), leading to war onset in 1959.

Finally, Bangladesh was born in war and secession from Pakistan in 1971. The leader of the Bengali movement was the leftist nationalist Sheikh Mujibur Rahman, who won Bangladesh’s post-independence 1973 election. Two major groups eventually took to war against the Mujib government—the JSD and PBSP. The PBSP, discussed below, shows little support for H2—it had been ultra-radical since the late 1960s and viewed parliamentary democracy as a sham. But the JSD is a different story. It emerged as a splinter from the ruling Awami League in October 1972, arguing that the AL was insufficiently radical. Unlike the PBSP, however, it was open to trying electoral politics in 1973, and became an opposition party that was comparatively more moderate than other ultra-left forces (Jahan, 1974, pp. 127–128; Mohaiemen, 2006, p. 301; Moten, 1980, pp. 72–81).

Mujib had created a paramilitary force, the Jatiya Rakkhi Bahini (JRB), that was deployed, arbitrarily and brutally, against the Awami League’s rivals—including the JSD. As Mujib moved toward a greater autocratization of Bangladesh, the JSD decided to go underground for self-defense in 1974: “It was the rise of the Bahinis that pushed JSD and others to also start arming themselves” (Mohaiemen, 2006, p. 301), founding an armed wing (the Biplobi Gonobahini) and aiming to overthrow Mujib (Masoom, 2000, pp. 39–40; Maniruzzaman, 1980, p. 184; Riaz, 2016, p. 54). This is clearly a case of state repression blocking democratic participation by leftist forces, contributing to a turn to insurgency. The JSD revolt—unlike its near-simultaneous contemporary, the PBSP—works for H2 better than H3.

I now turn to the cases that support H3, which argues leftist insurgency is most likely when left movements face divisive choices about whether to embrace electoral co-optation. The four cases above that provide support to H3 and the first JVP revolt in 1971 (set aside as a theory-building case) leave us with seven cases to examine. I find that six of them provide robust support for H3, while one (the 1968–69 NPA revolt in the Philippines) provides only partial support. Ideology does important work in explaining why these specific rebels, but not their contemporaries in the broader left, thought that democracy could not deliver meaningful policy change, even in political systems in which the left held power or was advancing through the electoral system.

The final case provides support to both H2 and H3: the CPP/NPA insurgency emerged in the late 1960s Philippines in a context of both blocked redistribution under democracy and visceral intra-left rifts over parliamentary politics. The Communist Party of the Philippines (CPP) was formed as an avowedly revolutionary party in late 1968. Its founding by Jose Maria Sison was a repudiation of the existing PKP, which had failed during its involvement in the Huk insurgency in the 1940s/early 1950s before returning to being mostly non-violent afterwards.

Albertus and Menaldo summarize the complexities of politics during this period: “universal suffrage enabled mass participation in elections, and the constitution enabled those elections to produce strong executives. But on the other hand, the dominant social actors—the agro-commercial oligarchy and the business class, many with economic or political ties to the United States—monopolized the seats in both houses of the national legislature” (2017, p. 262). The result of this combination was a consistent failure to redistribute (Slater et al., 2014, pp. 366–367).

This looks to be a promising case for H2. Yet the rise of the CPP/NPA is not a story of rightist repression and leftist reaction. First, there was profound disagreement within the left about whether armed rebellion was necessary or likely to be effective. The PKP in the late 1960s “renewed organizing for ‘parliamentary struggle’” (Abinales & Amoroso, 2005, p. 199) and “was content to slowly build front organizations, among students, workers, and peasants, not for the purpose of forcing a confrontation with the government, but for use as a springboard in a parliamentary struggle” (Jones, 1989, p. 25). Indeed, the KPK presciently claimed that violent rebellion would only “invite intensification of state repression and infiltration” (quoted in Fuller, 2011, p. 112).

Arrayed against this was the Mao-inspired emerging CPP, a young and “overwhelmingly petty-bourgeois” (Fuller, 2011, pp. 68–69) set of students and left activists led by Sison. They were believers in “armed revolution being the only road” (Fuller, 2011, p. 76). The “battle for influence between the old and new parties” (Abinales & Amoroso, 2005, p. 200) escalated from intra-KPK feuding in 1967 to the formation of the CPP in December 1968/January 1969 and then NPA in March 1969. The CPP/NPA violently clashed with KPK cadres as the split emerged (Fuller, 2011, p. 105). This process was remarkably similar to the pathway identified in H3—the KPK was working toward mainstream politics, while the more radical CPP saw mainstream electoral politics as a dead end.

Second, while Ferdinand Marcos was not a leftist, and policy did not move to the left during this period, he was also not in the late 1960s engaged in any particular crackdown against the left (that would come later). Instead, his main rivals were other elites with whom he battled over taxation levels and political power (Slater et al., 2014, p. 367). This is not a situation like Laos in 1958–9 or, indeed, the Philippines in 1946, of intensive state repression of the left that triggered revolt. Thus the onset of the NPA/CPP’s revolt thus provides evidence for both mechanisms.

Here I briefly explore other trajectories of non-insurgency identified in Table 1. We face a problem in finding cases of fully non-violent co-optation, however, since almost all of the southern Asian countries faced some degree of violence around the incorporation of the left. A clear out-of-sample supporting case is the PCI in Italy, but southern Asia’s countries either did not have full leftist incorporation or it was accompanied or preceded by leftist insurgency. This shows just how difficult it was to seamlessly or non-violently bring the left into the political arena—incorporation was a double-edged sword that both staved off revolution and triggered violent revolts.

Instead, I identify two cases in which incorporation windows did not exist and no leftist insurgency emerged (before the destruction of the non-violent left closed off future possibilities). The Indonesian Communist Party (PKI) captures the “riding two horses” outcome in Table 1—a disciplined, cohesive party that did not face a hard choice about democratic participation prior to its annihilation in 1965–66. It was an organized and popular electoral force, as evidenced by its showing in 1955’s national parliamentary elections and then regional elections in Java in 1957 and 1958 (Robinson, 2018, pp. 37–38). The 1955 elections provided the PKI with “previously unheard of political influence and opened the possibility that the party might come to power through parliamentary means” (Robinson, 2018, p. 44).

Yet the brief period of open national electoral politics in Indonesia came to an end in 1957 when Sukarno declared martial law and with the 1959 return to the 1945 Constitution and cancellation of a planned election. In the 1957–1965 period of Guided Democracy, the PKI remained a legal political party, but focused on mass mobilization: “the shift away from electoral politics and toward the politics of mobilization did nothing to weaken the party’s popularity. Indeed, the PKI proved to be exceptionally adept in the work of mass mobilization” (Robinson, 2018, p. 44).

The party continued to pursue power through non-electoral means, becoming an increasingly close ally of Sukarno (Slater, 2010). At the same time, the PKI “was a legal party with no armed wing and no plans for revolution” (Robinson, 2018, p. 43). Thus there was no incorporation trade-off at hand—it could hold open the door to electoral involvement while also keeping a foot in non-electoral mass mobilization and elite maneuvering. The PKI was then cut down by the army at the height of the PKI’s power in 1965–66, and the hard left violently eliminated from Indonesian politics. No incorporation window emerged prior to the left being wiped out by the military.

The Pakistani left in the 1940s and early 1950s is an example of a divided and marginal political force that never had a opportunity to make major mainstream political advances: it is an instance of the lower-right cell in Table 1. The left in Pakistan was divided deeply along linguistic and geographical lines. The peasant mobilizations of East Bengal were very different than the Punjab heartland of the Pakistan Communist Party (PCP), while the PCP also faced criticism from other leftist intellectuals and political parties (Ali, 2011). There was little chance of a major leftist breakthrough in the face of both its own limitations and a profoundly hostile state apparatus: “The repression of Communist activities in Pakistan since 1947 has severely hampered the ability of party leaders to organize a large following” (Franda, 1970, p. 589).

Scattered peasant uprisings in East Pakistan, alleged Communist involvement in the Rawalpindi Conspiracy of 1951, and the increasingly pro-American stance of the Pakistani state elite combined to limit communist freedom of maneuver. When the PCP made some inroads in East Pakistan in 1954, it was rapidly banned by the state and repressed, in a preview of the military-bureaucratic nexus that would soon place Pakistan on the road to authoritarian rule (Jaffrelot, 2015).

There simply was not a real incorporation window, and the CPP did not lead a mobilized and cohesive left. Thereafter, even under formal periods of democracy, leftists, whether legal parties like the National Awami Party (NAP) or underground communist cells, were never able to carve out much space in a polity dominated by an anti-communist army and bureaucracy. Though some leftists became involved in the Baloch insurgency of the 1960s, the left was not a prime mover in this revolt. Instead, leftist forces were focused in student and labor groups that could generate street power but had a limited electoral role.

As in Indonesia, state policy ultimately made the idea of a leftist electoral breakthrough irrelevant to Pakistan. In both countries, the Cold War and powerful anti-communist militaries severely limited the space for leftist parties in mainstream politics. By contrast, the comparatively more insulated contexts of India, Sri Lanka, Burma, and Bangladesh are where we see the greatest freedom for the left in mainstream politics, and both the greatest involvement by leftist electoral forces and most robust leftist insurgencies. This international context helps us understand where and why incorporation windows emerged.

Influential research on revolution and regimes has framed revolutionary forces as reactive—the state and the elite set the stage that determines the incentives of the rest. This paper has suggested greater agency among prospective revolutionaries and their ideological rivals. There is a systematic path to revolt that finds its origins in disputes within the left about electoral politics. Rather than a reaction to rightist repression or the specter of authoritarianism, this was often a course consciously chosen by ideologically committed cadres with plausible alternative strategies. These other strategies, in turn, were often pursued by social democratic rivals in structurally-similar contexts. As a result, a surprising number of leftist rebels launched rebellion not against right-wing oligarchs, but rather center-left and socialist elected governments.

This finding does not invalidate H2’s alternative; there are indeed cases that align with the “undemocratic democracies” mechanism of leftist revolt, as in Laos, 1987 Sri Lanka, 1974 Bangladesh, and the Philippines in 1946. However, it appears to be less dominant than existing literature has allowed. Leftist insurgency is characterized by “equifinality” (George and Bennett, 2005)—there are multiple pathways to onset that cluster into a small number of clearly-identifiable mechanisms. Methodologically, the research design here allowed both a careful examination of cases and a set of comparisons that provided a valuable mix of breadth and depth. This blend of medium- and small-N work may be worth pursuing more broadly in political violence research—conflicts are sufficiently rare and heterogeneous to make deep investigation worthwhile, but common enough to make comparison essential.

These dynamics may not be unique to southern Asia. A systematic global study of this phenomenon will have to await future work, but a brief external validity check suggests that there may have been similar dynamics in Cold War Latin America. As in southern Asia, there were certainly leftist revolts under autocracy: Castro’s Movement of July 26 in Cuba, the FMLN in El Salvador, ELN in Bolivia, and several insurgencies in Guatemala occurred under dictatorship. Yet in Colombia (ELN, FARC, M19), Peru (Sendero Luminoso, MRTA, MIR), and Venezuela (Bandera Roja) we see plausible cases of electoral democracy being the site of armed leftist mobilization. There are also more ambiguous cases, as with leftist groups in Argentina and Uruguay in the 1960s and 1970s that existed, with varying levels and forms of violence, across different regime types.

As in southern Asia, Cold War dynamics were very important both in inspiring revolts and shaping regimes’ responses to them, but these international forces interacted with domestic contexts in which regime type and political systems played a key role in determining whether and how open revolt broke out (Policzer, 2009; Wickham-Crowley, 1992). There seem to be similar pathways to radical redistributive insurgency across regions, tentatively suggesting external validity to the arguments advanced in this paper.

Understanding the conditions under which we see radical and violent redistributive insurgency requires further empirical research. First, further cross-national comparisons are necessary to probe, deepen, and complicate the theory while moving beyond the limitations of this paper’s research design. A deeper examination of Latin America during the Cold War is a natural candidate for systematic further exploration. Western Europe also experienced both leftist incorporation and (admittedly small) leftist terrorist campaigns during the Cold War (Della Porta, 1995). These cases provide an opportunity to more systematically explore the domestic and international conditions affecting the ideological structure of intra-left competition and the existence of incorporation windows.

More provocatively, Kalyvas (2015) has argued that radical Islamist groups have come to occupy the role that leftist revolutionaries occupied during the Cold War. While these groups have largely emerged in non-democratic countries, there still might be relevant similarities that can extend the basic logic of the argument: they have demanded radical changes to the political status quo, articulated fundamentally different understandings of the polity, and sought to capture the commanding heights of state power. There are a wide variety of contexts in which movements face a difficult choice between the revolutionary road and the parliamentary path.

Second, there is a need for more fine-grained sub-national comparisons that can improve our understanding of where and when precisely the radical left successfully mobilized for war, and when and where these movements were unable to generate momentum (i.e., Chandra & García-Ponce, 2019). This can help us understand, for example, why West Bengal was so violent in the 1960s while Kerala was substantially less so, despite both having vibrant left movements, or why some areas of the Philippines proved more receptive than others to both the Huk and NPA insurgencies. This is an area in which the micro-level strand of civil war research (Kalyvas, 2006) can be productively fused with the more meso -level approach of this article to uncover how broad political trends actually played out on the ground.

There are two broader theoretical insights that emerge from this article. The first is that democracy can have complex, often cross-cutting, effects on conflict and mobilization (Jones & Lupu, 2018). In a number of the cases, we see a fascinating dynamic in which electoral democracy breaks apart a potential revolutionary coalition, but in doing so pushes the radical rump toward open revolt. This forestalls the overthrows of regime that scholars like Skocpol have focused on—but opens the door to grim, often decades-long, insurgent and counterinsurgent wars that carry a high human cost. The incorporation of the left in 20^th-century Asia was a bloody, protracted, and ambivalent process.

The second theoretical point is the role of political ideology in the debates that triggered insurgent revolt. Individuals and groups with similar structural characteristics often adopted radically opposed views of politics, in ways that cannot be easily mapped onto their position on an income distribution. There is nothing obvious about Siraj Sikdar’s class status in 1972 Bangladesh that would make him seem objectively more likely to pursue Maoism than Mujib-ism. Transnational ideological currents were grafted onto local political contexts in highly contingent ways, with enormous consequences. The study of conflict needs to more systematically theorize and measure the origins, diffusion, and evolution of ideology to better understand how these understandings of political life became fused to both revolutionary violence and parliamentary incorporation (Costalli & Ruggeri, 2015; Gutierrez- Sanín & Wood, 2014; Hoover Green, 2018; Leader Maynard, 2019; Staniland, 2015; Stewart, 2020; Straus, 2015). The leaders and foot soldiers of revolution, counter-revolution, and center-left incorporation all pursued their visions at high personal and political cost, often in the face of tremendously long odds; their ideas deserve to be taken seriously.

The author thanks Michael Albertus, Sarah Daly, Jon DePoyster, Evgeny Finkel, Stathis Kalyvas, Monika Nalepa, Dan Slater, Drew Stommes, Andres Uribe, Adam Ziegfeld, three anonymous reviewers, and the editors of Comparative Political Studies for excellent feedback. Will Howell organized a stimulating faculty workshop on the manuscript in the University of Chicago Department of Political Science. Noah Schouela, Winston Berg, Liya Khan, Bryan Popoola, Catherine Grothe, Dan Gamarnik, Madeleine Stevens, Erik Mueller, and Wenyan Deng provided exceptional research assistance.