Terrorism, counterterrorism and ‘the rule of law’: state repression and ‘shoot-to-kill’ in Northern Ireland

ABSTRACT Authors have argued that counterterrorism must be consistent with ‘the rule of law.’ Often associated with this approach is the assumption that plural political structures limit the state’s response to terrorism and that state agents will be held accountable if their response is excessive. Scholars who focus on social movements reject this assumption. We examine the state’s response to anti-state violence in Northern Ireland between 1969 and 1994. In 1982, Sinn Féin did much better than expected in an election to the Northern Ireland Assembly. Following the election, it is alleged that state agents followed a ‘shoot-to-kill’ policy and shot dead Irish republican paramilitaries instead of arresting them. We find evidence suggesting such a policy and consider the implications.

When responding to "terrorism", democratic states face the dilemma of defending themselves without "destroying the values for which they stand" (Ignatief 2004, p. vii;Merom 2003).
Indeed, several authors have argued that the state's response to terrorism must be consistent with "the rule of law": that counterterrorism should be rooted in a justice system with checks and balances that protect everyone, including anti-state activists who deny the legitimacy of legal structures designed to protect even their civil liberties (Chalk 1995;Pedahzur and Ranstorp 2001;Wilkinson 2006: 61-94; see also Kalyvas 2003).
In contrast, social movement theories reject the assumption that liberal democracies are governed by plural political systems that guarantee due process for everyone. As Davenport and Inman have commented, "when authorities are challenged with some form of conflict, they engage in some form of repressive action-simply, threatened governments normally respond with force" (2012: 622; see also, Davenport, 1995;Gamson 1990Gamson [1975 ;Gurr, 1988;Tilly 1978;Alimi, Demetriou, and Bosi 2015;Earl 2011;Lichbach 1987;Khawaja 1993;White and White 1995;Rasler 1996).
Scholars generally agree that democracies have a complex set of repertoires with which to respond to challenges. These include accommodative means that allow for participation and contestation (Gurr 1988, p. 54;Davenport 2007, p. 11). In democracies, leaders are under more scrutiny as an independent press can publicize illegitimate uses of force and they may be voted out of office. Democratic leaders also must worry about the cost of using inappropriate repressive measures that lead to public outrage-the "backfire" effect of repression (Hess and Martin, 2006). They must employ tactics that demobilize movements while simultaneously mitigating or eliminating backfire effects (Smithey and Kurtz 2018, p. 185). Smithey and Kurtz (2018) argue that elites have a continuum of repressive tactics available to them, ranging from violent sanctions that induce fear among challengers to persuasive inducements that increase the chances that challengers will internalize a regime's legitimacy (p. 190). The response to a domestic threat will depend on many factors, including the type of the challenge, the identity of the challengers, the challengers' goals, and the regime type. Lethal force is but one of the many tactics leaders may use to repress opposition. We investigate the possibility that a liberal democracy, when threatened, and like its more authoritarian counterparts, will systematically employ lethal force against its own citizens. The following is a case study of state violence against Irish republican paramilitaries-"terrorists"in Northern Ireland. 1

CONTEXT
Between 1969 and the 1998 Good Friday Agreement, more than 3,600 people were killed by political violence in Northern Ireland (White 2017). Broadly, there were three actors in the conflict: anti-state Irish republican paramilitaries, pro-state loyalist paramilitaries, and the security forces (such as, the British Army and the Northern Irish police, the Royal Ulster Constabulary (RUC). The "Provisional" Irish Republican Army (IRA) was responsible for almost 1,800 fatalities, including more than 900 members of the security forces and 500 civilians (of whom approximately 350 were Protestants). Other republican paramilitary organizations active during these years include the "Official" IRA, the Irish National Liberation Army (INLA), the Irish People's Liberation Organization (IPLO), and the "Continuity" IRA. Loyalist paramilitaries were responsible for almost 1,000 fatalities, and most of their victims were civilians (650 Catholic and 120 Protestant;McKittrick, et al. 2004).
In general, there was "strategic consistency" in the approach of successive British governments that acknowledged a political dimension to the violence, but also viewed it as criminal behavior (Cunningham 2001, p. 155). This approach was rooted in past practice (including the British response to previous insurgencies in Ireland and in colonial settings), interacted with the republican paramilitary campaign, and included security-based and politically-based initiatives (see Bell 1976;Barkan 1984;McAdam 1983;Neumann 2003;Smith 1995). For example, in 1971 internment without trial was introduced in Northern Ireland for the fourth time in the 20th century (Beggan 2006;Bell 1993;De Fazio 2018). These efforts were complemented by political initiatives that included negotiations with various Irish governments, the Sunningdale Agreement (1973), which led to the brief establishment of communal powersharing , and a Provisional IRA-British truce in 1975 (English 2003;Ó Dochartaigh 2015). Over time the security forces became more sophisticated and discerning (see Table 3).
Paramilitaries also adapted. In 1976-77, the Provisional IRA reorganized for a "Long War" strategy. They scaled back their bombing campaign, embraced a smaller and more secure cellular structure, and adopted a "total strategy" of coordinated military and political activity (Smith 1995: 110-12;152-161;English 2003;Moloney 2007: 175-8; see also Morgan and Smith 2016;White 2017).
In directly confronting terrorism, the security forces were more aggressive against antistate republican paramilitaries than they were pro-state loyalist paramilitaries, and also killed a large number of civilians. McKittrick et al. (2004Table 18) 2 Police officers and soldiers were subject to rules of engagement but in several instances were prosecuted for the excessive use of force, including four soldiers found guilty of murdering civilians.
Our focus is on an apparent change in the republican paramilitary-security force dynamic in the early 1980s. We address the development of an alleged shoot-to-kill policy, directed against anti-state republican paramilitaries, under which the rules of engagement were seemingly compromised.

SHOOT-TO-KILL
On 1 March 1976, "special category status" for paramilitary prisoners was ended.. The Irish republican protest against their "criminalization" peaked in 1981 when ten prisoners fasted to death in an attempt to force the British into conceding political status (Kirkpatrick 2018a;2018b;O'Hearn 2006). Although the prisoners failed to win immediate concessions, the support mobilized outside of the prison was much greater than expected. In a Northern Ireland byelection, hunger-striker Bobby Sands was elected the MP for Fermanagh/South Tyrone; his subsequent funeral drew an estimated 100,000 people. In June 1981, two IRA prisoners were elected to the Dublin parliament, contributing to a change in government. In the wake of the hunger strike, the republican movement as a whole experienced a large-scale mobilization, especially the Provisional IRA and its political wing, Sinn Féin.
To counter the hunger strike mobilization and support the Social Democratic and Labour Party (SDLP), the moderate Catholic/nationalist alternative to Sinn Féin, the British government sought to restore a government for Northern Ireland, beginning with an election to a new Northern Ireland Assembly in October 1982. In order to register a protest vote, and after pledging they would not take their seats in an illegitimate British assembly, Sinn Féin contested this election and received 64,191 first preference votes (10.1% of the total). Of five Sinn Féin elected representatives (out of 78 total), three of them-Gerry Adams, Martin McGuinness, and Danny Morrison-were assumed to be leading members of the Provisional IRA (Moloney 2007 one third of the nationalist community supported a war to expel them from Northern Ireland" (O'Leary and McGarry 1993, p. 213; see also Neumann, 2003, p. 116;Patterson 1997, pp.194-95;Smith 1995, p. 162).
This was the political context in which, on 11 November 1982, Seán Burns, Gervais McKerr, and Eugene Toman, three members of the Provisional IRA's "North Armagh Brigade", were shot dead by the Royal Ulster Constabulary's E4A Special Support Unit. Police officers claimed that the men killed had tried to drive through a roadblock. Forensic evidence showed that the car was stationary when they were shot. The fact that the paramilitaries were unarmed suggested to some that they had been executed. This was the first of a series of alleged shoot-tokill incidents involving RUC and British army special covert operations units (see also Bew 8 2014).
Our investigation is two-fold. First, we use quantitative data to test the hypothesis that following the 1982 Northern Ireland Assembly election there was a change in the nature of republican paramilitary-security force confrontations. 3 Second, we present a chronological analysis of alleged shoot-to-kill incidents between 1982 and 1992.

DATA AND METHODS
Our data are primarily drawn from

Time Frame
We examine the security force-republican paramilitary dynamic between the "Lower Falls Curfew" in Belfast, 3 July 1970, and the Provisional IRA's ceasefire of August 1994. The curfew marks the beginning of direct conflict between republican paramilitaries and the security forces (Bell 1993). The ceasefire marks the beginning of the end of the Provisionals' military campaign. Between July 1970 (our starting point) and November 1992, there were 106 incidents in which the security forces shot dead 140 republican paramilitaries and three civilians. The youngest paramilitary shot dead was a thirteen-year-old member of Na Fianna Éireann, a youth group aligned with the Provisional IRA. The oldest was 52 years of age. The mean age of paramilitaries shot dead by the security forces was 23. Of the 140 paramilitaries shot dead by the security forces, 108 were members of the Provisional IRA, ten were members of the Official IRA, ten were members of Na Fianna Éireann, nine were members of the INLA, two were members of Cumann na mBan (an independent women's organization), and one was a member of the Irish People's Liberation Organization (IPLO). The vast majority of the victims (136 of 140 paramilitaries) were men (see Bloom, Gill, and Horgan 2012;Reinisch 2019;Reinisch forthcoming). 5 More of these incidents (53) occurred in Belfast than in any other location (see Table 1).
Combining with Derry city (16), just less than two thirds of the incidents occurred in urban areas (69/106=65%). There was one international incident. On 6 March 1988, soldiers of the Special Air Service (SAS) shot dead three IRA members in Gibraltar. The rest of the incidents were scattered among the small towns of Armagh, Downpatrick, Dungannon, Newry, and Strabane, and the rural areas of Counties Antrim, Armagh, Fermanagh, and Tyrone. A multivariate analysis allows us to determine if there was a qualitative change in British counterterrorism following the 1982 Assembly election.

Dependent Variables
British authorities have consistently denied that there was a shoot-to-kill policy in Northern Ireland. Our measures are by necessity indirect and are in essence an attempt to assess the existence of a latent variable-that is, a shoot-to-kill policy implemented after the 1982 Assembly election (Bollen 2002). 6 We measure observable outcomes of an alleged shoot-to-kill policy.
If there was a shoot-to-kill policy, the evidence suggests it was implemented by the special operations branches of the security forces: the RUC's E4A Special Support Unit, the British Army's 14th Intelligence Unit, and the British Army's Special Air Service (SAS). 7 In Secret Victory: The Intelligence War that Beat the IRA, William Matchett (2016) presents a "chronological list of all covert operations … that resulted in one or more fatality from 1974 to 1992" (Matchett 2016, pp. 108, 198-205, 221-22). Our first dependent variable is based on thirty covert incidents in which republican paramilitaries were shot dead. This is a binary variable (1=covert operation; 0=all other incidents). 8 Our second dependent variable is also binary and measures the number of paramilitaries shot dead in each incident (0=one paramilitary was killed; 1=two or more paramilitaries were killed). 9 Under a shoot-to-kill policy, the emphasis would be on killing paramilitaries rather than arresting them. This should result in a significant increase in fatalities per incident. In 86 of 106 incidents, only one paramilitary was killed. The first incident with multiple fatalities occurred on 23 October 1971. Maura Meehan and Dorothy Maguire, sisters and members of Cumann na mBan, were driving through West Belfast and warning people that a raid was underway when

Independent Variables
Our key independent variable measures whether or not a given incident occurred before or after the 1982 Northern Ireland Assembly election (0=before the election; 1=after the election). We control for the level of violence that preceded each incident, the geographic location of each incident, and the possibility that informers influenced security force operations.
The number of people killed between incidents is an indicator of the level of paramilitary activity prior to a given incident. We suspect that the more people killed between incidents, and the more members of the security forces killed between incidents, the more likely the security forces would respond with force. We control for the total number of people killed between incidents and the number of security forces killed between incidents. Because these variables are correlated, they are entered into separate equations. 10 These variables should control for the possibility that periods of increased paramilitary activity, such as after the hunger strike, led to more aggression from the security forces.
A binary variable is used to indicate whether incidents occurred in the urban areas of Belfast and Derry city or the rural areas and small towns of Northern Ireland (1=urban; 0=rural/small town). This variable should capture differences in the republican paramilitarysecurity force dynamic associated with urban versus rural guerrilla warfare/terrorism (e.g., Schultz 1978). Finally, we control for 78 alleged informers killed by republican paramilitaries between 1972 and 1992. 11 The loss of information provided by an alleged informer may have influenced the ability of the security forces to engage in counterterrorism. Executing an informer may also have made it less likely that other informers would provide additional information to the security forces. We anticipate that the more informers killed, the less likely there would be an incident. Table 2 presents the results of binomial logistic regression analyses on two dependent variables. In Models 1 and 2, the dependent variable is whether or not paramilitaries were shot dead in a covert operation (1=covert operation; 0=other incidents). In Models 3 and 4, the dependent variable is whether or not multiple paramilitaries were shot dead in an incident (1=more than one fatality; 0=one fatality). Separate models include the total number of persons killed between incidents (Models 1 and 3) and the total number of members of the security forces killed between incidents (Models 2 and 4).

[TABLE 2 ABOUT HERE]
The results across all four models show a significant effect for the 1982 Northern Ireland Assembly election. The likelihood of a covert operation perpetrated by the security forces in which republican paramilitaries were shot dead increased significantly after the election (Models 1 and 2, p < .001). Similarly, the likelihood that multiple paramilitaries were shot dead in a single incident increased significantly after the election (Models 3 and 4, p < .01). Based on Models 1 and 2, after the election it also became significantly less likely that there would be a fatal covert operation in Belfast or Derry city (p < .01). In Models 3 and 4, the urban or rural nature of the incidents was not significant, although the direction of the coefficients is negative.
Controlling for the other variables, the total number of persons killed and the number of security forces killed are insignificant across all four equations. These findings suggest that changes in the level of violence did not directly influence incidents in which republican paramilitaries were shot dead by the security forces. Similarly, the number of alleged informers killed did not directly influence incidents in which republican paramilitaries were shot dead.
Because coefficients are expressed in terms of the logged odds of the dependent variable and are non-linear in logistic regression models, their substantive interpretations are not straightforward (Long and Freese 2014, pp. 133-84). To better understand the influence of the Assembly election and the influence of geography on what appear to be shoot-to-kill operations, we generated their predicted marginal probabilities based on the findings for Model 1. Based on the predicted marginal probabilities, prior to the election there was a 26% chance that a covert operation with fatalities would occur in a rural area or small town but after the election there was an 84% chance that a covert operation with fatalities would occur in a rural area or small town.
It is unclear why the alleged shoot-to-kill incidents were more likely to occur in rural areas and small towns. Control of political violence may be easier in the limited space of urban areas. And by 1982 the security forces may have effectively contained republican violence in Derry city and Belfast and intentionally chose to focus on the countryside, especially along the border with the Republic of Ireland which offered a potential safe-haven for paramilitaries.
However, as shown in Map 1, relatively few of the covert incidents that occurred between 1982 and 1992 were along the border with the Republic of Ireland, and there were incidents in Belfast and Derry city.

[MAP 1 ABOUT HERE] [TABLE 3 ABOUT HERE]
A crosstabulation of persons killed by republican paramilitaries by geographic location suggests that the violence in Belfast and Derry city was never contained. As shown in Table 3, lethal republican activity in these cities increased relative to 1980 (as a percentage of total persons killed by Irish republicans) and then decreased in Belfast between 1983 and1986. 12 The decrease in Belfast, however, was caused by internal decisions made by the Belfast IRA and not the successes of the security forces. As part of Sinn Féin's mobilization and the decision to contest elections, it is well-documented that the Belfast leadership diverted funds to Sinn Féin and limited operations that might alienate voters (see Moloney 2007, pp. 242-45, 287-97, 328;English 2003;see also, Neumann 2006, pp. 959-64). This led to a schism in Belfast that was a precursor to a split in the Provisional IRA and Sinn Féin in the autumn of 1986. By significantly increasing Provisional IRA activity in Belfast in 1987, the leadership demonstrated their commitment to pursuing a military campaign while also embracing constitutional politics via Sinn Féin.
There may also be a geographic explanation of this finding. Implementing a shoot-tokill policy in a rural area may be more attractive because there are likely to be fewer witnesses and it is easier to conceal a team of soldiers or police officers in the countryside. Alternatively, there may be a higher likelihood of civilian casualties in an urban area. Different responses to informers may have influenced the geography of incidents. Of 29 alleged informers executed by republican paramilitaries between 1982 and 1992, seven were from Belfast and five were from Derry city, but only two were from Tyrone. Between 1982 and 1992, 23 members of the Provisional IRA's "Tyrone Brigade" were shot dead in 11 incidents.
The key finding from our quantitative analyses is that after the 1982 Assembly election there was a significant increase in covert incidents in which Irish republican paramilitaries were shot dead and there was a significant increase in incidents with multiple fatalities. These findings are consistent with allegations that the authorities adopted a shoot-to-kill policy following the 1982 Assembly election. Because this is a controversial interpretation of the findings, we considered alternative explanations. There is the possibility that in the early years of the conflict paramilitaries were more likely to operate as individuals and in small groups but in the later years they were more likely to operate in battle teams or cells. This might account for an increase in multiple fatality incidents. However, data presented in Table 4 rule out this explanation.
The IRA reorganization was in place by December 1977 and after this the size of active service units should have been relatively constant. 13 Table 4 presents information on incidents in which republican paramilitaries were shot dead in the post-reorganization era. Consistent with the findings, the monthly rate of total incidents, of covert incidents, and of multiple fatality incidents is greater after the Assembly election. The results in Table 2 are not a function of republican paramilitary organizational changes in 1976-77. 14

[TABLE 4 ABOUT HERE]
A quasi-experimental/"difference-in-difference" design (Angrist and Pischke 2008;Cook and Campbell 1979) allowed us to assess the possibility that the increase in republican paramilitaries killed was caused by a more general and aggressive approach by the security forces in the 1980s. Drawing on data from Malcolm Sutton's Bear in Mind These Dead (1994) and the CAIN Web Service, Table 5  Whether or not there was a shoot-to-kill policy, after the Assembly election the security forces became more aggressive when confronting republican paramilitaries.

[TABLE 5 ABOUT HERE]
Consistent with the view that a shoot-to-kill policy was implemented, our findings show that there was a qualitative change in the nature of counterterrorism in Northern Ireland after the 1982 Assembly election. Independent of the findings in Table 2, the findings in Tables 3, 4, and 5 confirm this change. A detailed examination of covert incidents between 1982 and 1992 also suggests that a shoot-to-kill policy was implemented.

Models 1, 2, 3, and 4 underestimate the complexity of British counterterrorism in Northern
Ireland. Alleged shoot-to-kill operations were undertaken by three different organizations-the RUC's E4A Special Support Unit, and the British Army's Special Air Service and 14th Intelligence Unit-in the same environment in which there was routine policing by the RUC while other segments of the British Army and the Ulster Defence Regiment (UDR) concurrently provided support and their own counterterrorism efforts. Paramilitaries could be shot dead in a variety of circumstances. As an example, on 6 February 1992, an IRA unit ambushed an offduty and part-time member of the UDR. The soldier, who was seriously wounded, shot dead one of the paramilitaries (McKittrick et al. , p. 1280. This incident is similar to several of the pre-election incidents and unlike many of the alleged shoot-to-kill incidents.

[TABLE 6 ABOUT HERE]
In order to better understand the republican paramilitary-security force dynamic in the post-Assembly election era, we begin with Table 6 which presents information on a subset of 21 covert incidents in which the RUC's E4A Unit, and the British Army's 14th Intelligence Unit and Special Air Service (SAS) shot dead republican paramilitaries (Matchett 2016, pp. 222-30).
Between July 1970 and the election in October 1982, there were 77 incidents in which 85 republican paramilitaries were shot dead by the security forces (85/77=1.10 fatality per event).
In the post-election period there were 29 incidents in which 55 paramilitaries were shot dead, which was a statistically significant increase in fatalities per incident (85/77=1.10 vs. 55/29=1.90; p<.001). As shown in Figure 1, a subset of 21 of these incidents producing 48 fatalities was even more deadly, with an average of more than two fatalities per event (48 fatalities /21 incidents=2.28). The primary victims were members of the Provisional IRA, who lost 40 members in 16 incidents. The primary perpetrators were members of the British army's Special Air Service (SAS) who killed 36 paramilitaries in 14 incidents. November 1982-autumn 1983, during which the first three incidents raised public concern that there was a shoot-to-kill policy, Sinn Féin benefitted from the public outcry, andif there was a shoot-to-kill policyit thus appears that a decision was made to switch implementation from the RUC to the British Army; Period 2) autumn 1983-spring 1985, during which the British and Irish governments began negotiating the Anglo-Irish Agreement and (consistent with the presence of our latent variable) incidents that could be perceived as shoot-to-kill operations temporarily ended; Period 3) spring 1986-autumn 1988, when the British government put forward a series of initiatives designed to curtail Sinn Féin and again what could be perceived as shoot-to-kill operations were temporarily ended; and, Period 4) spring 1990-November 1992, during which the British and Irish governments were both involved in secret negotiations with republicans that led to a Provisional IRA ceasefire.

Period 1: November 1982-Autumn 1983
The first three incidents here occurred in quick succession-two in November 1982 (11th and and 24th of the month) and the third on 12th December 1982. All three incidents occurred in North Armagh, involved the same RUC E4A team, and for each there was evidence of preplanning. In the second incident, for example, the E4A team staked out an arms dump in a hay shed. After two curious teenagers entered the shed they were shot by the E4A team; one teenager was killed, the other seriously wounded (McKittrick, et al. 2004, pp. 920-21, 926, 929-930;Murray 1990;Rolston 2000). The fourth incident, on 2 February 1983, was very different.
Two members of the INLA who were moving a weapon from one location to another in Derry confronted a person following them. It was an undercover member of the British Army's Special 14th Intelligence Unit who shot dead one of the INLA operatives and wounded the other (McKittrick et al. 2004, p. 936).
By the time of the fourth incident, the first three incidents had raised widespread concern that RUC officers were sidestepping due process and executing suspected paramilitaries. Such allegations caused problems for the authorities, and played into Sinn Féin's argument that the criminal justice system was compromised in Northern Ireland. In June 1983, Gerry Adams, was elected MP for West Belfast, andSinn Féin's overall vote increased to 43% of the nationalist electorate.
The state's response to the continued rise of Sinn Féin and the public outcry over alleged shoot-to-kill incidents was multi-faceted, and seemingly consistent with the "rule of law." In liberal democracies police officers, coroners, and prosecutors are public officials who to some degree are held accountable by journalists, elected representatives, and the public. In September 1983, the Director of Public Prosecution opened an inquiry into the first three incidents and police officers involved were charged with murder. In court, testimony outed an informer and revealed that RUC officers had lied to cover their activities. In a controversial decision, the officers were acquitted in April 1984, but the revelations led to an embarrassing investigation of the RUC directed by John Stalker of the Greater Manchester Police, beginning in May 1984 (Stalker 1988). If there was a shoot-to-kill policy, it appears that at some point in 1983 a strategic decision was made to shift operations from the RUC to the British Army, especially the Special Air Service (SAS) (see also Bew 2014). Following the third incident in this period, RUC E4A officers did not shoot dead another republican paramilitary until 10 April 1991.

Period 2: Autumn 1983-Spring 1985
The role of a police force, even a militarized force like the RUC, is very different from the role of an army that is trained to use deadly force to defend a state and prosecute war. The fifth alleged shoot-to-kill incident (4 December 1983) was very similar to the first three but with one important change-two IRA members were shot dead by the SAS as they approached an arms dump hidden in a hedge. In contrast to the aforementioned trial of RUC officers, the initial inquest into this incident was adjourned because three of the SAS soldiers involved did not appear. When the inquest reconvened, the only soldier from the SAS unit directly involved in the shooting who did appear in court had not actually witnessed the shooting (see Murray 1990, pp. 292-95;Tíghrá 2002, pp. 256-57). Alleged shoot-to-kill operations, as led by the SAS, continued until the tenth such incident, on 23 February 1985. After this, they stopped for more than a year. Political moderates were threatened by the growth of Sinn Féin and shoot-to-kill allegations,which undermined nationalist faith in the political system in Northern Ireland. The Irish government sought a role in the affairs of Northern Ireland, which they believed would help the SDLP and also assuage nationalist concerns. The British government, led by Margaret Thatcher, initially resisted but, faced with ongoing violence coupled with the rise of Sinn Féin, agreed to negotiate with the Irish government. The negotiations culminated with the signig of the Anglo-Irish Agreement (AIA) in November 1985, which gave Dublin a formal political role in Northern Ireland, including a voice on security policy. As shown in Figure 1, from November 1982 to February 1985, there was roughly one alleged shoot-to-kill incident every 11-12 weeks (119/10=11.9). After this, for the 38 weeks prior to the signing of the AIA, and for 23 weeks thereafter, there were no alleged shoot-to-kill incidents. If there was a shoot-to-kill policy, it seems to have been suspended in order to support the AIA process.

Period 3: Spring 1986-Autumn 1988
Covert operations in which Irish republican paramilitaries were killed re-started in the spring of 1986, with the eleventh such incident on 26 April 1986, and continued until 30 August 1988, which witnessed the fourteenth such incident. Accounts show that the security forces had advance warning of the republican operations leading to these incidents (e.g., Murray 1990, pp. 365-68;376-84, 396-409;439-48).
The twelfth incident in this period is particularly important because it clearly demonstrated a commitment to kill rather than arrest paramilitaries. On 25 April 1987, an IRA bomb killed Sir Maurice Gibson, the Lord Justice of Appeal in Northern Ireland, and his wife. In their statement claiming responsibility, the IRA specifically mentioned that it was Gibson who had acquitted RUC officers accused of murdering their comrades in what was seen as the first shoot-to-kill incident in November 1982. 16 Two weeks after Gibson's death, on 8 May 1987, the Provisional IRA's East Tyrone Brigade set out to bomb what they believed was an unstaffed, part-time police station at Loughgall, North County Armagh. The security forces acquired advance warning, but instead of taking steps to arrest those involved they set up an ambush.
When the IRA team began their attack they were met with five minutes of continuous fire in which the SAS shot an estimated 1,200 rounds. Eight paramilitaries were killed. Two civilians were also wounded, one fatally. Each of the IRA casualties had a head wound, including two paramilitaries who were shot fleeing from the scene. The SAS also fired without warning on the vehicle carrying the civilian casualties (Magee 2011, p. 3490;McKittrick et al. 2004McKittrick et al. , pp. 1077McKittrick et al. -1080Murray 1990: 380-83;Urban 1996: 227-37). The authorities expressed regret over the civilians targeted, but not the paramilitaries. The Secretary of State for Northern Ireland commented, "Those who launch such attacks have to face the consequences" (Rolston 2000, p. 130).
One interpretation of the next (third) incident in this time period (and thirteenth overall) is that it demonstrates official endorsement of shooting suspected terrorists dead instead of arresting them. Based on information that the IRA was planning a bombing attack in Gibraltar, Prime Minister Thatcher approved deploying an SAS team. On 6 March 1988, three suspected IRA members known to the authorities were spotted crossing the Spanish-Gibraltar border.
About an hour later, the suspects were heading back toward the border when they were confronted and killed by the SAS. The British government released a statement that there had been a "fierce gun battle", but the SAS team was flown 1,500 miles back to England before they could be interviewed by the Gibraltar police. The government subsequently acknowledged that the three "terrorists" were unarmed, but the soldiers involved claimed that the deceased had made movements that were threatening. Witnesses contradicted this and stated that two of three deceased had their hands in the air when they were shot (McKittrick et al. , pp. 1112Murray 1990;Rolston 2000, pp. 155-74). The Gibraltar event remains one of the most controversial of the alleged shoot-to-kill incidents.
The final such incident during this period (and fourteenth overall) is important because it seemingly demonstrates a complex relationship between alleged shoot-to-kill incidents and other repressive measures. On 20 August 1988, an IRA bomb killed eight British soldiers near Ballygawley, in County Tyrone. The state's response was both immediate and long-term. Ten days after Ballygawley, in "a carefully planned ambush", the SAS shot dead three members of the Tyrone IRA who they suspected had been involved in the bombing (McKittrick et al. , pp. 1143Patterson 1997, p. 211). And over the course of the next year, the British government engaged in "smart repression" against Sinn Féin (see also, Smithey and Kurtz 2018a, pp. 185-86). Their goal was to undercut the growth of Sinn Féin and force the party into a more moderate and constitutional direction while simultaneously limiting "backfire" from people worried that the government was taking steps to deny fundamental civil liberties (see also O'Leary and Silke 2011).
In the autumn of 1988, the Thatcher government restricted the press by introducing a broadcasting ban on statements from persons who supported proscribed organizations, restricted a suspect's right to silence in Northern Ireland, (allowing judges in non-jury courts to "draw inferences" of incrimination if alleged paramilitaries remained silent when questioned), and restricted free speech by making it a criminal offense for candidates and elected officials to express support for proscribed organizations. Although there was generally broad support in parliament for the government's approach to the conflict in Northern Ireland, this package of measures was controversial and met with opposition from the Labour Party and civil libertarians (Cunningham 2001, pp. 60-61;Thatcher 1993, pp. 411-15). In combination with other features of the state counterterrorism campaign, they were also effective. Sinn Féin's vote fell in the 1989 local elections, and during the 85 weeks (approximately 20 months) folllowing the fourteenth such incident, there were no alleged shoot-to-kill incidents in Northern Ireland. Period 4: Spring 1990-November 1992 After the above mentioned legal restrictions were in place, more incidents occurred that revived allegations of a shoot-to-kill policy. Between 18 April 1990 (the fifteenth such incident) and 25 November 1992 (the final incident) there were six such incidents over 31 months, averaging approximately one every five or six months (31/6=5.1; one incident every 22 weeks).
Political dynamics are also evident in the ending of such incidents. By the time of the final incident, the British and Irish governments were both involved in secret negotiations with the IRA. In 1992, republican paramilitaries killed 42 people, and there were two alleged shootto-kill incidents by the security forces, who killed nine people overall. In contrast, in 1993 republican paramilitaries killed 39 people, but there were no alleged shoot-to-kill incidents and indeed, for the first time since 1968, the security forces killed no one. Between 1 January 1994 and the start of the Provisional IRA ceasefire on 31 August, republicans killed 27 people, but only one fatality was attributed to the security forces, of a man who died from injuries sustained more than a decade earlier (McKittrick et al. , pp. 1345(McKittrick et al. -46, 1475. What did not happen in 1993 and 1994 is consistent with the notion that a shoot-to-kill policy was suspended in order to support an incipient "peace process" (see also, Gupta 2007). 17

DISCUSSION
In Northern Ireland, a key aspect of the government's counterterrorism strategy was the argument that republican paramilitaries were criminals. Sinn Féin's success in the 1982 Northern Ireland Assembly election undermined that argument and showed the potential of a "dual military/electoral strategy" for republicans (Smith 1995, pp. 219, 161-94;Patterson 1997, p. 193-95). Our contention is that British elites saw this as a serious threat that had to be countered.
Quantitative analyses show that after the election there was a statistically significant increase in covert security force incidents in which republican paramilitaries were shot dead, and a significant increase in incidents in which multiple republican paramilitaries were shot dead.
These results are consistent with the notion that a shoot-to-kill policy was implemented following the 1982 Assembly election. Indeed, whether or not there was a shoot-to-kill policy, it is clear that the nature of the republican paramilitary-security force dynamic changed significantly in 1982.
Mark Urban writes that the "key role in advocating ambushes [was] played by middleranking police and army officers" and that the attitude of top-level police and military officers was "one of acceptance than of initiating a wave of ambushes" (1992, 241;see also Neumann 2003, pp. 107, 131, 144). Our view is that top-level police and military officers, and British political elites, did more than accept what was happening.
In the context of the Assembly election, RUC E4A officers on their own initiative probably started the tit-for-tat targeting of suspected paramilitaries (see also Balian and Bearman 2018). In the first incident (October 1982), the victims were prime suspects for an attack that killed three RUC members two weeks earlier. The third incident (12 December 1982), in which two INLA volunteers were killed, occurred a week after an INLA bomb killed eleven soldiers and six civilians. But what accounts for the end of such incidents involving the RUC and the transition to similar incidents involving the British army amidst a storm of public concern and the continued growth of Sinn Féin and the Provisional IRA?
It is possible that, on their own initiative, middle level SAS officers followed the RUC's lead and adopted a shoot-to-kill policy. And three of the relevant SAS incidents also have titfor-tat aspects to them. In September 1983, Seamus Campbell was one of 19 IRA members to escape from Long Kesh high security prison; the escape was a major political embarrassment.  2011, pp. 383-401;Van Der Bijl 2017, p. 179;Murray 1990, pp. 377-79, 439-40). Without question, however, sending the SAS to Gibraltar and withdrawing them immediately after the incident there was decision made at the highest level (Murray, 1990, p. 410-13). Perhaps most telling, for extended periods at highly politically sensitive times (Feb 1985-April 1986August 1988-April 1990November 1992-August 1994 there were no such incidents, even though republican paramilitaries remained active throughout. Either conditions were such that the security forces did not have any opportunities to directly confront the paramilitaries during these time periods or, more likely, there was some political influence or consideration in the stopping and starting of such operations.
In summary, we believe that that there was a shoot-to-kill policy and that it unfolded in the following way. Sinn Féin's dramatic success in the 1982 Northern Ireland Assembly election threatened the status quo to such an extent that RUC officers on the ground and seeking revenge for republican attacks initiated a shoot-to-kill strategy of their own accord. Either soldiers on the ground or higher up the chain of command then transferred this strategy to the SAS and the British Army's 14th Intelligence Unit. Once it was in place, we believe that British political and military elites then pursued and controlled the policy for their own ends. Ultimately, we cannot prove or disprove the hypothesis that the security forces pursued a shoot-to-kill policy in Northern Ireland. However, our findings are consistent with the argument that such a policy was adopted following the 1982 Northern Ireland Assembly election. 18

IMPLICATIONS FOR SOCIAL MOVEMENT THEORY
Our analyses suggest that liberal states, when necessary, will systematically side-step the rule of law when confronting anti-state protest (Morgan and Smith 2016;Bennett 2010). This finding has implications for social movement theory. Scholars of social movements assume that contention lies on a continuum. At one extreme are non-violent political activities like voting and leafletting and at the other extreme is political violence. These may be different behaviors, but they are all political behaviors (Gamson 1990(Gamson [1975Tilly 1978;Tilly and Tarrow 2007, pp. 9-11, 69-87, 136-61;Demirel-Pegg 2014). Based on this case study, scholars should assume something similar with respect to state repression. At one extreme is "soft repression", including measures designed to intimidate those involved in protest, such as the police asking for the names and addresses of people attending a rally. At the other extreme is "hard" and deadly repression that ignores due process and civil liberties (Earl 2003;Kurtz and Smithey 2018b).

27
Our findings are qualified because they are restricted to the repression of anti-state insurgents. In Northern Ireland, the state's response to anti-state violence was different from its response to pro-state violence (White 1999). There are, for example, no allegations of a systemic shoot-to-kill policy directed at pro-state loyalists, even though they were responsible for almost 1,000 fatalities. The literature on repression would benefit from additional examination of the pro-state paramilitary-security force dynamic.     9 Unfortunately, we cannot construct a ratio variable of the number killed by those arrested and/or a ratio of the number killed by the total number of paramilitaries involved in any given incident. Such variables are impossible to construct given the clandestine nature of the incidents and the fact that in some events paramilitaries involved left the scene undetected while others were in the relative background, e.g., in "scout" cars. 10 In counting deaths between incidents, we exclude those killed in the specific incident.
Between 10 February 1975 and 22 September 1975, there was a bilateral British-IRA Truce. We include the final incident prior to the truce  but exclude an incident on 5 June 1975 in which an IRA member was shot after a sectarian attack (#1389, pp. 545-46). For the independent variables, we do not count fatalities during the truce and re-start from 22 September 1975. 11 The incident on Gibraltar is coded as an urban incident. Alleged informers are identified in