This thesis was inspired by the question of how the state addresses irregular challenges for its survival and reputation. It used an example of the confrontation between illegal armed groups (IAGs) operating in the North Caucasus and the Russian state in 2007-2014. Investigation started by asking to what extent do repressive and reconciliatory counter tactics decrease the level of violence produced by illegal armed groups? The thesis was situated in-between of deterrence and backlash theories to examine (in)effectiveness of repressive and reconciliatory policies. It accounted for (in)effectiveness by investigating whether the policy decreases or increases the level of insurgency-related violence; namely, whether it causes deterrence or backlash, in the case of repression or, alternatively, whether it causes conformity or backlash in the case of reconciliation. The thesis operationalised its main variables by disaggregating the strategy into separate repressive and reconciliatory tactics. It considered, on the one hand, three types of IAGs tactics: armed assault, bombings (suicide bombing, vehicle bomb, bomb placement, bomb tossing, firing, fake bomb) and hostage taking. One the other hand, the state tactics were divided into four categories: repressive indiscriminate (regime of counterterrorist operation, clash, and shelling), repressive discriminate (special operation, shooting, arrest, seizure, and detection), reconciliatory indiscriminate (involvement of civil society through dialogue, and socio-economic development), and reconciliatory discriminate (amnesty and reintegration). The thesis expected targeted repressive operations to suppress active IAGs members, whereas socio-economic incentives to contribute to maintaining the success of violent repressive operations. To test these hypotheses, the thesis relied on large empirical data, specially collected from the open sources, including 3270 episodes of IAG-initiated violence and 6114 governmental repressive actions. Data for reconciliatory efforts was taken from official statistics. The thesis used a generalized linear negative binomial and a generalized additive negative binomial model to assess the relationship between governmental policies and the level of violence. The thesis found that discriminate violence does indeed decrease attacks. However, it causes an immediate strong backlash effect at first, and only with considerable time and magnitude of repression eventually leads to the reduction of violence. The more discriminate repression is applied the less backlash it causes. Unlike repression, reconciliatory tactics produce a decrease in attacks. Thus, the thesis found partial support for both deterrence and backlash models. It, however, showed that deterrence effect overcomes initial backlash reaction.