Accounting for diverging paths in most similar cases: corruption in Baltics and Caucasus

This paper discusses the conditions under which post-Soviet states succeed in fighting corruption. The method of paired comparison of most similar cases, Estonia and Latvia on the one hand and Georgia and Armenia on the other, is used to tease out the variables that vary within and across pairs and produce divergence. It is argued that young, and structurally and ideologically cohesive, groups in power that are antagonistically predisposed toward the former colonial patron and free from the influence of the old guard are more likely to reform while enduring political-economic networks undermine anti-corruption reform.


Introduction
In the literature on political transitions it has been argued that some posttransition countries may become trapped in a 'gray zone' when 'transitions get stuck' [1,2]. Similarly, empirical evidence suggests that countries often get stuck in vicious cycles of malfeasance and corruption 1 despite economic Crime Law Soc Change (2017) 67:187-208 DOI 10.1007/s10611-016-9658-y 1 For the purposes of this paper I adopt Philp's (2011) definition of corruption: Corruption in politics occurs where a public official (A), acting in ways that violate the rules and norms of office, and that involve personal, partisan or sectional gain, harms the interests of the public (B) (or some sub-section there-of) who is the designated beneficiary of that office, to benefit themselves and/or a third party (C) who rewards or otherwise incentivises A to gain access to goods or services they would not otherwise obtain. This definition is most comprehensive because it accommodates often-overlooked issues such as influence trading, systems of patronage, the use of legal means to deliver favours and illegal funding to groups rather than individuals that can indirectly lead to private gain (Yadav 2011). Also, it does not exclude the acts that are not restricted by the legislation such as state capture.
transformations or political developments. Strikingly, besides a few notable exceptions cases of highly corrupt countries becoming less corrupt are rarely observed [3]. Two of these contemporary anti-corruption achievers, Estonia and Georgia, are located in post-Soviet Eurasia -both of which have demonstrated remarkable transition from highly corrupt regimes to less corrupt environments. This paper looks at these success stories and compares them with their neighbours that failed to reform to the same extent.
The question 'under what conditions post-Soviet states implement anti-corruption reform?' is most interesting to explore in countries that resemble one another on a number of political and economic indicators but differ in the extent to which they have pursued anti-corruption policies. In particular, this paper asks why countries that demonstrate so much similarity -Estonia and Latvia on the one hand, Georgia and Armenia on the other -pursued anti-corruption agendas with different levels of rigour and commitment, which hence explains why they significantly differ on corruption scales? The method of paired comparison of most-similar cases allows for the controlling of a number of variables [4] that usually affect levels of corruption. This includes political systems or the degree of democracy/authoritarianism [5], political and economic competition (Schumpeter, Washington Consensus), colonial heritage and age of democracy [6], institutional legacy [7], the existence of natural and mineral resources [8] and size of the country [9]. The case selection also controls for economic capacity, prospects for European Union accession, and international trade openness that some researchers argue are crucial in explaining diverging outcomes in corruption levels across Eastern Europe [10]. 2 By gaining variation on the dependent variable (anti-corruption reform) the paper teases out two elite-level independent variables that vary within and across the pairs and produce divergence: these are a radical break with the past and political elite cohesiveness. The goal of this paper is to give more depth and detail on this variation and the way in which different extents of break with the past and elite cohesiveness are affecting corruption. Even though this paper sheds some light on how these two factors affect corruption levels, this paper's ambition is mainly to provide avenues for hypothesis building regarding causal pathways.
First I outline the puzzle and establish that there is variation in corruption levels in respective pairs of countries in the specific periods under study. Second, I look at the extent to which the transition resulted in a radical break with the past and how the extent of this break affected corruption levels. Third, I look at political elite cohesiveness and explain how this cohesiveness varied and how it impacted anti-corruption policies. I conclude by elaborating on the findings of the comparison within as well as across the Baltic and Caucasian pairs.

Puzzle
The two cases in the Baltics (Estonia and Latvia) and the two in the Caucasus (Georgia and Armenia) demonstrate a striking similarity in a number of aspects. Historically and religiously, even culturally, Estonia and Latvia bear a strong resemblance to each other, while Lithuania deviates. Estonia and Latvia share a religion (Lutheran), common historic heritage (of being suppressed under a number of foreign rulers for many centuries), and experience of being part of the Tsarist Empire, independence in the Inter-war period and Soviet invasion in 1940. During the USSR they were regarded as the 'Soviet West.' Nowadays Estonia and Latvia are both members of the EU with burgeoning economies and pluralistic democratic systems (Table 1).
Similarly Armenia and Georgia share a history of invasions and having experienced being part of the Tsarist Empire and then the Soviet Union. Both have been a battleground for expanding empires including the Romans, the Persians and the Ottomans. Their languages are different, as are their alphabets, but in both cultures the allegiance to Christianity provided a way to sustain identity with both countries having a strong sense of being small nations that survived difficult conditions. Both ethnic identities remained strong throughout Soviet times despite Russian assimilation policies, with the table below summarising some of the similarities (Table 2).
Hence the two Baltic and two South Caucasian neighbours demonstrate great resemblance in a number of political and socio-economic aspects and the path of state formation. However the countries diverge on corruption indicators. Estonia is a distinguished achiever and excels regionally and internationally while Latvia still faces corruption problems and is more corrupt than its neighbours. Since 2004 Georgia has implemented major anti-corruption reforms in the field of public governance, has cut red tape dramatically and reformed its criminal justice system although Armenia remains largely un-reformed and relatively more corrupt than its northern neighbour (Chart 1). The chart below demonstrates the variation.
The variation in the Baltic cases dates back to the first years of the post-Soviet era 3 and hence I focus on the early and mid-1990s in order to figure out the direction of causal arrows. Even though the Baltic States had similar starting points they adopted different policies in the early 1990s that resulted in divergent paths of development. Estonia was quickly regarded as a success story [13] and the prospect of EU membership in the late 1990s stimulated greater convergence with other EU member states thereafter ( [14] p.9). The starting point in the Caucasus was much different. Georgia and Armenia had a reputation for being the most corrupt republics in the Soviet Union as opposed to the Baltic countries where corruption levels had been much lower even in Soviet times, with much of the illegal practices having survived the aftermath of the Soviet breakup. Armenia was never a reformer despite differing corruption levels in the 1990s. An important moment of divergence happened after Georgia became the 'anti- 3 The victimisation surveys that identify the percentage of the respondents who actually experienced corruption is an useful tool to quantify the variation in corruption levels. Chatterjee and Ray [12] use International Crime Victim Surveys (ICVS) to come up with the 'bribe rates' that is the number of bribe victims divided by the number of respondents. The victimisation is based on the individual's response to the question: [During the past year] has any government official, for instance a customs officer, police officer or inspector in your own country, asked you or expected you to pay a bribe for his services? The data shows striking variation between Estonia and Latvia. In 1995 the bribe rate for Estonia is 0.036 as opposed to Latvian score of 0.126 and in 1999 the rate for Estonia is 0.051 compared to Latvian rate of 0.136. Further, DEMSTAR survey provides evidence to further quantify the variation. According to the survey (semi-structured interviews conducted faceto-face interviews in 2000-2002) of the high-ranking administrators and politicians holding office in the 1990s in Estonia and Latvia, 28 percent of Estonian respondents agreed that misuse of state position was common in Estonian politics and 70 percent disagreed. This needs to be contrasted to 72.5 percent of Latvian elite reported the misuse of status and 25.5 percent denying it. corruption achiever' after its Rose Revolution and hence I concentrate on this process unfolding from 2003 onwards. 4 The Rose Revolution in Georgia is an obvious intervening variable, a critical juncture that provided a window of opportunity for the new political elite but this is not self-explanatory because similar 'revolutions' led to dire consequences in post-Soviet Kyrgyzstan and Ukraine where new governments largely failed on the anti-corruption front [17]. There are a number of factors that made the Rose Revolution a breaking point in fighting corruption and I focus on those variables instead. In Georgia a group of young reformers led by US-educated lawyer Mikheil Saakashvili launched radical reform and had soon been credited by the World Bank's's [18] Doing Business report as a top reformer. In contrast Armenia experienced only superficial reform and success against corruption has also been limited.

Independent variables and causal mechanisms
It is beyond the scope of this article to discuss the incentive structure of Georgian and Estonian elites relating to why they chose to fight corruption rather than benefit from it. The thinking of political elites about the reform especially its pace and depth was important since in both cases there was a shared understanding in top leadership that window of opportunity was short and the change had to be as fast and as radical as possible ( [19]. Hereby I will not seek to account for these 'causal beliefs' ( [20], pp.27-29) but I rather take the willing and committed leadership as given and look at the conditions under which this willingness and commitment led to the reform. The elites in other post-Soviet countries, for example in post-Maidan Ukraine could have also capitalized on window of opportunity and invest their political capital in fighting corruption. Moreover, there are some indications that the certain members of 4 According to CRRC [15] survey 6 % of the Armenian respondents report paying bribe over the last 12 months compared to 0 % in Georgia. The international crime victimization surveys of 2010 and 2011 indicate that Georgia shows one of the best results in Europe -only 0.2 % of the population mentioned that one or more officials had asked them to pay a bribe in the course of last year, down from 16.8 % in 2000 and 3.5 in 2004/5 [16]. EBRD surveys further triangulate the variation.In Georgia the percentage of the firms indicating corruption as a major constraint decreased from 41  governing elite of post-Maidan Ukraine are willing to change the corrupt system but they have been failing so far due to surviving power of oligarchs and lack of elite homogeneity [21,22]. In cases of Estonia and Georgia we are looking at the transition away from particularistic regimes that is defined as ' a mode of social organization characterized by the regular distribution of public goods on a nonuniversalistic basis that mirrors the vicious distribution of power within such societies' ( [23], p.87). This condition of affairs is referred as 'patrimonialism' by Weber [24], 'limited access order' by NWW (2009) and 'extractive institutions' by Acemoglu et al. [25]. However none of these authors provide clear insights regarding the conditions under which the transition is possible.The logic implied in Acemoglu and Robinson as well as in NWW is that most of these transitions rests on one crucial factor -the empowering of larger segment of society that leads to more inclusive institutions. Ironically this does not necessarily apply to post-Soviet achievers. In their phases of radical reform Estonia systematically excluded ethnic Russians and both Georgia and Estonia excluded Communist time networks and most of the ranks of previous government. In both cases one can observe a closely-knit group of only a handful of reformers rather than broad coalition of different groups.
I focus on two inter-related independent variables: radical break with the past and elite cohesiveness. A leadership that is not constrained by organised interests has a better chance of bringing change. The success of some authoritarian states in fighting corruption has been partially explained by elite homogeneity. As Scott [26,27] has argued historically corruption under cohesive elites (early Ayub Khan period in Pakistan, early Naser Egypt, early Ataturk Turkey) were relatively low compared to  [28][29][30]. One root of the link between fractionalization/heterogeneity and corruption can be intra-governing elite fragmentation that leads to infighting and turf wars and undermines the government efficiency. Homogeneity may also neutralise the so-called 'spoiler trap' because of 'the spoilers' limited access to power and hence a lack of opportunity to oppose reform. This is commensurate with the literature on 'Weak Government Hypothesis' [31][32][33][34] that argues that fragmented governments are less effective at responding to challenges, and that fractionalization decreases their decisionmaking capabilities. Timothy Frye (2010) in his seminal book Building States and Markets after Communism argues that the degree of political polarization, defined as the policy distance on economic issues between the executive and the largest opposition faction in the parliament helped set the pace and consistency of reform in the post-communist era. Frye sides with the earlier work of other scholars who argued that nomenclature elites are interested in dysfunctional states or 'partial reform equilibrium' that would maximize the concentration of rent to themselves [35]. This is inherently linked with the potential need for break with the past that would allow meaningful reform to happen ( [36], p. 12). Libman and Obydenkova [37] show that Russian regions where the share of communists was larger in the Soviet period were also much more corrupt twenty years after the collapse of the USSR. This can be partially explained by the survival of representatives of the ancien régime in power. Arguably counter-elites that are less beholden by the remnants of a previous regime have better chances to reform [38]. A radical break with the past allows less politically-polarised entities to emerge that reform faster (Frye 2010) and reduces the number of veto players that may block reforms [39,40].
Here is the casual pathway for each of the independent variables: first, 'nonpacted' political transitions 5 (2003 Rose Revolution in Georgian case and transition in the aftermath of Soviet breakup in Estonian case) produced radical changeover of governing elites. 6 This changeover also meant the breakdown of the collusive links between the state and the business. Elite reshuffle and breakdown of state-business nexus dramatically decreased the influence of old guard in posttransition milieu and hence neutralised its capacity to undermine reforms. The unrestrained leadership was able to deliver. Second, elite cohesiveness caused little intra-elite polarization, hence political leadership having consensus on major goals acted efficiently and could implement reform.
These independent variables are related with each other. Before taking power the counter-elites in Estonia and Georgia were a closely-knit group of individuals established under the strong belligerence to the existing regimes and this was a reason why they excluded most of the old guard from governance structures once they prevailed. On the other hand, the nature of transformation, that is revolutionary transition in both cases of Estonia and Georgia, allowed the counter-elites to fully consolidate power without the negotiated pact with the ancien regime. A break with the past was possible because the new elite was cohesive, while the radical nature of that break gave the elite that was behind it an incentive to cohere. The dependent variable here is the anti-corruption reform. The reform in Estonia and Georgia was 'Big Bang' type of all-encompassing reform that implied transformative change in all spheres of public governance while in Latvia and Armenia the reform was very limited, if any. Hence we are looking at broader public sector reform rather than an isolated anticorruption policy only. In fact there is no anti-corruption policy that would effectively address the problem but rather the transformative change in political economy is required [44]. This essay will now try and explain this variation.
Explaining divergence -enduring networks vs. radical break with the past This section will demonstrate the degree to which the old guard influences politics in the post-transition period, which has varied in Estonia compared to Latvia and in Georgia compared to Armenia. It will further show that unrestrained leadership was much more efficient in implementing anti-corruption policies while the influence of the old guard has undermined reform efforts.
In Estonia reforms had already been lunched by the Savisaar government in 1991. The flagship policy of the government was price liberalisation reform but the government was weak enough to succumb to some pressures from various lobby groups that led, for instance, to the softening of agricultural reform. Tiit Vähi's government followed up with easing bureaucracy, implementing currency reform, lifting various regulations and liberalising trade. However the Vähi government harboured a number of officials with a more traditional socialist outlook and Vähi and his team favoured the notion of a third way between capitalism and socialism [45].
The role of former top and middle level communist officials in the Estonian political leadership at the end of 1991 was quite significant [46,47]; however this had changed with the first post-communist parliamentary elections in 1992 and the nomination of 32 years old Mart Laar as prime minister. The elections were won by the right wing Pro Patria party that got 29 out of 101 seats. Pro patria has become the leading force in governing coalition established with two other political parties. Importantly the coalition had only one vote majority [48] and unlike the Georgian case described below Laar's government had to work hard to push the legislation through the Parliament.
The party elite supported radical marketisation and Westernisation of Estonia's economy. The Estonian leadership viewed the fight against corruption through promoting market economy reforms and deregulation, and by also cutting ties with the country's Soviet past [49,50]. The radical overhaul of bureaucracy was implemented that brought mainly new and young people in government institutions [51]. Laar [49] recalls in his memoirs that it would have been impossible for the old appratchiki to adapt to new requirements: 'People who have based their entire career not on honest work but on lies and deceit, it is unrealistic to expect them to change… the old system can only transplant corruption'.
Radical reforms started immediately after Laar's election -flat rate was introduced, large scale privatisation launched, and import and export tariffs were drastically reduced [52]. The number of ministries decreased by one third and their staff and functions rationalised. Some institutions like county councils, where Soviet nomeklatura held especially strong positions, were abolished [53]. The government eliminated subsidies, cheap loans and other such privileges that could be obtained through bribery of public officials. The Estonian government widely used external expertise partly because of a fear that more experienced local cadres could sabotage reform initiatives [54]. Estonia passed a radical policy towards Soviet era judiciary, replacing most of it and starting practically anew with new magistrates trained within a matter of a few months [55]. The de-facto lustration, which took place almost immediately after independence and the lustration law in 1995, eliminated the influence of the old guard while in Latvia these efforts had been much more fragmented, and according to some researchers [56] the Soviet secret police, the KGB, had infiltrated a number of new structures post-independence in politics and the economy.
In contrast to Estonia's experience, Latvian reforms advanced very slowly with much confusion and a lack of cohesiveness. Monetary reforms and a liberalisation of prices were successful but as discussed below privatisation policies had its drawbacks and the country's dated industries were slow to adopt. For instance, many of the civil servants from the old system retained their posts (Nørgaard 1999 p. 78). Between March 1990 and June 1993 two individuals dominated politics in Latvia: Anatolis Gorbunovs, the Supreme Council Chair and ex-ideological secretary of the Latvian Communist Party, and Prime Minister Ivars Godmanis. These two individuals were among the core of the new political party, the Latvian Way, which won a relative majority of seats (36 out of 100) in the first post-communist parliamentary elections in 1993. Individuals like Gorbunovs was said to have helped the old elite in the new system [57]. Sixty-eight percent of Latvian MPs in the early 1990s had previous membership in the Soviet Communist Party compared to 24 % of Estonian parliament members [58], which suggests a greater break with the past in the case of Estonia.
Due to the survival of old elites in power the state-business nexus was more entrenched in the case of Latvia, than in Estonia. Soviet industrialization policies also played a role that put an emphasis on developing light industry in Estonia (textile, woods and food processing) and heavy industry in Latvia (machine building, metal working, electrical and electronics). As a result Estonia had 40 out of 300 enterprises that were Moscow-controlled all-union enterprises while Latvia had 140 out of 400 of the same type (Nørgaard 1999, pp. 124-126). 7 The predominance of heavy industry in Latvia, that was more loyal to outside influence, generated powerful interest groups that opposed liberal market reforms. These enterprise managers pursued their own ends in the transition period of 1989-91 when the control from Moscow has been declining and local authorities did not consolidate power yet [60]. The neoliberal policies in Estonia in 1992-94 meant that non-competitive Soviet enterprises could not continue to function, which in turn allowed for a fast restructuring of the economy.
In Latvia the survival of many communist elites in power allowed the apparatchikcum-entrepreneurs to capitalise on their previous links to establish corrupt relationships with the state. The dominance of the older cadres in the economy meant the survival of old personal networks and unwritten rules of corruption from the Soviet era. More than 60 % of Latvian private businesses reported having contacts every month or more often with members of parliament compared to only 9 % in Estonia ( [58], p. 122). Norgaard and Hansen [61] showed that the actors outside the state played a large role in Latvian politics, in particular in the area of economics and finance. Several ministers reported that informal structures outside the state play a crucial role in decision-making. The impact of businesses on Estonian politics was never that substantial. For instance, Laar's neoliberal government did not require additional lobbying efforts because its ideology was already crafted to support businesses (Huang 2002) while powerful interest groups had been locked out the inner reform cycles -one of the key determinants of Estonia's success [62]. Pro Patria lacked links with local businesses partially because the party was assisted by Western partner parties which provided support through training and possibly the money for the campaign (Muuli 2012). Laar's government also refused to hand out subsidies to the businesses that was another important policy aspect that decreased opportunities for corruption [51].
The lack of influence of the old guard and the breakdown of collusive links between the state and the business allowed the Estonian government to implement a cohesive, transparent and disciplined process of privatisation under the control of one central agency that differed from the poorly-organised, opaque and unclear institutional framework in Latvia [14,46,60,63]. In Latvia privatization decisions were initially left up to the branch ministries and municipal governments before the establishment of a centralised agency in 1994. This created corrupt opportunity structures and unlashed a 'grabbing hand' of entrenched economic bureaucracy ( [58,64], p. 115). After more than three years the Latvian government was still vacillating about the method; what would be targeted for acquisition, who would manage the process and what would be needed for acquisition [65]. Despite the stricter regulations in the mid-1990s, insiders already had enough time to benefit -50 % of industry had already passed into private hands when the privatization process became more regulated ( [63], Nørgaard 1999 p.146-147). As such, the delayed privatization process allowed old nomenclature networks to regain control over the economy [62].
Interestingly some of the dynamics between Estonia and Latvia have been replicated between Georgia and Armenia. The significant divergence in corruption levels between Armenia and Georgia took place after the 2003 Rose Revolution in Georgia. In fact the degree to which corrupt networks have been entrenched in the two countries explains the occurrence of revolution in Georgia and the failure of anti-government protests to oust the regime in Armenia. Former Armenian presidents Levon Ter-Petrosian (President of Armenia in 1991-1998) and Robert Kocharian (President of Armenia in 1998-2008) both managed to reign in these interest groups to their own political and material advantage; however former Georgian President Eduard Shevardnadze (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003) failed to do so and this erosion of power in Georgia was one of the key factors that accounts for the success of the counter-elites in the Rose Revolution [66]. The 'alliance of convenience' between militaries, major businessmen and regional political leaders was strong in Armenia and so was the perception that their interests would be threatened if Robert Kocharian would be overthrown (Iskandarian 2004).
The Rose Revolution reshuffled the governing elite in Georgia. Mikheil Saakashvili got 96.3 % of the presidential votes in 2004 and his political party United National Movement (UNM) obtained a constitutional majority in parliament. The incoming elites seemingly decided to capitalise on their popular mandate to implement some painful reforms. The political transition was not the outcome of negotiated pacts with the old guard unlike Ukraine where the reform momentum after the Orange Revolution (2005) was undermined by the 'spoiler trap.' The absence of significant influence of the old guard alongside a large margin of legitimacy coupled with a zeal to transform Georgia away from 'post-Sovietness' made it possible to implement wide sweeping reforms. Unlike Estonia where anti-corruption policy was a corollary of market building, in Georgia the anti-corruption policy was an integral part of Saakashvili's statebuilding project. In Georgia new anti-corruption legislation, a zero-tolerance policy to crime, and reforms of key institutions central to combating corruption, such as the police force and prosecutor's office. 8 In November 2003, from the very beginning of his rule, President Mikheil Saakashvili made it clear that building efficient state institutions was his priority, as he duly initiated sweeping public sector reform. Saakashvili's team demonstrated an impressive level of dedication and commitment. There was a large overhaul of staff in public institutions. Overall the number of public sector employees dropped by almost 50 % while the salaries of the remaining civil servants increased roughly 15-fold. Some government institutions and/or procedures that had been generating illicit rents were cancelled, which included an anti-monopoly agency, food security agency, and technical check-up of cars. This was supplanted by general economic liberalisation policies that reduced red tape and eliminated opportunities for bribery [68]. As a consequence, corruption was substantially reduced in sectors where citizens interacted with the state most frequently, including policing, registering property, licensing businesses, and tax administration. However the Saakashvili government faced continuous criticism of 'elite level' corruption -some of this corruption was detected; during 2003-10, about 1000 public officials faced charges of corruption, including six members of parliament and 15 deputy ministers, according to Georgia's Ministry of Justice.
Armenia chose the path of 'evolution' rather than revolution but reforms lacked sustainability and took place on the eve of elections or in their immediate aftermath [69]. Most of the government's anti-corruption work is invested in drafting strategic documents and legal acts that are rarely implemented [70]. Success against corruption has been marginal even though work has been done for the sake of simplifying licensing procedures, business regulations, and administration of tax and customs. Many corrupt officials have been sacked in the tax department, customs service and police; however personnel changeover has lacked consistency and thoroughness and hence has been regarded as merely symbolic [71][72][73]). In November 2012 an address to the Cabinet President Tigran Sargysyan admitted failure in the struggle against corruption and announced intention to employ more surgical methods against corruption [74].
Similar to the Estonia-Latvia comparison, the state-business nexus is much more entrenched in Armenia than in Georgia. Initially the incoming elites in Georgia were only peripherally related with large businesses and had only one or two financiers; for example natural gas trader David Bezhuashvili. Most of the licit and illicit economic resources were monopolised by members of the Shevardnadze family [59]. Hence incoming elites were less beholden by economic interests in the period during which most radical reform took place (2004)(2005)(2006)(2007). The parliament elected in 2004 mostly consisted of young activists from National Movement and only a few businessmen could be found among the ruling UNM (United National Movement) faction. However this changed as re-privatisation proceeded and as new incumbents developed vested interest. Re-privatisation destroyed the old oligarchic class but developed a new type of interest groups that either maintained close links to the incumbents or were part of the ruling team. The key indication of the vested interest and 'elite corruption' was the transformation of some members of the political elite into Georgia's wealthiest individuals [75]. The government is believed to have awarded public contracts to 'friendly' companies, in exchange for political support. The owners of the same companies would be found among key financial contributors to the ruling UNM [76,77]. Even though corrupt practices at the higher levels had been perfected and old interest groups had been replaced by new ones, petty corruption had been drastically reduced.
In Armenia corrupt networks were largely endured throughout the 1990s and 2000s. Hence a window of opportunity never emerged. After President Ter Petrosian lost power in 1998 due to the revolt of his inner cycle two subsequent presidents, Kocharian and Sargisian, became keen to co-opt powerful interest groups in order to hold on to power (Navasardyan 2012). The need to co-opt vested interests result in reforms facing constraints since breaking such vicious cycles is difficult because of tightly-knit networks that are interested in keeping the status quo. In 2009 Prime Minister Tigran Sarkisian complained that the existing system would not allow for the exposing and punishing of officials who maintained extensive and illegal business interests [78].
The post-Soviet practice of securing and developing businesses through informal, and sometimes illicit, deals produced corrupt networks of politicians, entrepreneurs and criminals that completely blurred the public-private dichotomy in the early 1990s. The early privatisation policies resulted in the concentration of nearly all of Armenia's national wealth in the hands of 45-50 families that now control 54-70 % of the country's national wealth (UNDP 2001, [79]). This was exacerbated by the corrupt interest of government officials at national and local levels that acquired direct, partial or indirect control over emerging private firms [80,81]. In early 2000s the average age of an oligarch in Armenia was 50 years old, with the youngest being 48, which indicates that most of these wealthy individuals are carryovers from the communist past [82].
The business sector is organised into several informal commodity-based cartels and semi-monopolies in specific economic sectors or geographic regions. This is especially true for trade in commodities, like sugar, flour and alcoholic beverages (Nranyan 2012, [78]). These individuals enjoy parliamentary immunity and 'represent a convergence between corporate, state, and in some cases even criminal interests' [78]. Some oligarchs have openly spoken about their use of elected office for safeguarding business against actual or potential political pressure 9 .
Oligarchs have helped political authorities to rig elections and suppress opposition. The involvement of their bodyguards in bullying opposition politicians, voters and journalists has been widely reported during elections [84]. In return for political support, Ter-Petrosian and Robert Kocharian allowed the economic elite to enter formal politics. The 2003 parliamentary and 2005 municipal elections welcomed dozens of wealthy individuals with connections to the Kocharian government into the national parliament and local administrations [85,86]. The proportion of entrepreneurs grew from the mid-1990s, of which one-third were MPs during the period 2003-7 [87].
In sum, a radical break with the past was significant in Estonia and Georgia while a continuity of entrenched networks can be observed in Latvia and Armenia. The absence of 'spoilers' in the reform process facilitated fast and efficient anti-corruption policies in Estonia and Georgia while the surviving corrupt networks have helped keep the status-quo and undermine reform in Latvia and Armenia.

Explaining divergence -elite cohesiveness vs. elite fragmentation
The extent of elite cohesiveness is another crucial factor that varies within and across the two pairs. This section will show that the structural and ideological cohesiveness of political elites was crucial for a high degree of consensus on development goals and organisation within the new authorities. This cohesiveness then significantly contributed to the overall efficiency of government policies especially in the field of anticorruption 10 .
In Soviet times nationally-oriented, indigenous elites were purged in Estonia and Latvia. In 1959 the older cadres and central leadership in Moscow eliminated younger Latvian leaders in the Central Committee that were subsequently replaced by Russians and Moscow-loyal Latvians. This accentuated Russian domination as well as destroyed more nationally-oriented political elites. Similar purges had already expelled native elites in Estonia in 1949-1952(Krumins 1990; [89,90], pp. 99-100), which had longer term implications. For instance, by the late 1980s, ethnic Estonians held 82.2 % of administrative managerial posts while ethnic Latvians held 63.1 % [91].
The crucial difference between Estonian and Latvia elites in the period under scrutiny is elite cohesion. Steen [58] found very close integration among Estonian elites to which he refers a 'tribe feeling.' As DEMSTAR surveys of Baltic elites show acting and former ministers in Estonia demonstrate a greater trust in one another compared to their Latvian counterparts [61] that suggest more elite cohesiveness in Estonia. Unlike Estonia more-heterogeneous Latvia lacked consensually-unified elites 9 Backhaus [83] documents a similar outcome in the former German Democratic Republic. The emerging business networks could keep close informal ties to the former communist members in the bureaucracy and political elite, leading to higher corruption. 10 Saakashvili was inspired by economic model of Singapore and his role models were strongmen/statebuilders Mustafa Kemal Atatürk and Charles de Gaulle rather than democratisers Vaclav Havel or Thomas Jefferson [89]. While Laar [49] was inspired by Ronald Regan's 'Reaganomics', Lescek Balcerowicz's shock therapy in Poland, Margaret Thatcher's reforms and Milton Fridman's writings on neoliberalism. and had first to establish a kind of platform for coordinated national actions that was known as Club 21 in the late 1980s, which later formed the basis of the major political party, Latvian Way. Mart Laar's team was a group of young liberal intellectuals. Laar, a former history teacher, was a radical with clear views on general policy priorities [92]. Laar's team had roots in embryonic civil society of Soviet times that formed in 1985-87 around issues of protecting the environment and cultural heritage and displayed a large degree of ideological homogeneity [52]. Inspired by neo-liberal market reforms elsewhere Laar put an emphasis on the importance of big-bang approach that aims to replace a planned economy with a market economy in a single burst of reforms or reform 'in one breath' as he puts it. Laar claims that his team's youth and simple pro-market zeal was crucial to its success [49]. Indeed the Estonian ruling elite was much younger compared to its Latvian counterpart. The Estonian Minister of Defense Indrek Kannik was 28 and the Minister of Foreign Affairs Juri Luik was just 27. This generational change touched general ranks of the bureaucracy -in 1994, 23 % of government employees were aged 22-34 compared to only 10 % in Latvia (Steen 1994). In Estonia and Georgia younger elites were important in achieving the political sea change, which would take place. 'Maybe I was too young' Mart Laar said in an interview recently, 'but maybe, because I was so young, I was just crazy enough to dare.' 11 Younger staff frequently lacked experience but had the motivation, new ideas and alternative visions, which were significant assets in having a positive impact on the efficiency of government institutions. Youth mattered greatly as it acted as a measure to the extent to which individuals were still captives of the Soviet past. Age can also be related with strategic time horizons when younger leaders are more interested in curbing corruption for the sake of 'establishing the brand name' and having a safer exit strategy (Manow 2005). More research however is needed on causal mechanisms on how the age of elites can influence anti-corruption policies.
In contrast to Estonia, the Latvian political elite in the early 1990s was an amalgam of young career seekers, members of the Latvian diaspora and old nomenclature members [93,94] and hence was rather an electoral platform lacking an ideological profile, unlike its Estonian counterpart Pro Patria. The role of émigré Latvians drastically decreased among the party leadership partially because these individuals were not considered as insiders by the former Soviet nomenclature ( [95], p 68).
Successful reform effort in Estonia hence was led by a structurally-and ideologically-cohesive, integrated, and young political elite lacking a communist background. This guaranteed consensus over goals and a high degree of organisation within the elite increased government efficiency. As a result the Estonian government demonstrated a striking capacity to manage the transition process, including responding to the challenges of privatisation as described above, drafting new legislation, etc. For example a 1995 EBRD transition report scores the effectiveness of legal rules in Estonia at four (out of five) compared to two for Latvia [96]. Efficiency was undermined in Latvia by the heterogeneous elites with Soviet nomenclature hangovers and a lack of ideological unity and intra-elite trust.
Similar to Estonia, there was a generational change in Georgian ruling elites. Many members of Saakashvili's team had had a civil society background -for instance Interior Minister Ivane Merabishvili, Secretary of the Security Council Giga Bokeria and Tbilisi Mayor Gigi Ugulava. Like Laar's government, the core of the team in Georgia demonstrated a high degree of ideological coherence rooted in neoliberal ideology with strong right-wing and market-friendly elements [97,98]. Saakashvili was only 37 years old when he took office and many of his cabinet members were in their late 20s or early 30s. Furthermore, the government succeeded in attracting young educated personnel in middle and lower level of bureaucracy with competitive salaries. Many of these young bureaucrats were highly motivated, full of new ideas and, more importantly, zealously committed to institutional integrity [99]. In Armenia under Serzh Sargysyan (post-2008) some young people under their 40s had been hired in government structures especially in the presidential administration, central bank, and the ministries of justice, finance, economy and education. According to International Crisis Group [100], Sargysian wanted to distinguish himself from previous governments as a reformist. However 'young reformers' is not a cohesive group, and the notion had been increasingly at odds with the old guard and oligarchs. This process of recruitment of Westernised young people was met with fierce opposition of entrenched interest groups that largely perceived it as a threat to the dominant informal system (Navasardyan 2012).
In contrast to Georgia, Armenia has been ruled by a group of individuals who had made their careers in the ranks of the young communists (Komsomol) and had hailed from the 'Karabakh clan.' The two presidents, Kocharian and Sargisian, were both from Nagorny-Karabakh, formally Azerbaijani territory under Armenian control. 12 The impact of the Karabakh conflict over the formation of Armenia's political class is important. Unlike Georgia that had been defeated twice in the 1990s in ethnic conflicts, which took place in Abkhazia and South Ossetia, the Armenian state had to reward its national heroes who gained reputation and prestige during the conflict. Many former leaders of armed movements obtained key government position in law enforcement and other structures as well as lucrative shares in various business sectors ( [101], Navasardyan 2012). The origins of some of the contemporary oligarchic structures can be traced back to the 'power ministries' during the conflict especially the Ministry of Defence and Ministry of Interior [79]. Traditionally the heads of power ministries have been the key actors in distribution of spoils ( [102], Stefes 2006). Most importantly these ministries, especially the defence sector, dominated by older officers have remained most resistant to any internal reform perceived as undermining the power of the old guard [103]. These conflict dynamics have crystallised the dominance of political and military leaders of the Karabakh movement, 'corporate body of veterans' in Iskandaryan's [104] terms, over the formal state in Armenia.
Georgia experienced a radical elite changeover that brought to power a new generation of young leaders committed to an idea of building a strong state with low corruption levels while in Armenia significant elite continuity is observed. The neoliberal agenda of political elites has played an important role in Estonia and Georgia especially in terms of cutting bureaucracy and simplifying the business environment.
Importantly, apart from neoliberal agenda, the Estonian and Georgian elite also differed from their respective neighbours in terms of their attitudes toward their former colonial patron Russia. This was a significant driver in terms of forging elite cohesiveness in these countries, providing incentive to diverge from 'Sovietness' or 'post-Sovietness' [that often implied having a resemblance to Russia] and cleaning up the criminal justice system due to the perceived threat of foreign influence. Some researchers have argued that in Estonia the consensus over the main aims of transition was achieved largely because of a general fear and actual threat from Russia to restore its rule (Sootla G. and Lääne S. 2013). This resulted in more integrated national elites in Estonia compared to in Latvia, which was less homogenous. The 2004 Baltic Barometer population survey revealed that 39 % of ethnic Latvians thought Russia posed no threat compared to only 10 % of ethnic Estonians (Rose 2005). In Latvia the ethnic Latvian and ethnic Russian populations were more integrated that further undermined the possibility of homogenous political elite. Twelve percent of Latvian elites were married to ethnic Russians while none of the spouses of the elites in either the Estonian parliament, bureaucracy or judiciary were Russian in the early 1990s ( [58], pp. 48-51). In Estonia a stronger emphasis was put on getting away from 'Soviet-ness'. Similar to Georgian case Russia as an 'anti-model' played a crucial role. In a recent interview Mart Laar [48] said 'we learnt what not to do by looking at Russia. ' Similarly Georgian elites' cohesiveness and drive away from post-Sovietness was significantly delineated by antagonism toward Russia. Saakashvili reiterated several times that the Georgian leadership was building an 'alternative governance model in post-Soviet Eurasia,' meaning a government marked by low levels of corruption, in contrast to the way Russia functions (Civil Georgia 2008, [105]). Concerns about the co-optation of Georgian security forces established strong incentives to crack down on corruption in law enforcement structures and partially delineated a major clean-up in the Georgian police and security forces (Light 2013, Slade 2013). In contrast, Armenia is politically and economically more dependent on Russia and its elite are not in full agreement on the need to get away from its former colonial patron. The study of Armenian elites [87] reveals only a slight difference among those who prefer EU integration (26 %) and those who prefer integration with Russia and CIS countries (22.5 %) while the absolute majority of the respondents from Georgian political parties interviewed in 2010 looked positively at EU integration [106]. One crucial implication is that there is no external push to change anything in the way the Armenian government runs the country (Stefes 2009) while the experience of Central/Southeastern Europe suggests that close ties with Europe provide a means and motivation for EU democratic conditionality that in turn makes political systems more transparent and accountable [107].
However this is not to say that international environment is more important than domestic agency. Russian leverage and linkage does not have much explanatory power beyond being an 'anti-model' that is motivating [or not] local leaders to reform. For example in terms of economic leverage -Georgia hosts many large-scale Russian investment projects as does Armenia. Estonia had pragmatic and cooperative view towards Russian investment from the very beginning compared to Latvia where less liberal economic ideology fuelled resistance to selling national companies to the foreigners. Russian investors owned most of the port and terminal infrastructure in Estonia already from the 1990s but the deals have been transparent compared to Latvian experience where privatisation of gas industry involving Russian actors was notoriously corrupt [108]. This indicates that internal conditions have more significance and delineate the behaviour of external actors.
In the same token, the emulation of Scandinavian model in Estonia and some impact of American model and various European models on Georgia is clear enough however the impact of EU and international conditionality on reform process is often overestimated [109]. Indeed Estonia had stronger linguistic ties with Finland and the possibility to watch Finnish television in Soviet times and regular ferry connections bringing Finnish tourists provided Estonians with a window towards Europe which was denied to the Latvians ( [94], Björkman 2002). Estonians had bigger diaspora in Sweden (30,000 have Estonian roots compared to only 5000 Latvians) that promoted greater links with Finland and Sweden during transition period (Nørgaard 1999). This also promoted more 'Nordic identity' of Estonia and political elites often emphasised that 'Estonia is a Nordic country.' In the 1990s top Latvian diplomats came up with the idea of 'Amber Gateway' for the Baltics emphasising Latvian view of the more autonomous Baltic region that was somewhat at odds with close alignment of Estonia with Finland ( [95], p.131). This had some impact during the transition process in the form of investment, exchange of ideas and normative diffusion [51]. However Estonian leadership had its own strong views on reform that was frequently not commensurate with the viewpoint of international advisors. In a recent interview Mart Laar [48] emphasizes that his team was not listening much to Western advise.
In the same token in Georgia the models of emulation have been diverse with confusing and contradicting references to Estonia, Singapore, Switzerland and a host of European countries [110]. Government officials would often go against the advice of international institutions and advisors. Interviews with some decision-makers suggest that the advice of European experts often emphasised a more moderate approach while the authorities had been in favour of quick and radical action. For example, the mass dismissals in 2003 were undertaken against Western counsel (Burakova 2011). European experts 'mainly wanted us to write plans and strategies' said Shota Utiashvili, former official of the Interior Ministry in an interview and 'strongly favoured slow pace of reform.' 13 Kakha Bendukidze the author of libertarian reforms was widely reported to dismiss most of the recommendations of international advisors including USAID and IMF [111].
To summarise young and consensually-and structurally-coherent political elites in Estonia and Georgia have managed to implement efficient anti-corruption policies while less coherent elites in Latvia and Armenia have been less effective in reform endeavours. The success in both cases of Georgian and Estonia is determined mainly by domestic agency and not external factors (see also [11], p. 205). 14 13 Estonia's Mart Laar: Just do it-and as quickly as possible, interview with Mart Laar, AntiCorrp. 14 Armenia fought Azerbaijan over the region of Nagorno Karabakh in 1991-94 and won the conflict. Since then the conflict is frozen and the territory remains under Armenia's control.

Conclusion
This paper looked at the most similar cases in the Baltics and the Caucasus and asked the question why similar neighbours diverge on corruption levels. It teased out two domestic elite-level factors that vary within and across the pairs: radical break with the past and elite cohesiveness. I argued that these factors drive the variation in anticorruption reform in these countries. The 'formula' that can be derived from this paper is that a cohesive and young group in power free from the influence of the old guard and antagonistically predisposed towards the former colonial patron is more likely to reform. However, it still remains to be seen whether this assertion can be generalised through other small postcolonial states somewhere not in post-Soviet space. 15 The post-Soviet experience also highlights the importance of 'big bang.' 'Just do it' said Laar in 2002, which he probably advised the same to Saakashvili after taking the role of Georgian presidential advisor in 2006. Saakashvili encouraged the new Ukrainian authorities in 2014 'to move fast in order to uproot the deeply rooted corruption' (Kyiv [112]). Slow and piecemeal reform can be undermined by the spoiler trap. However radical reform requires a willing political leadership, which may prove too costly because it may alienate different factions of the political elite and some parts of the population. This may also explain why big bang is rarepoliticians are often not willing to accept the cost of being voted out of office or becoming a victim to a palace coup. The neoliberal undertone of Estonian and Georgian reforms also played a role especially in terms of cutting red tape and eliminating low-level bribery. This goes against the argument that the main factor influencing the incidence of corruption is neoliberal reform itself [113]: even though the neoliberal agenda failed in Russia in the 1990s, the experience of Estonia and Georgia suggests that it can succeed if the process is well-managed by good leadership.
The break with the past contributes in at least two important ways. First, it led to the temporary breakdown of collusive and corrupt networks uniting leading politicians and important businessmen. The key criticism of the applicability of the World Bank's corruption definition -'the use of public office for private gain' -is that the distinction between public and private often does not exist in the post-Soviet space. This article reaffirms this assertion and also demonstrates that less influence and integration of business elites with the ruling team in Estonia and the temporary breakdown of powerful political-economic groupings in Georgia eliminated crucial corrupt opportunity structures and facilitated reform. Second, a break with the past opened up opportunities for the younger generation to occupy leading government positions. Estonia and Georgia both demonstrated a high degree of break as counter-elites prevailed in Estonia in 1992 and in Georgia in 2004. Other cases of less intense changeover of political elites when the old guard maintained influence demonstrate that a radical break is important. For instance in Kyrgyzstan and Ukraine there was some reshuffle of elites (since the Orange Revolution in Ukraine in 2004 and the Tulip Revolution in Kyrgyzstan in 2005) but this did not lead to meaningful change partially because reform efforts were watered down by 'pacted transition' or a continuing influence of the old guard [17]. The more the old guard influence, the greater the power of spoilers to keep the status-quo and undermine reforms. In Latvia and Armenia the continuity of entrenched networks impeded reform processes.
The break with the past is related with another variableelite cohesiveness. The two reinforce one anothera radical break with the past was possible because the counterelites were cohesive and vice versa, while a cohesive political elite is an outcome of a radical break with the past. Saakashvili's team resembled Laar's governmentboth teams were ideologically and structurally coherent, composed of long-time friends with roots in civil society and having a high degree of inter-personal trust. The cohesiveness of elites results in efficient management of the reform process and well-coordinated action to deal with related challenges in both countries. Laar and Saakashvili jumped on the windows of opportunity provided by the transition and invested their political power in creating well-functioning institutions.