Access to Justice in Clientelist Networks

Despite its importance for poverty reduction, the poor face barriers when accessing justice. However, the literature, looking mostly at developed countries, focuses on the institutional underpinnings of these restrictions, and thus ignores social structural barriers. This paper deals with one type of social barrier; clientelism and its impact on poor clients’ access to justice. I argue that asymmetric power distribution enables patrons to bar clients from accessing formal institutions. Such barriers can’t be rectified through institutional reforms. Making use of an original dataset from Pakistan this paper recommends an unorthodox policy solution; increasing connectivity. I dem-onstrate how exit options, brought about through connectivity, alter bargaining powers, thereby limiting patrons’ ability to block clients from accessing formal justice institutions.


INTRODUCTION
Access to justice is a pressing issue in the poverty reduction debate. The United Nations classifies it as one of its Sustainable Development Goals for 2030 and sees it as part of its wider aim of promoting a just, peaceful and inclusive society by building "effective, accountable and inclusive institutions" (UN SDG 16). The World Bank too has focused on initiatives to improve access to justice, taking the form of training programs, auxiliary support -through capacity building programs -and aid packages. These all stemmed from the realisation that poor households do not always have unrestricted access to formal dispute resolution bodies. Therefore, the aim is to 'legally empower' the poor to assist them in pulling themselves out of poverty.
This idea gained so much traction that the UN established a Commission on Legal Empowerment of the poor, defining legal empowerment as: "the process of systemic change through which the poor are protected and enabled to use the law to advance their rights and their interests as citizens and economic actors" (pg 2, UNDP 2009). 1 It found 4 billion people around the world to be poor because they are excluded from the rule of law and therefore stressed the importance of legally empowering these citizens to improve their welfare. While scholars have critiqued the claims of the Commission of Legal Empowerment (see e.g., Banik 2009) there is general agreement on the importance of access to justice, not just for promoting peace but also for ameliorat-ing poverty. The literature outlines a broad range of barriers that impact the poor's access to justice -this includes barriers to accessing both the police and courts. These barriers are categorized in different ways, with some differentiating between physical barriers (e.g. distance to police stations and courts, infrastructure), financial barriers (e.g. costs of services), educational barriers (e.g. lack of knowledge about legal rights and services), and barriers stemming from a lack of institutional legitimacy (e.g. corruption, unequal treatment, harassment) (see e.g., McLaughlin et al. 2010). Others distinguish between barriers to the supply of justice that manifest through formal institutional inadequacies, and barriers to the demand for justice that discourage citizens from seeking out formal institutions to resolve disputes (see e.g., Sandefur 2010;Jensen 2011). Interestingly, most of these studies are based on developed countries, resulting in barriers in developing countries being understudied.
These typologies, while useful, tend to focus exclusively on identifying and categorising characteristics of formal legal institutions that generate barriers to access. In other words, almost all of the above-identified barriers can be addressed through institutional reforms, e.g., by strengthening the rule of law, increasing legal aid, reducing costs, eradicating corruption, and improving police practices. Indeed, the vast majority of programmes aiming to increase access to justice in developing societies consist of top-down reforms of formal legal institutions (McLaughlin et al. 2010).
While institutional reform is undoubtedly important to improving access to justice, this perspective tends to ignore the social structural barriers that can inhibit individuals from exercising their right to justice. These barriers stem from social conditions or inequalities that exist outside of formal legal institutions (Sandefur 2010). For example, sociological research has highlighted how the structure and strength of social networks can facilitate or inhibit access to public goods, such as justice (Slocum et al. 2010). Characteristics such as socioeconomic status, ethnicity or caste, education, and gender can hinder awareness and access to formal justice procedures (Hatıpoğlu-Aydın and Aydın 2016; Vapnek et al. 2016). While of course some forms of social barriers are linked to formal institutional barriers -e.g. social inequality is often linked to gender (women tend to have restricted access as compared to men), and ethnicity and caste often determine the extent of usage of formal institutions (Sandefur 2010;Jensen 2011) -barriers arising from social structures, networks and characteristics tend not to be addressed by formal institutional reforms.
This paper aims to fill this gap by dealing with a particular form of social barrier; the presence of asymmetric power relations and their impact on poor households' access to formal justice both through access to the police and courts. In the context of rural development, 2 the existence of clientelist networks -an exchange relationship based on high levels of inequality -means that households are often dependent on their patron for access to goods and services, such as justice. This access can take the form of either the patron providing informal dispute resolution or determining how clients can approach the police and courts. The outcome is patrons exercising considerable control over conflict management -both formal and informal -which they can dole out to selected clients (Auyero 2000). This results in the erection of additional barriers to accessing formal justice which not only cannot be rectified through institutional reforms but may even limit any potential benefits accrued from such reforms. This paper investigates these claims using unique data from rural Punjab, Pakistan. Specifically, I examine the direct and conditional effects of social barriers on inhibiting access to justice, as measured by the use of police, by exploring the conditions under which hierarchical networks are detrimental to poor villagers' access to formal justice. Given that institutional reforms have a limited impact on social barriers to justice, the paper looks at alternative ways of increasing clients' access to formal policing. Specifically, I look at how socio-political outcomes in a rural economy change when the village is connected to the external market.
The paper aims to make two contributions to the literature. Firstly, by focusing on Pakistan it adds to our understanding of barriers to justice in the Global South. Secondly, it highlights the important role social barriers play in limiting poor citizens access to justice. The remainder of the paper proceeds as follows. Section 1 gives a brief overview of the literature, highlighting the theoretical and empirical research on barriers to justice. Building from the typologies outlined previously, I distinguish between two sources of barriers to justice: institutional and social. Institutional barriers incorporate all barriers that stem from deficiencies in formal institutional characteristics, including physical access, financial costs, and lack of legitimacy. While social barriers can be of many forms -social structural conditions, social capital, and status to name a few -I focus on the role hierarchical relationships play in poor households' access to justice. Section 2 uses data from rural Pakistan to understand how asymmetric power relationships impact poor households' access to formal justice. Making use of a natural experiment in the form of a motorway constructed in 1998 the section tests the validity of using connectivity as a means of improving poor villagers' access to justice. Section 3 concludes the paper.

BARRIER S TO ACCESSING JUSTICE Institutional barriers to accessing justice
Scholars have identified a wide range of institutional barriers that inhibit access to justice. While these barriers emerge from a range of institutional characteristics and points of contact -e.g. geography, legal structure, procedure, treatment, and organizational culture -the underlying theoretical mechanisms deterring access are similar. According to the legal mobilization literature, individuals make decisions on whether or not to access justice based on a calculation of costs and benefits (Sandefur 2010). Certain institutional characteristics can increase or reduce costs. For example, the availability of free legal services and ease of access to services can reduce the costs of seeking justice. Likewise, an organizational culture of corruption and harassment may increase the costs of going to the police because individuals fear having to pay bribes and/ or not being heard by the police. This section reviews three major types of institutional barriers highlighted in the literature: physical, financial, and legitimacy barriers. Physical barriers prevent individuals from accessing judicial buildings (Vapnek et al. 2016). In many developing societies, these are often located in cities far from rural citizens. Travelling to them can take hours or even days, which can result in lost income and high costs for a single trip (UNDP 2005). In Afghanistan, for instance, most of the judicial services and courthouses are located in major urban centres, meaning rural Afghans must travel long distances on largely unpaved and dangerous roads ( Jensen 2011). In addition, Khan (2000) notes how this distance impacts quality as judges have limited access to legal texts and have little oversight.
The financial costs of judicial services can be prohibitive for potential users, particularly for the rural poor who incur costs related to travel as well as services (Vapnek et al. 2016). Slow judicial procedures and backlogs can prolong court cases for years, further increasing costs (Maru 2009). Therefore, programs aiming to reduce financial barriers typically involve reforms to reduce the length of proceedings, providing free legal aid and services, and implementing "mobile courts" that operate in rural areas free of charge (Maru 2009). However, aside from formal costs, users may also face high informal costs due to corruption within judicial institutions. For instance, a survey in Southern Punjab, Pakistan, found that most respondents had to pay extra fees to judges and court administrators which they believe reflect corruption (Galway Development Services International 2015).
Access to justice may also be restricted due to diminished demand. Judicial institutions that lack legitimacy can foster distrust and fear of formal procedures and agents. Nivette and Akoensi (2019) highlight how lacking confidence and trust in police and courts can reduce the demand for formal legal remedies when engaging in conflict resolution. A lack of legitimacy may stem from multiple institutional deficiencies, such as corruption, unfair treatment by police or courts, harassment and misconduct, as well as ineffectiveness (Tankebe 2010). Avakame et al. (1999), for instance, find that inequality in America affects a household's decision to call the police, and it affects the police's decision to make an arrest. Moreover, Jackson et al (2014) highlight how pervasive issues of corruption, harassment, and unequal treatment among police can alienate individuals from formal institutions. While much of this literature focuses on developed countries, there is growing evidence to suggest that distrust in legal institutions is a significant barrier to accessing justice, particularly the police, in developing societies (see Tankebe 2008 andJackson et al. 2014).

Social barriers to accessing justice
Aside from institutional barriers, the poor can also face social barriers that prevent them from accessing formal dispute resolution bodies. These range from barriers arising due to caste, race, gender, social status, the absence of social networks, inequality, etc. Literature dealing with justice and policing, while acknowledging some of these barriers, tends to focus on their institutional underpinnings, thereby ignoring their social causes. In particular, institutional reforms fail to address barriers arising due to asymmetric power relationships such as clientelism.
Clientelism is an exchange relationship between a group of unequals born out of necessity and mutual benefit (Mason 1986). On the one hand, the network consists of clients who own very little, if any, assets and are under constant threat from fluctuations in their environment (Powell 1970), which they try to ameliorate by approaching a figure of authority to act as their 'patron' . The patron, on the other hand, is someone with a command over resources -both material and social -who can ensure that, at the very least, clients' basic needs are met by offering goods and services such as employment, housing, credit, social insurance and dispute resolution (Scott and Kerkvliet 1977). In exchange for these services, the client is expected to provide a guaranteed supply of cheap labour, social following, and political support by agreeing to vote in accordance with the patron's wishes (Scott and Kerkvliet 1977).
The rate at which these goods and services are exchanged is determined by the relative bargaining powers of the two parties. By definition, of course, the patron has the upper hand in this relationship as the needs of the clients tend to be vital, while those of the patron are desired, but by no means life-threatening (Eisenstadt and Roniger 1980). Also, the dysfunctional nature of most developing country governments increases clients' dependence on their patrons (Stokes et al. 2013). Thus, while as a collective, clients are very valuable to patrons, individually the value, and thus the bargaining power, of each client is considerably low (Mason 1986). It depends on the needs they aim to fulfil, the number of other clients with whom they have to compete for patronage and the alternatives they have for fulfilling their needs. Therefore, the greater the economic and social inequality, the stronger will be the patron's control (Waterbury 1977).
This imbalance allows patrons to extract excessive rents form the exchange. In fact, Basu (1986) argues that patrons with monopolistic control over assets can influence clients' independent third-party interactions. 3 This can allow them to block any interaction that threatens their power. For the purpose of this paper, it means that monopolistic patrons should be able to restrict their clients from accessing the police and courts directly, and instead require clients to approach them with their problems. Clients have no option but to comply as not doing so risks them losing access to patrons' resources.
Patrons' incentive for doing this is to consolidate power by increasing clients' dependence on them. Their ability to limit clients' access to the state is aided by their privileged access to the state themselves. Harriss-White (1997) argues that the elite in developing countries create a "shadow state" where they place themselves between the state and the poor as an intermediary. This intermediary class can then use their connections with the police and politicians to extract rent from their poor clients (Khan 1997).
However, Shami (2012) argues that Basu's model of clientelism is one extreme of the power spectrum, with the other end consisting of a relationship more cooperative in nature, where patrons help broker goods and services for their clients. A core driver for these two types of clientelism is whether clients have outside options -either in the form of other patrons or opportunities outside of the village economy. When they do, patrons find their exploitative powers curtailed and instead need to provide greater benefits to entice clients to join, and stay in, the network. I argue that one of the services patrons can provide is informal dispute resolution. 4 Interestingly, while both extremes result in social barriers arising against the poor using formal dispute institutions, they differ significantly in the mechanisms through which the barriers are erected. In the case of monopolistic patrons, clients find their access blocked by patrons requiring interaction with formal institution be directed through them. Conversely, in the case of cooperative networks clients may find it more effective to approach their informal network. Therefore, while the former structure restricts access, the latter results in limited demand for formal institutions. The effect of institutional reforms on these two types of social barriers would also be different. In the case of exploitative clientelist networks, institutional reforms, even when successfully implemented, should have a limited impact on clients' access. For cooperative clientelism, on the other hand, institutional reforms should have a beneficial impact on demand as clients start to see the effectiveness of the judicial system. Therefore, for police reforms to be successful, local power relations need to be taken into consideration, with some form of social reforms built into the policy package.
Often policies tackling clientelism have focused on politically contentious land reforms. However, I argue that patrons' monopolistic control stems not from inequality alone, rather it comes from an interaction of inequality with isolation. It is the combination of the two that leaves clients with no option but the patron for most of their needs. This dependence forces clients to meet patrons' exploitative demands, even when they interfere with their independent relationships. Therefore, I stipulate, that for most outcomes, the balance of power can be altered by connecting villages to the external market, thereby breaking rural isolation. Connectivity would bring with it new opportunities both outside and within the village, all of which would, at the very least, reduce clients' dependence on the patron for employment.
As mentioned previously, while the services of the clients are not vital for patrons' survival, they are highly desirable. One such service is a guaranteed labour supply, which in an agrarian economy is extremely valuable. Moreover, clients also provide their patrons with a political following, by voting for the patrons' chosen candidate (Stokes et al. 2013). This gives patrons leverage over politicians which can be used to bargain for public and private goods. I argue that when faced with the possibility of losing the benefits of their clientelist following patrons will alter their demands to maintain the integrity of the relationship.
Connectivity is an interesting policy tool for altering power relations as interactions need not change. However, what changes is the nature of these interactions. The presence of the external market alters clients' opportunity costs from starvation to market prices. This should limit patrons' ability to infringe on clients' access to justice systems.
The benefit of connectivity is that it should show results in the short to medium run, making it a viable alternative for countries struggling with redistributive policies. Moreover, it is unlikely to be opposed as it benefits patrons by increasing property values and it allows politicians to claim credit for a big impact project. The unintended benefits for clients should not be obvious apriori and therefore would not be a cause of resistance. The next section makes use of empirical evidence to test the validity of this unorthodox solution to reducing social barriers to justice.

EMPIRICAL ANALYSIS Empirical strategy
The main question this paper aims to answer is: what role do asymmetry of power relations play in households' access to formal dispute resolution bodies? The hypothesis tested is that clientelism creates barriers for accessing justice, by either restricting access or limiting demand. I measure access to justice as a willingness to report a crime to the police. 5 Moreover, I focus on both complete and partial restrictions. Partial restrictions would allow clients to have general access to the police but limits who they can file a case against. In order to test the validity of this claim, I use data from rural Punjab, Pakistan. Pakistan is a good choice for this study because of its high levels of economic and social inequality, which is particularly pronounced in rural areas. Within Punjab I focus on rural Hafizabad which is a riceproducing district. The district has high levels of inequality, resulting in most of the incomegenerating assets being concentrated in a few hands (GHK 2005). This inequality allows for the establishment of clientelism (Shami 2012). What makes Hafizabad particularly interesting is the variation in land distribution amongst the rich across the different villages. All villages in Hafizabad are either extremely unequal, with most of the land owned by a maximum of three households -I classify these as landlord-dominated villages -or have land ownership spread over 15-20 households -classified as peasant-based villages. 6 This distribution was historically determined by the British revenue collecting machinery and has not changed significantly since independence 7 (Nelson 2010).
While monopolistic control over land given patrons excessive power in landlord-dominated villages, 8 competition for clients in peasant-based ones severely limit individual patrons' power. Therefore, if monopolistic landlords restrict clients' access to the police then we should see higher levels of police usage in peasant-based villages when compared to landlord-dominated ones.
What further makes Hafizabad a good choice for this study is a motorway, along with its link roads, constructed in 1998 connecting previously isolated villages to towns and cities. The increase in connectivity brings with it an increase in opportunities for villagers outside of the village economy. The mere presence of these opportunities, even when not exercised, are beneficial as they make defection a credible threat. As mentioned earlier, patrons value clientelist networks and therefore, when faced with clients' changing opportunity costs should alter behaviour so as to prevent its disintegration.
The motorway is an exogenous shock to the district as its placement was determined by the federal government with no interference from actors local to Hafizabad. While the eventual location of the road differs from the initial plans, these variations arose due to concerns of geography, connectivity, and defence. The aim was to minimize the distance travelled between two major cities while avoiding some treacherous mountains and remaining far from the other main highway. Also, the initial route changed as it fell very close to an Air Force bombing range.
There were rumours of the road being moved due to elite pressures, but these do not affect this study as a) the rumours were largely unsubstantiated and b) the area in question was much further down the motorway. Furthermore, the local Member of Provincial Assembly (MPA) belongs to the party which opposed the motorway project and therefore lacked the ability to influence its placement. Lastly, the patron in Hafizabad, while 'big men' in their villages, belong to the lower middle class, thereby lacking the ability to influence a national-level project. 9

Methodology
Within the district I look at 8 villages varying by distance from the road and land tenure system. Sample villages were selected by splitting the population of villages between landlorddominated and peasant-based villages, and then randomly selecting 8 -with each type being both connected and isolated. This resulted in four types of villages -connected and isolated landlord-dominated villages and connected and isolated peasant-based villages. Landholdings in sampled landlord-dominated villages varied from 100 to 300 acres. Conversely, in peasantbased villages the largest landholding was 23 acres. The incidence of landlessness amongst the villagers was the same in the different tenured villages.
Turning to connectivity, while connected villages had the link road running through them, isolated villages were between eight and eleven kilometres from the link road. Bad infrastructure made travel time to these villages between 40 and 90 minutes by car. But the biggest hurdle for isolated villagers was the absence of transport facilities, as few outsiders frequented them. This meant that those without private transport facilities were limited in their ability to leave the village, and only did so if it was absolutely necessary. 10 Connected villages, on the other hand, now enjoy an influx of traffic which makes travelling easy and cheap. This opens up the possibility for villagers to look for work outside of the agricultural sector -a sector that is controlled by the landlord. Additionally, the increased traffic brings with it new employment opportunities within the villages -these are found in the form of car repair shops, tea shops, souvenir shops etc. The general opening of the village economy was quite evident to my surveyors and me as in connected villages we blended in and only drew the attention of roadside hawkers wanting to sell us something. In isolated villages our presence was a rarity, resulting in us being followed around by little children curious about the 'outsiders' in their village.
Within the village, data was collected from a stratified random sample of 20 per cent households. Stratification was done along biradery (kinship group) lines as the literature documents this as being a good proxy for social power in South Asia (Alavi 1972). Moreover, it is also argued that households have different relationships with the patron based on their biradery -upper class biraderies are treated better, even when they are poor, when compared to lower class biraderies. Stratification by biradery allows me to test whether within village inequality impacts households' ability to access the police. The sampling process involved a local geographer and me mapping the village and identifying the biradery of each household. The Figure below is an example of these maps. The letters in the boxes represent the biradery of the households. 9 This was evident by their spending behaviours. 10 Key respondent 2.

Connected Villages Isolated Villages
Landlord-dominated villages 2 2 Peasant-based villages 2 2 Each household was given an identification number which was then entered into STATA to generate a stratified random sample -along with extras for replacement purposes. The resulting sample size is shown in Table below. I collected the data in 2013-14 along with 12 undergraduate student volunteers. 11 The survey was conducted orally -thereby catering for illiterate respondents -and included mostly close-ended questions, and a few open-ended ones. Overall, 367 surveys -20 per cent of the total population -were completed successfully. About 7 surveys 12 had to be replaced because of respondents losing interest mid-way, or them not being available. The survey response rate was 97 per cent. 13 Also, guaranteed anonymity helped put respondents at ease when it came to answering sensitive questions. Sociodemographic characteristics of the respondents is given in Table 1. All respondents were over 18.
Alongside household surveys, I conducted qualitative key respondent interviews were conducted with the local school headmasters, shopkeepers, Maulvis (Imam), Landlords, and journalists. Information from these was used to understand the mechanisms of change.
11 They were required to speak fluent Punjabi (local language) and were given extensive training to ensure consistency. Moreover, I personally checked every survey to ensure quality. 12 No village needed more than 2 surveys replaced. 13 High response rate was due to the care and diligence on part of the students.

Sample drawn
Sample collected Aside from land tenure systems and levels of connectivity, the 8 sampled villages are fairly similar. From Table 2 we can see that close to 80 per cent of households work in the agricultural sector, yet over 50 per cent are landless. Moreover, roughly 20 per cent of households work as agricultural day labourers; a low paying and exploitative form of employment. Around 90 per cent of households self-report as being part of a clientelist network. Households' clientelist membership was measured through two indicators. Firstly, respondents were asked to name influential people in the village whom they considered their leader and thus approached regarding everyday concerns/issues. 14 Secondly, households were asked if it was part of a vote block -a group that votes collectively under the direction of a leader -and who the head of that block was. The high percentage of households self-reporting to be part of a clientelist network -in both connected and isolated villages 15 -is not surprising given the ineffectiveness of the Pakistani State, as State dysfunctionality is seen as a driving force behind the emergence of clientelism.
Furthermore, households' average monthly spending is $140 in connected villages and $152 in isolation. This sustains an average of 8 people and allows over 75 percent to consume 3 meals a day. Also, half the sampled households are headed by an illiterate person, but still over 75 per cent send their kids to school. Lastly, the nearest police station is 9 km away from connected villages and 5 km from isolated villages. 16 Therefore, if the distance is a factor, then the disadvantage should be felt more strongly by connected villages.
Ideally, studying the effects of connectivity would use data from before and after the construction of the road. However, such data does not exist. Instead, I try and overcome this disadvantage through the research design. The relative similarity of these villages allows me to use distance as a proxy for time, as it can be argued that before the construction of the motorway outcomes in connected landlord-dominated villages would be similar to those in isolated landlord-dominated villages. Any substantial changes can be attributed to the road. The use of peasant-based villages as a control allows me to rule out the argument that low outcomes are being driven by police corruption and inefficiency, as that should be felt irrespective of village type. 17 Thus, this setup lets me see (a) whether levels of police usage vary based on village land tenure systems and (b) if connectivity helps outcomes in landlord-dominated villages converge towards those in peasant-based ones. The quantitative analysis is conducted through the following logistic regression model:

Police access (Y vi )
Households' access to the police can be complete, partial or none at all. This is measured through a series of questions: Would the household file a case in the event of a dispute? Would they approach the police or their patron first in the event of a dispute? Would they approach the police alone or with their patron? Lastly, was there anyone in the village against whom they would not file a case? If yes, then the household was asked to name the person was. 18 Apart from the last question, they were all binary questions; coded as Yes = 1 and No = 0. Households with complete access are willing to file a police case against anyone. Those with partial access name someone they would never file a case against. And those without access are not willing to ever register a case.

Control for village type (MW vi , LL vi , MW*LL vi )
The model includes a dummy for isolated villages (coded 1), with connected villages being the reference category (coded 0). A second dummy captures belonging to a landlord-dominated village (coded 1) -peasant-based villages are the reference category (coded 0). Lastly, I interact with the two dummies to isolate the effect of belonging to an isolated landlord-dominated village -connected peasant-based villages are the reference category. The inclusion of the interaction term means that the dummy for isolation now captures the effect of distance on peasantbased villages alone 19 and the dummy for landlord-dominated villages captures the effect of land tenure systems only in connected villages. 20 18 The question specified that it had to be a non-family member. 19 The un-interacted term assumes that the landlord-dominated village dummy is 0. The reference category is connected peasant-based villages. 20 The un-interacted term assumes that the distance dummy is 0. the reference category is connected peasant-based villages. Controls for households' economic status (X vi ) X vi is a vector of controls for households' economic status. It includes two variables. The first is a wealth index. Following Shami's and Majid's (2010) calculations wealth brackets were computed through principal component analysis (PCA). Using the information on the structure of the residence as well as durable assets ownership such as refrigerators, fans, TVs etc, Filmer's and Pritchett's (2001) method is used to extract orthogonal linear combinations that capture common information most successfully. In the current analysis the first principal component which captures the greatest information that is common to all the variables, explains 20 per cent of the variation. This component is used to create a wealth index utilising 'scoring factors' for each of the original asset variables. The analysis includes a dummy for rich households (coded 1) and uses the middle and poor groups as the reference category (coded 0). The second variable is land ownership -a continuous variable. 21 This is included separate from wealth as in South Asia land has social power (Alavi 1972), thereby enhancing owners' bargaining power.

Controls for households' social status (S vi )
S vi is a vector of control for households' social status. It includes an index for the strength of households' social networks. This index uses PCA described above. 22 It is calculated using the following questions: How often the household meets other households? Do they socialize with people from other biraderies? Do they discuss the patron? Do they discuss politics? Do they feel any restrictions on their socialising? Granovetter's (1973) argues that diverse social networks have diverse resources/skill sets they individuals can rely on to improve their wellbeing. The model includes a dummy for strong social networks (coded 1) and uses medium and weak networks as the reference category (coded 0). Lastly, I include controls for the two main upper-class biraderies (Bhatti and Kharral) (coded 1, respectively) and two maim lower class biraderies (Muslim Shiekhs and Ansari) (coded 1, respectively). These are each included as dummy variables and their reference categories are all other biraderies in the village (coded 0).

Robustness check (NP vi )
The motorway should lead to a general increase in awareness. Therefore, the beneficial effects of the road may not be due to connectivity, but due to increased awareness. Vapnek (2016), for instance, argues that poor households' access to justice is limited due to their low levels of awareness. Therefore, Equation 2 includes a control for awareness as a robustness check. Households were asked if they had access to a newspaper, either by reading it or having it read to them if they are illiterate. Access to newspapers is included as a dummy variable (coded 1), with no access being the reference category (coded 0). The error term, v , is clustered by villages. Figure 1 looks at households' willingness to register a case with the police in the event of a serious dispute. I find that a considerably larger percentage of households in peasant-based villages are willing to use formal dispute resolution mechanisms than those living in landlorddominated villages. 23 The difference in usage is even starker when we look at the socio-economic composition of households willing to register a case in the different villages. Table 3 highlights how within this group, those residing in landlord-dominated villages are better off on every indicator when com-pared to peasant-based villages. This is despite self-reported crime rates being similar across the different villages-around 19 per cent of households report having a major dispute over the last 5 years. Hence, the data indicate that poor households in landlord-dominated villages tend to shy away from using formal dispute resolution bodies.

Analysis
However, willingness to register a case does not indicate the absence of social barriers or that a household is legally empowered. Villagers might be able to access the police but still either face restrictions/limitations or have a preference for their informal dispute resolution network. Therefore, to get a better sense about social barriers, I asked households: In the event of a serious dispute would you approach the police or your patron first? While 67 per cent of households in landlord-dominated villages said they would approach their landlord first, 55 per cent claimed the same in peasant-based villages. Figure 2 further splits the data by distance, revealing that it is households in isolated landlord-dominated villages which are most likely to approach patrons first; 76 per cent claim to approach the patron first. This is contrasted by claims from those residing in just as unequal, but connected, villages; 56 per cent claim the same in connected landlord-dominated villages. These results are statistically significant and, as Table 3 shows, robust to the inclusion of controls. Table 4 presents the results of the logistic regression model to explore whether a) village type matters for households' decision to approach their patron for dispute resolution and b) which households within the villages are willing to go directly to the police. Column 1 presents the results of Equation 1 without the interaction term. This allows us to see the effect of land tenure systems and isolation on the entire data set. 24 Starting with village-level indicators, Column 1 shows that both inequality and isolation matter -belonging to a landlord-dominated village and being isolated makes it more likely that households will approach their patron before the police. However, including the interaction term takes away the significance of both these variables, highlighting that it is households residing in isolated landlord-dominated villages which are most likely to approach their patron rather than the police. This result is robust to the inclusion of controls for awareness in Column 3. Moreover, I find that while socially active households approach their patron first, Bhattis tend to go straight to the police. Bhattis are part of the socially upper class in Hafizabad and therefore are less likely to face constraints from using formal institutions.
The beneficial impact of connectivity on villagers' legal emancipation comes out starkly when questioned why the patron is their first stop in the event of a crime. In connected landlorddominated villages 56 per cent claim to approach him because he has connections which he uses to get their work done, only 26 per cent felt the same in isolated landlord-dominated villages. Common responses in the latter villages regarding why households approach the patron were; "he is our landlord" or "he is the village head". Households in these villages communicated quite clearly that everything had to go through the landlord as he was the village head. Most households in these villages could not fathom defying the landlord. For instance, when asked why they continued to abide by the wishes of a patron who barely catered for their needs the general response was: "where would we go if we defy him" 25 or "how would we feed our children if he cuts us off". 26 Connectivity seems to have changed this as most households in connected villages claim to approach the patron out of choice rather than compulsion. This was particularly evident from one key respondent interview where I asked why he thought the landlord was proactive in improving villagers' welfare: "he does work for us because he wants our political support". 27 Respondents in these villages had an appreciation of their value in the exchange. This change in the value of villagers' services was something patrons themselves also acknowledged. When asked how the motorway had altered things, one patron from a connected landlord-dominated village responded: "Things are very different since the construction of the motorway. Previously all I had to do was express a need and villagers used to come running to help. Now it is different. Now the villagers prefer to pursue the new alternatives made available by the motorway. As a result, I can no longer expect them to do whatever I say, rather I have to give them something in return." 28 Amongst households approaching the patron around 75 per cent say they would register a case with the police. However, this does not make the patron redundant. They can still have a role   to play by either accompanying villagers themselves, or sending one of their cronies with them, to the police station. I speculate that their incentives for doing this can range from wanting to remain relevant/important in villagers' lives, to offer it as an additional service to their clients 29 or to potentially control their interaction with the police. Therefore, I inquire how households choosing to use the formal justice system approach it; i.e. individually or through the patron. Amongst households willing to register a case 56 per cent say they would go with the local patron. However, as can be seen from Figure 3, the distribution varies across the different villages; 67 per cent of households in isolated landlord-dominated villages report going through the patron and only 35 per cent of households say the same in connected landlord-dominated villages. Close to 50 per cent of households in peasant-based villages ask their patron to accompany them. Interestingly, households in connected landlord-dominated villages are least likely to go to the police with their patron. One explanation for this might lie in the changing nature of this network. Since previously this was an exploitative network, the loosening of the landlords' control may have resulted in some households choosing to limit engagement with the landlord -particularly when it comes to third party interactions. This resonates with Shami's (2012) findings, where connectivity makes clientelist networks more functional in nature, but past experience lowers villagers' desire for interacting with their patron. Interpretation of this variable is ambiguous as the dysfunctionality of the Pakistani judicial system means that some households actively ask the patron to go to the police with them. In my sample, 70 per cent of households said they found the police corrupt and ineffective. However, drawing on Shami's (2012) findings, I argue that monopolistic patrons, in a bid to maintain control, may also demand that clients take them along to file police cases. In these situations, the presence of the patron at the police station can have a censoring effect on clients. Therefore, I stipulate that how villagers approach the police can indicate varying levels of trust, independence and bargaining power. Households approaching the police alone have faith the system will solve their problems without harassment. 30 Moreover, they signal their independence from the patron -particularly in landlord-dominated villages. Thus, at face value it is difficult to know exactly why households are going with the patron; out of choice in a bid to improve their chances of success or out of compulsion.
The research design helps answer this question. Peasant-based villages are unlikely to have patrons powerful enough to demand that they accompany clients to the police station. This was evident from household interviews which claimed that the patron's presence was requested in a bid to be heard (see interviews below). 31 Given the relative similarity of these villages, there is no reason to believe that villagers in isolate landlord-dominated villages find the police more corrupt. Therefore, if we treat outcomes in peasant-based villages as a baseline, then anything above it is arguably out of compulsion rather than choice.
Looking at the results Table 5 I find that distance, rather than inequality, is an explanatory variable. Column 1 finds that households living in isolation are 16 per cent more likely to go to the police with their patron. Given the relative similarity of these villages, it can be argued that the greater involvement of patrons in isolation is not due to police dysfunctionality, but rather due to patrons' wanting to maintain control. Including the interaction term in Column 2 takes away the significance of the connectivity variable, thereby highlighting how isolation has a limited impact on peasant-based villages. I argue that once again competition amongst patrons means that clientelism in these villages is more functional when compared to those found in landlorddominated villages. What is significant is the effect of connectivity on landlord-dominated villages. Households residing in isolated landlord-dominated villages are 47 per cent more likely to have the patron accompany them than those residing in connected landlord-dominated villages. This difference is significant at the 1 per cent level. 32 These results are robust to the inclusion of a control for awareness in Column 3.  When questioned why households had the landlord accompany them to the police station explanations varied significantly by village type. Answers in isolated landlord-dominated villages ranged from: "he is our landlord" 33 to "he is the big man in the village" 34 to "he is the village head, and everyone listens to him". 35 However, the patrons' power was not due to inequality alone. In just as unequal but connected villages, households' responses were quite different. While here too people referred to the patrons' general level of influence, but this was mostly tied to him also being functional. Answers in these villages ranged from "he is the landlord and he is honest" 36 to "he is the honest head of the biradery" 37 to "he is accessible". 38 Moreover, as mentioned earlier, aside from a feeling that their patron used his connections to improve their welfare, respondents in these villages also had an appreciation of their value in the relationship (see Key Respondent interview above). Therefore, while in isolation monopolistic patrons had the ability to require clients to involve them, in connected villages that was only possible by invitation.
Moreover, landowning households are significantly less likely to go with the patron. I stipulate that their economic standing gives them the ability to approach formal bodies independently, without fear of harassment. The interesting finding in this table comes from aware households and those with strong social networks approaching formal institutions through the patron. However, analysis on the split data reveals that this is only the case in peasant-based villages. 39 Therefore aware households possibly ask for patrons' involvement with the police so as to maximize their chances of success.
Lastly, villagers can face partial restrictions to their access to justice by the landlord placing limitations on who they can file a case against. This would clearly infringe on villagers' legal empowerment, as empowerment entails being free to go to the police against anyone who harms them. Therefore, I ask whether there is anyone in the village against whom the household would not file a formal complaint, irrespective of the nature of the crime. I stipulate that households answering no to this question not only communicate complete access to the formal system, but also signal their independence from powerful individuals in the village. Amongst households willing to register a case 31 per cent say there is someone against whom they would never register a formal complaint. 40 The results are shown in Table 6.
Villagers in landlord-dominated villages generally are more likely to name a person they would not file a complaint against. Moreover, those residing in isolated landlord-dominated villages are 26 per cent more likely than those in connected peasant-based villages to abstain from formally charging someone with a crime. The significance of the isolation variable in Columns 2 and 3 indicates that amongst peasant-based villages, households in isolation are less likely to abstain from filing a case against someone than those connected. This was a bit surprising at first. However, unlike in landlord-dominated villages, here the person mostly named was an MPA not from the village. This politician was held in high regard due to his pro-poor policies.
Household-level characteristics highlight two important, and interconnected, findings. Firstly, household wealth matters -households which own land and are classified as rich are significantly less likely to refrain from filing a case against anyone. Secondly, lower-class Muslim Sheikh households are more likely to not formally file a case against some individuals. to name an individual against whom they would not file a case, when compared to households residing in peasant-based villages. The effect of connectivity is to reduce the magnitude of the effect. Column 1 highlights how those in isolated landlord-dominated villages are 13 per cent more like to have restricted access when compared to households in connected landlord-based villages.
Turning to household-level variables we can see that wealthy households are unlikely to refrain from reporting a crime against anyone. However, from Columns 3 and 4 we can see that the advantage afforded by wealth is only found in isolation. Also, Muslim Sheikh households are strongly disadvantaged in isolated landlord-dominated villages. As mentioned before, these households belong to the socially lower class and therefore are extremely disadvantaged in the village economy.
Another interesting result is found in Column 2 -nothing comes out significant in peasantbased villages. I argue that competition amongst patrons in these villages means that no single individual is strong enough to restrict villagers' access to the formal justice system. Therefore, household characteristics don't place them at a disadvantage. Lastly, in connected villages the only variable significant is land ownership -land owning households are significantly less likely to find their access restricted. This resonates with the discussion above about how land ownership carries social value in South Asia and gives the owner considerable bargaining power vis-à-vis those in power.
Clientelism has historically been described as a "loop-sided friendship" or a relationship born out of necessity and mutual benefit. However, extreme inequality in the distribution of power enables patrons to make excessive demands of their clients, some of which can be detrimental to their welfare. As Bhaduri (1973) noted, in an extreme case interaction with the landlord can leave villagers worse off than before. The problem was that once the relationship had been ini- (3) ***p < 0.01, **p < 0.05, *p < 0.1. tiated villagers were unable to break out of it. While the situation in the villages visited was not as exploitative as Bhaduri's (1973) findings, but in villages where landlords had monopolistic control patrons were able to make demands off their clients which infringed on their well-being. The data revealed how, in an effort to maintain control, monopolistic patrons restricted their clients from accessing formal dispute resolution bodies, and instead required villagers to approach them with problems. This helped increase clients' dependence on their patrons, thereby tipping the balance of power further in patrons' favour. These barriers are immune to institutional reforms and reduce the benefits those reforms hope to achieve.
However, as has been demonstrated, the power enjoyed by asset-rich patrons came not from inequality alone, but from an interaction of inequality with isolation. It is the combination of the two that limited villagers' exit options and therefore enables patrons to maximize their rent seeking behaviour. The role of connectivity has been to dampen this effect by allowing outcomes in landlord-dominated villages to converge towards those found in peasant-based villages. As the data revealed, patrons in highly unequal but connected villages were forced to cater for their clients' needs, as not doing so would cause the network to break down. Therefore at least in the short to medium term, this is an effective strategy for helping improve poor clients' access to justice.

CONCLUSION
Access to justice is vital for improving the welfare of the poor and therefore is included as one of the Sustainable Development Goals for 2030. Access to justice should limit the chances of the poor being victimized. However, despite this realisation the poor continue to see constraints on their access in both developed and developing countries. Most previous literature focuses on the institutional underpinnings of barriers to access and therefore offers solutions in accordance with this. This is possibly because of the current literature focus on developed countries rather than developing ones. While some of these findings can be generalized to developing countries, there is a fundamental difference in the intuitional setup of the two which limits extrapolation. Weak states combined with ill functioning institutions mean that the poor in developing countries face different, and arguably more, barriers to their access to justice.
The literature is particularly silent on the impact social structures have on poor citizens access to judicial systems. While these citizens too face constraints arising due to institutional failures, they have additional barriers imposed by powerful members of their community. This paper has demonstrated how the presence of clientelism can result in powerful patrons restricting clients'' access to the police. Their incentive for doing this is to maintain control and, by extension, be able to extract maximum rent of their clients. Such barriers are resilient to institutional reforms and instead need to be addressed directly. However, most conventional policies tackling inequality are politically contentious. This paper has proposed an unorthodox solution for increasing villagers' access to formal policing. I have argued that patrons enjoy exploitative powers not because of inequality alone, but because of inequality interacted with isolation. It is the combination of the two which significantly restricts villagers' options outside of the landlord. I have argued that connecting villages to the external market curtails patrons' control over their clients and thereby enabling them to access the police more freely. Villagers in connected landlord-dominated were more likely to go directly to the police, less likely to take the landlord with them and less likely to refrain from reporting someone who harmed them. In fact, for most indicators outcomes in connected landlord-dominated villages are converging towards those found in egalitarian villages. While in isolation villagers spoke of a landlord who was powerful and controlling, in connected villages similar households spoke of a landlord who helped improve their welfare and got work done for