key: cord-0669095-zwivmuyp authors: Jain, Ritika; Biswas, Shreya title: The road to safety- Examining the nexus between road infrastructure and crime in rural India date: 2021-12-14 journal: nan DOI: nan sha: 3db8d6b84b05527b201af68165de5530a33742e3 doc_id: 669095 cord_uid: zwivmuyp This study examines the relationship between road infrastructure and crime rate in rural India using a nationally representative survey. On the one hand, building roads in villages may increase connectivity, boost employment, and lead to better living standards, reducing criminal activities. On the other hand, if the benefits of roads are non-uniformly distributed among villagers, it may lead to higher inequality and possibly higher crime. We empirically test the relationship using the two waves of the Indian Human Development Survey. We use an instrumental variable estimation strategy and observe that building roads in rural parts of India has reduced crime. The findings are robust to relaxing the strict instrument exogeneity condition and using alternate measures. On exploring the pathways, we find that improved street lighting, better public bus services and higher employment are a few of the direct potential channels through which road infrastructure impedes crime. We also find a negative association between villages with roads and various types of inequality measures confirming the broad economic benefits of roads. Our study also highlights that the negative impact of roads on crime is more pronounced in states with weaker institutions and higher income inequality. Assurance of safety is one of the fundamental aspects of civil society (Iyer et al., 2012) . Living with a feeling of unsafety and the fear of being subjected to crime leads to perverse impacts on the quality of life (Dutta and Hussain, 2009) . People tend to become more risk-averse when facing significant changes in the external environment, such as a natural calamity, civil unrest, pandemic or violent crime (Brown et al., 2019) . Moreover, a crime-induced rise in risk aversion has adverse implications by restricting human mobility and reducing access to job and educational opportunities (Cook et al., 2013; Dutta and Hussain, 2009) . Accordingly, even the Sustainable Development Goal (SDG) 16 highlights that reducing various levels of organized crime is an essential tool to achieve other SDGs 1 . The extant literature underscores several other adverse economic effects of crime. Crime activities increase uncertainty, are related to inefficient resource allocation, and deters investment (Detotto and Otranto, 2010) . There is a negative correlation between crime, economic growth, and employment (Detotto and Pulina, 2013; Goulas and Zervoyianni, 2013) . Although important for developed economies, the adverse impact of crime is of utmost importance in developing economies already plagued with low growth, low investment trap, and uncertain economic environment. Further, due to informal markets, weak institutions and poor quality of infrastructure, the probability of being caught and convicted of a crime are lower in developing countries than the developed ones. Studies also document that countries that are characterised by higher inequality have a higher incidence of crime due to the following reasons (Ehrlich, 1974; Fajnzylber et al., 2002) . Firstly, in societies with a high degree of inequality, the legal wage for low-skilled workers may be too low compared to expected earnings from indulging in illegal activities. Second, the cost of crime is the combination of the probability of being caught and prison time. Again when inequality is high, one may argue that the quality of life within and outside the prison may not differ substantially, making crime a high expected return and low-risk activity compared to low inequality scenarios. Since developing countries have higher inequality (Van der Hoeven, 2019), the vulnerability of individuals and exposure to crime is likely to be high in this setup. Frequent criminal activities may affect productivity in developing economies hindering life. Thus, research on factors and interventions that may control and deter crime may be highly relevant for the developing world. 1 Source-https://www.unodc.org/unodc/en/sustainable-development-goals/sdg16_-peace-and-justice.html, https://www.swp-berlin.org/publications/products/comments/2015C45_vrr_bsh.pdf (Accessed on July 11, 2021) Becker (1986) pioneered the economics of crime and suggested that criminals were rational agents deciding whether to indulge in criminal practices based on their benefits and costs. Following Becker (1986) , a large body of literature has emerged that has explored the determinants of crime (Fajnzylber et al., 1998; Cahill and Mulligan, 2003; Imrohoroglu et al., 2006; Bunanno and Montolio, 2008) . These papers identify several attributes such as unemployment rate, urbanization extent, the fraction of foreigners, previous incidence of crime and quality of institutions. On the other hand, sociological literature focuses on how the social theory of relative deprivation may be one of the significant determinants of crime (Merton, 1968; Bernburg et al., 2009 , O'Mahony, 2018 . The theory posits that more impoverished and more unequal societies have higher crime counts due to people feeling deprived relative to their peers. Besides these socio-economic factors, spending on road infrastructure may also influence crime rates (Hughes, 1998) . Our paper also attempts to examine how building road infrastructure may impact criminal activities. We explore various channels through which roads may influence crime. First, local development through roads may lead to better employment opportunities. A revisit to Becker's (1968) model then implies that the opportunity cost of crime rises with better employment opportunities. As a result, individuals may substitute their time spent on crime with formal employment. Hence, Becker's (1968) framework implies that building road infrastructure should impede and deter crime. However, if the economic benefits of employment due to roads disproportionately favour the skilled and endowed individuals more, the unskilled ones may still indulge in criminal activities. In some instances where the benefiting group forms a minuscule share of the population, it may lead to a rise in criminal activity. Another channel that determines how road infrastructure may influence crime stems from the infrastructure development implementation itself. Roads reduce the time costs and increase mobility, both critical for criminal activities and economic activities. A well-connected road network may catalyze movements of criminals to potential hot spots with ease. These contrasting channels provide an interesting backdrop to test the empirical validation of how building road infrastructure may affect crime in developing economies. Against this background, we attempt to examine how building road infrastructure impacts crime in rural India. The focus on rural India stems from the weak infrastructure and scarce non-farm opportunities (Jha, 2006) 2 . With heavy reliance on agriculture for employment, infrastructure development in rural parts of India has been slow. Despite receiving attention in several development plans and policies since independence, the slow pace had been persistent until the late nineties. In 2000, the central government of India introduced the Pradhan Mantri Gram Sadak Yojna (PMGSY) that aimed to connect all villages with an all-weather pucca road in a phased manner. Rule-based population cutoffs determined the sequence of phases. However, in 2011, the PMGSY rollout was extended to all the villages in India. We use data from the India Human Development Survey (IHDS) conducted in two waves-2004-05 and 2011-12. Among several socio-economic attributes, the survey explored whether the household faced any type of criminal activity in the last twelve months. We use multiple measures of crime as our dependent variables. Both waves of IHDS also have a separate questionnaire for village-level amenities, population composition and occupation structure, among other attributes. Using information from the village questionnaire on whether the village was accessible through an all-weather pucca road or kaccha road or was inaccessible, we construct our focal variable-the presence of a pucca road in the village. We employ an instrumental variable estimation strategy to account for omitted unobservable factors that may simultaneously influence road and crime measures. We find that households living in villages connected with an all-weather pucca road experience 5% less criminal activities than households living in villages without it. Our effect size doubles when we control population composition, inequality and income uncertainty at the village level. A closer examination of the specific type of criminal activity reveals that the effect is limited to types of crime that have a higher possibility of happening outside home premises-female harassment and burglary. We explore several channels that may drive our main findings. We posit that roads as a deterrent to crime may work through two channels-direct effect of better street lighting, higher likelihood of bus service and increased employment opportunities. Additionally, it will also bear the indirect benefit of higher income for the households and lower inequality at the village level. We test these channels by examining the impact of a pucca road on street lighting, and bus stops in the village, employment and income status of households. We find strong evidence that households in villages with better-connected roads have greater access to public programs related to street lighting and bus services. Additionally, we also find evidence for increased employment, higher income and equal land distribution for villages with better roads. These results outline the primary channel through which roads reduce crime. We extend our model in two broad ways. These extensions are based on institutional factors and pre-existing socio-economic conditions in the state where the village is located. In the first extension, we attempt to examine if our impact is uniform for states with better quality of institutions vis-à-vis states that do not. We use measures that capture the efficacy of crime deterrence and management at the state level and divide our states according to high and low categories. We find that road infrastructure reduces crime only in states that have a lower quality of institutions. This outlines the critical importance of roads in helping these states catch up with better institutions. As a second extension, we test if our effects are conditioned by the level of inequality and the coverage of public employment programs at the state level. Again, we find that our impact is limited to relatively more unequal states and that have better coverage of public employment program. These results underline the importance of building road infrastructure in rural India. The paper is organized as follows: Section 2 lays down the relationship between crime and road infrastructure. We discuss the Indian experience in Section 3 and data and descriptive statistics are presented in Section 4. The econometric methodology is discussed in section 5. We present our results in Section 6 and conclude in section 7. The need for infrastructure is of paramount importance in developing countries due to weak institutional factors (Sawada et al., 2014) . These countries invest a large amount of resources in restructuring and building a broad and well-connected road network. An established road infrastructure setup leads to a wide range of impacts on urbanisation, population and environment. These impacts may be beneficial or harmful depending on the broader context (Khanani et al., 2021) . For instance, the building of roads and transportation services accompany peripheral residential development, the emergence of commercial establishments and other forms of spatial segregation. Additionally, infrastructural development also enhances mobility, consequently reducing barriers to labour force participation (Akee, 2006; Lei et al., 2019) , improving access to schools (Adukia et al., 2020) and health care facilities (Aggarwal, 2021) . The socio-economic benefits of roads are driven by the primary channel of reduced transportation costs and higher mobility. The resultant ease of access may open up job opportunities and public services that were previously inaccessible. For instance, Khandker (1989) finds that government investment on roads in Indian districts between 1961 and 1981 was associated with new non-farm employment and higher wages. Several recent studies have also documented the Indian experience of the creation of non-farm jobs due to rural road infrastructure investment (Aggarwal, 2018; Asher and Novosad, 2016) . However, roads may have some adverse outcomes as well. Past studies have examined the direct negative effect of environmental degradation, higher chances of landslides and road accidents (Forman and Alexandar, 1998; Paul and Meyer, 2001; Slabbekoorn and Peet, 2003; Coffin, 2007) . Besides these, it may also have non-uniform impacts across various sections of society. For instance, people who own land and vehicles may use the roads to their advantage, whereas the landless may not benefit as much. Consequently, road infrastructure development may lead to rising inequality with the impact of roads favouring the rich and endowed more than the poor (Aggarwal, 2018) . However, Ferriera (1995) argues that if infrastructure investment in underdeveloped areas increases connection between core economic activities, it may lead to more productive opportunities for the poor that may reduce inequality. In summary, the direction of the impact of roads on inequality is ambiguous depending on the context and the country. An additional aspect related to road networks that has received relatively less attention is criminal activity. Criminal activity is as spatially segregated as economic opportunities. The ease of mobility and better connectivity due to road infrastructure that benefits economic opportunities is also relevant for illegal activities. Further, the choice of committing crime stems from limited viable economic opportunities (Becker, 1968) . In a developing country like India, where 11.90% of the population is unemployed 3 , and a mere 20% of employed individuals are employed as waged and salaried workers 4 , most individuals face a lack of dignified job opportunities. While there are several contributing factors to the employment situation in India, weak and inadequate infrastructure exacerbates it. Consequently, it can be related to a higher possibility of low or unskilled individuals indulging in criminal activities in given lack of job opportunities along with low wage returns. Against such backdrop, investing in road infrastructure may be an effective tool in reducing barriers to entering the labor force and reducing criminal activity. Further, if infrastructure benefits the elite and non-poor more than the poor and economically vulnerable sections of the society, it may make the society more unequal. Road infrastructure may then increase crime as a consequence of a rise in inequality. Becker (1968) explains that to catch up with high-income individuals, crime is an easier tool for low-income individuals than the low returns from the labor market. Several papers have extended this model (Ehrlich, 1973; Block and Heineke, 1975; Chiu and Madden, 1998) . Bourguignon (2000) also finds a positive association between observed inequality and levels of crime. Bourguignon (2001) revisits this issue with a particular focus on urbanization and reiterates that crime is a byproduct of uneven economic development or processes. Inequality and poverty, even if transitory, have large and persistent societal losses through crime. A similar positive association between crime and inequality is presented in Merton (1938) . Discussed in the sociological literature, the strain and disorganization theory by Merton (1938) posits that individual alienation due to low income, marginalized status, or discrimination may lead to indulging in criminal activities. The empirical evidence of these theories, however, remains inconclusive. While and Bourguignon et al. (2003) find support for it, Land et al. (1990) , Kelly (2000) and Kang (2016) find an insignificant relationship between crime and inequality. Bourguignon et al. (2003) investigate the relation for the seven largest cities in Colombia and find that probability of being a criminal was higher for individuals living in households that had a per capita income below 80 percent of the mean. In contrast, Kang (2016) emphasizes that crime is primarily driven by economic segregation instead of within neighborhood inequality. A second dimension to the relationship between road infrastructure and criminal activities is related to the physical network of roads. According to criminology literature, road infrastructure may work as a skeletal structure to criminals that may aid in identifying and easily accessing the hot spots (Davies and Johnson, 2015) . This may perpetuate criminal activities in areas that are well connected. However, one may argue that infrastructure activities require casual employment that benefits the unskilled category of the local population. Hence, considerable investments in infrastructure projects like road construction may boost employment among the poor and unskilled, leading to lower criminal activities. The above discussion highlights the possibility of the association between road infrastructure and criminal activities to move in either direction. Against these contrasting channels, we attempt to evaluate the Indian experience of the impact of road infrastructure on crime. There are two possible ways to gather information on criminal activities in India-crime records maintained by the respective nodal agency in the country and information collected from victims directly through a survey. The former source of information is layered by several agents between the crime scene and the nodal agency. The Code of Criminal Procedures deals with criminal cases and procedures in India. According to this code, crime-related information reported to the police needs to be compiled and written in a First Information Report (FIR), then signed by the victim. Based on the FIR, the police start investigating the crime and making arrests if necessary. Once the police forces have arrested an accused individual, they are produced before the magistrate charges the person with the specified crime or releases them. Most crimes such as extortion, theft, harassment, murders, terrorism and so on are punishable under the Indian Penal Code. National Crime Records Bureau (NCRB) is the nodal agency in India that collects, compiles and disseminates information related to crime. Published by the Ministry of Home Affairs, NCRB records annual data on various aspects of crime at the state and district levels of India. The primary source of information of this data is the FIR that is filed with the police force. The information is collated at the District Crime Records Bureau and sent to the respective State Crime Records Bureau from the police station where the FIR is filed. Finally, the NCRB consolidates all this information and compiles it. Using police records of crimes from NCRB may have several limitations in capturing the pattern of actual crime, since various crimes in India go unreported due to poor quality of infrastructure, weak institutional factors and social stigma. In fact, reporting of crime is a problem across countries-according to the International Crime Victim Survey data only 40 percent of committed crimes are reported at the global level. However, under-reporting of crime is more pervasive in developing economies. Ansari et al. (2015) report that people in India do not report crimes due to the paucity of police stations, lack of awareness and inadequate trust in the criminal justice system. The study also documents that while the trend of violent crimes such as murders in India is not very different from the developed world, other criminal activities such as burglary and theft have a higher likelihood of not being reported as compared to the rest of the world 5 . Further, under reporting of crime in India, this issue is a dual consequence of victims choosing not to report and police deciding not to record it. While police records are valuable information, using victim-reported crime may reduce the scale of an understatement. Such data is gathered from surveys directly by asking respondents if they faced any crime in the past year. Prasad (2013) documents that while police-recorded crime patterns are well represented in Indian regions where institutions function efficiently, the difference in the two types of crime measures is enormous in the rest of the country. Despite experiencing high economic growth in the early 2000s, India has been grappling with a weak and inadequate infrastructure network. While necessary for urban areas, access to better infrastructure is critical for poverty alleviation and economic development in the rural sector. Within the infrastructure sector, roads have been at the forefront of economic development in India, with rural road development plans receiving attention since independence. Despite receiving considerable attention, a lack of planning, improper design and low monitoring lead to several deficiencies in the rural road network (Samanta, 2015) . Inadequate embankment and poor drainage network implied that most of these roads were not accessible during rough weather. Against this background, a centrally sponsored scheme, PMGSY, was launched in 2000. The scheme's primary objective was to provide rural habitations (defined as a cluster of the population that resides at the same location along the lines of a hamlet) with an all-weather pucca road within 500 meters. While it was a scheme introduced by the central government of India, state and local governments were active participants in the implementation of the project. PMGSY was implemented according to population criteria in a phased manner. Villages that had a population of 1000 or more were prioritized in the first phase with few exceptions. The 5 These are deduced from two rounds in which select Indian cities were included in the International Crime Victim Survey in 1992 and 2003. Ansari et al. (2015) discusses this issue in detail. second phase involved villages with a population of 500 and finally the third one for villages with 250 6 . However, in 2010 the scheme was universally opened to all villages. Additionally, the roads being built were to be connected with the core network of roads within the state. In 1951, a mere 20% of Indian villages had access to an all-weather road 7 that increased to 60% in 2000 (Lei et al., 2019) . As of 2019, the access has spread to 73% of Indian villages 8 . This broad coverage of road networks reflects that road infrastructure since the 2000s has grown. Using measures based on the PMGSY data to capture road infrastructure quality provides an accurate picture of capturing how uniform the growth has been across various regions. We use the India Human Development Survey (IHDS), a nationally representative survey of more than 40,000 households. With comprehensive coverage of socio-economic variables such as health, education, gender relations, social networks, crime, confidence in institutions and so on, the IHDS dataset is well suited for addressing the impact of road infrastructure and crime. The survey was conducted twice, wherein the first wave was in 2004-05 and the second wave was in 2011-12. Thus, the dataset consists of a household-level panel for two years. IHDS also collected information at the village level in both the waves from focus group discussions among village officials, people in business, and similar people in the village. The information spans various issues such as infrastructure, public programs, occupation structure, and population composition. We merge this village-level information with the household questionnaire. Since the village level information is collected only for rural areas, we drop all households dwelling in the urban sector 9 . Our final data set comprises more than 27,000 rural households for 2004-05 and 2011-12. The dependent variable for our analysis is criminal activities experienced by victims (households) in the recent past. IHDS has four questions on whether households faced burglary, threats, female harassment and breaking into homes or not. Based on these criminal activities, we construct two measures of crime-(i) a simple average of each of the four types of crime 6 There were few exceptions to the rule. For instance, if a habitation with less than 1000 population lies on the straight path of a road that was built for a habitation with higher than 1000. 7 Sourcehttps://niti.gov.in/planningcommission.gov.in/docs/aboutus/committee/wrkgrp12/transport/wgrep_rural.pdf and (ii) a dummy variable that takes a unit value if anyone in the household is a victim of either of the four types of crime and zero otherwise. To test the robustness of our model, we also use each of the types of crime as dependent variables in separate models. Our interest variable is related to road infrastructure in Indian villages. An ideal approach would have been mapping the PMGSY scheme implementation data with our dataset. However, due to the lack of village identifiers in IHDS data, we are unable to use the implementation of the PMGSY scheme at the village level. As an alternate resort, we exploit the stock nature of road infrastructure and use measures from the IHDS directly to measure the connectivity of villages with all-weather pucca roads. The village questionnaire asks whether the village is connected with a pucca (all-weather) road, a kaccha road or is not connected by a road at all. We use this information to construct a dummy variable for villages that are connected by an all-weather road. Consequently, villages not connected by any road or connected by a kaccha road get a zero value. We use a set of control variables at the village and household levels that may influence crime. These variables broadly span across the presence of police stations in the village, confidence that households have in institutions critical to crime deterrence and management, land distribution in the village between the largest and the rest of the caste and religion groups and other socio-economic variables. We present the definition, measures and basic summary statistics of each of these variables in Table 1 . As a first step in measuring aggregate patterns between road infrastructure and crime, we present how various economic outcomes differ between villages with pucca roads and villages without them. We compile the results in Table 2 . Table 2 presents that most crime measures have a lower value for households dwelling in villages with an all-weather pucca road than households that reside in villages without it. Further, we find that villages with pucca roads also have a higher probability of getting street lights through a public program. Similarly, this group of villages also have a bus stop that is closer to the village than the group of villages that do have a pucca road. Finally, families dwelling in villages with well-connected roads exhibit better labour force participation rates and higher family income. These patterns lean towards the possibility of rural road infrastructure being effective in tackling crime in India. We explore this further in the next section, wherein we discuss our identification strategy. We aim to examine the impact of road infrastructure on crime. To meet this objective, we estimate the following model- is the error term that captures the impact of all unobserved omitted factors. For obtaining estimates of the impact of road infrastructure on crime, road infrastructure must be exogenous in our estimation model. However, the possibility of a set of unknown variables influencing both crime and road infrastructure may lead to an omitted variable bias and consequently render our road infrastructure variables endogenous to criminal activity. This requires using an instrumental variable estimation strategy wherein the first stage we estimate road infrastructure measures using a set of exogenous variables used in (1) and an additional instrumental variable. We estimate the following equation as the first stage of our model- where all notations denote the same variables as in equation (1). The instrumental variable, denoted by , is based on the rationale that it affects road infrastructure but has no direct impact on criminal activities faced by households in that village. Since provisioning of different types of public goods, is correlated with each other at the village level, we use the proportion of households with access to piped drinking water as an instrument for building all-weather roads in villages (Banerjee and Somanathan, 2007) . We posit that piped drinking water access and providing road infrastructure may have a strong positive association, but access to piped drinking water does not have an apparent effect on crime incidence. The predicted values from model (2) is then used as an explanatory variable in the second stage, denoted by eq (1). In the presence of valid and relevant instruments, this estimation yields causal impact of road infrastructure on crime. We discuss our findings in the next section. Table 3 presents the estimates of the impact of road infrastructure on criminal activity. In Models 1, 2 and 3 we present the OLS estimates. The second stage of the instrumental variable regression models are presented in Models 4, 5 and 6.