Notes Towards a Place Based Approach for the Development of Southern Mexico ARCHIVES By MASSACHUSETTC INSTITUTfE OF FECHNOLOLGY He'tor Cuauhtli Flores-Ramirez JUN 2 9 2015 B.A. in Political Science El Colegio de Mexico, Mexico City, Mexico, LIBRARIES 2012 Submitted to the Department of Urban Studies and Planning in partial fulfillment of the requirements for the degree of Master in City Planning At the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June, 2015 @ 2015 H6ctor Cuauhtli Flores-Ramirez. All Rights Reserved. The author hereby grants to MIT permission to reproduce and to distribute publicly paper and electronic copies of this thesis document in whole or in part in any medium now known or hereafter created. Signature redacted Signature of Author DepaAjient of Urban Studies and Planning May I, Certified by S ignature VW K2 '/ 21, 2015 redacted Assistant Professor Gabriella Carolini Department of Urban Studies and Planning Th esi b S uPevisu Signature redacted " Accepted by %'- V Professor Dennis Frenchman Chair, MCP Committee Department of Urban Studies and Planning 2 Notes Towards a Place Based Approach for the Development of Southern Mexico By Hector Cuauhtli Flores-Ramirez Submitted to the Department of Urban Studies and Planning on May 2 1 th, 2015 in Partial Fulfillment of the Requirements for the Degree of Master in City Planning ABSTRACT The spatial structure of development is of both of theoretical and policy relevance given the feedback or network effects that material wealth or income inequality may have between neighboring spatial units. In order to investigate if development is spatially structured between municipalities in Southern Mexico, I perform global and local spatial autocorrelation analysis for income per capita and other key economic and social development variables. My analysis finds, through permutation tests, statistically significant evidence of spatial (positive and negative) autocorrelation clusters in these municipios for variables like GDP per capita and poverty rates, among others. This evidence supports the argument to investigate in more depth the role of space in development policy for this region of Mexico. My spatial analysis is focused on the municipios of Southern Mexico (Chiapas, Guerrero, Hidalgo, Michoacan, Oaxaca, Puebla and Tlaxcala). Southern Mexico has consistently been the poorest region in the country since the 1940s. Persistent underdevelopment and inequality is now giving rise to crime, violence and governance challenges that urge a rethinking of development policy in this part of Mexico. In general, economic and regional development research in Mexico has not explicitly modeled the spatial effect of development and it has not been conducted at the municipality scale. An increased understanding of both spatial and municipal information is crucial to implement more effective development policies. My thesis establishes a conversation both with literature on place-based development policy, and regional development literature in Mexico. Thesis Advisor: Gabriella Carolini Title: Assistant Professor 3 4 Acknowledgements For their never-ending love, I would like to thank my mother, Maria de los Angeles Ramirez Medina, and my sister, Gloria Flores Ramirez. I also thank my father, Hector Flores Quintana, and my dear Jee Seun Choi. For their financial support, I thank the Consejo Nacional de Ciencia y Tecnologia of Mexico, the Fulbright Scholar Program, the Banco de Mexico, and the Department of Planning and Urban Studies at MIT. I also thank the brilliant guidance received from my thesis director, Professor Gabriella Carollini, and the support I received from my friends at MIT. 5 Table of Contents 1. Introduction ............................................................................................. 8 1.1 What Do We Talk About When We Talk About the South?..... . .. . . . . . . 9 1.2 Why Does Development in Southern Mexico Matters? ....... . . . . .. . . . . . 11 1.3 Structure ------------------------------------------.............................................. 15 2. The Case for Place-Based Social and Economic Development .............. 17 2.1 M ethodology ................................................................................. 23 3. An Overview of Regional Development and Inequality in Mexico .......... 27 3.1 A Review of Research on Regional Income Per Capita Convergence and Inequality in Mexico ......................................... 30 3.2. Causes of Regional Inequality in Mexico and Policies to Combat It.. 36 4. The Spatial Structure of Economic and Social Development in Southern Mexico ..----.--------------------------------...................................................... 39 4-1 Spatial Structure at the State Level........................ 39 4.3 Spatial Structure at the Municipio Level..-----......-......45 Conclusion.................................. ................ 62 6 Figure 1. States mn Mexico's South AA. MM.S.A., SONOR cOAMuqLA 2 AUSoCO 4 3. OuAN wo Guaymmo 7. malmum GUATEMALA 2 4 WVC#OACA 1-A ICA CHMAPA GUATEMA 7 1. Introduction This thesis examines evidence of spatial dependence in key economic and social development variables at the municipio level for the states of Chiapas, Guerrero, Hidalgo, Michoacin, Oaxaca, Puebla and Tlaxcala in Southern Mexico.1 This analysis aims to contribute to a larger scholar and policy discussion regarding a regional development policy for Southern Mexico. By focusing on the analysis of spatial structure at the municipio level, I expect to inform the discussion regarding geographical targeting and spatial network effects in economic and social development. Regional development and inequality research in Mexico has focused in the study of state levels of income per capita measures. In particular, previous research has found evidence of a geographical cluster of relatively underdeveloped states in Southern Mexico, including Chiapas, Guerrero, Oaxaca, and Michoacan (Sastre-Gutierrez and Rey, 2008). Even though it is feasible to think of the South as a cohesive economic and geographic region (Hall and Humphrey, 2003), it is also true that, when inspecting data at the municipio level, regimes and patterns of spatial heterogeneity emerge. This finding is of both scholarly and policy relevance. In this thesis, I argue for the relevance of thinking regional development policy for Southern Mexico through the perspective of place-based policy and spatial analysis at the municipio level. Additionally, although past academic efforts have been focused on the theoretical dimension of regional development (endogenous growth theory, new economic geography, institutional economics), less attention has been paid to its implementation in real-world settings (Barca, McCann, and Rodriguez-Pose, 2011: 2). In this thesis, I also provide a preliminary outline of the policy relevance of incorporating space and geography in development policy. In particular, pondering the role of geographical targeting in development policies for Southern Mexico may have profound consequences in terms of its effectiveness, efficiency and resilience (Baker and Gorsh, 1994). For example, research on the role of geographical targeting in poverty alleviation in Mexico has suggests that the specific spatial unit utilized to implement a policy can have significant implications both on its overall efectiveness and the administrative costs of implementing these policies (Baker and Gorsh, 1994). 1 Mexico has three government levels: the federal, state and municpio level. There are 32 states and over 2,457 municipios. 8 1.1 What Do We Talk About When We Talk About the South? One of the most challenging issues regarding the research on regional development in Mexico, particularly when it comes to the South, is that there does not seem to be a generally accepted definition of which states constitute it. In fact, a key issue in the literature is the diversity of regionalization schemes for Mexico and the lack of clearly defined criteria to include certain states as part of the South and exclude others (SastreGutierrez and Rey, 2010). The importance of this issue cannot be understated given that, as has been demonstrated by Sastre-Gutierrez and Rey (2010: 295), the selection of a specific regionalization scheme can have significant effects on standard regional economic (income per capita) convergence analysis, and its fundamental conclusions. In practice, research done on regional development in Mexico has tended to overlook the importance of the regionalization scheme choice, and different alternatives have been selected thus preventing direct comparison across the literature. Among the regional schemes for Mexico that had been developed, some of them have incorporated explicit analytical criteria, like maximizing the homogeneity of states within a region given a certain variable, like poverty rates or income per capita. This does not mean that there is a clear better alternative than others, but providing an explicit rationale for the choice of one scheme over the other is helpful and important to compare and build a body of evidence in this research area for Mexico. Among this type of analytically supported regional schemes, in this thesis I will follow the max-p regionalization scheme used by Sastre-Gutierrez and Rey (2010). As the authors themselves explain, this approach: [G]enerates regions by grouping states together to respect a contiguity constraint and to maximize intraregional homogeneity with regard to per capita GDP. Additionally, the approach uses a base threshold, which as applied here consists of a minimum number of states for a region. In our case, we based this threshold on the average number of states observed in the five previous regionalization schemes. Given these constraints, the max-p is distinct from other regionalization algorithms in that the number of regions to be formed is endogenously determined, rather than having to be specified a priori." (Sastre-Gutierrez and Rey, 2010: 290) ".... In this thesis, and following the regional scheme developed by Sastre-Gutierrez and Rey (2010), I define Southern Mexico as comprising the states of Chiapas, Guerrero, Hidalgo, Oaxaca, Puebla and Tlaxcala (see Figure 2). My selection is based mostly on the fact that I have a special interest in economic development and the max-p algorithm scheme prioritizes the homogeneity in this regard (per capita GDP) above other criteria. 9 Figure 2. Southern Mexico: Chiapas, Guerrero, Hidalgo, Michoacin, Oaxaca, Puebla and Tlaxcala .4 ~'K3 Source: By the author. In Table 1, 1 present a selection of some descriptive statistics of Southern Mexico. The states of Chiapas, Guerrero, Hidalgo, Michoacan, Oaxaca, Puebla and Tlaxcala, comprise 23% of the total population in Mexico but they accounted for only 12% of the total GDP in 2012. Among these states, Puebla is the most populous (5% of the total population) and is also the one that contributes the most to Mexico's GDP (3.3%). On the other hand, Tlaxcala is the state with the least percentage of the national population and GDP (i% and 0.6%, respectively). In Table 1, the number of municipios by state is also presented. Mexico has over 2,400 municipios, however, the diversity between these municipios, both in terms of size and resources, is very high. A reminder of this is that 23% of all municipios are contained in one single of the selected states, Oaxaca, and that, collectively, Southern Mexico, as defined here, accounts for 50% of all the municipios in the country. This speaks to the degree of political and social heterogeneity of these states, as municipios division lines reflect the history of economic, political and ethnic conflict. 10 Table 1. Population, GDP and Municipios, Southern Mexico South Chiapas Guerrero Hidalgo Michoacan Oaxaca Puebla Tlaxcala Olhers 2595130 4796580 3388768 2665018 4351037 3801962 5779829 1169936 86383408 23.1 4.3 3.0 2.4 3.9 3.4 5.1 1.0 76.9 16129U2. 233226 12.S 1.8 124 185496 1.4 204227 293195 200377 424804 71637 11299944.9 1.6 81 84 113 570 217 60 1213 2.3 1.6 3.3 0.6 87.5 118 50.X6 4.8 3.3 3.4 4.6 23.2 8.8 2.4 49.4 Source: National Institute of Statistics and Geography, Mexico Finally, I would like to emphasize once more that the selection of this definition of Southern Mexico is one among a number of possible choices, and literature on regional development and inequality in Mexico has followed other schemes. In fact, some of the most influential studies on regional convergence in Mexico, in particular Esquivel (1999 and 2007) and Davila, Kessel and Levy (2002), have used schemes that correspond more "intuitively" with what academic and policy circles would refer as Southern Mexico, that is, the states of Chiapas, Guerrero, and Oaxaca, but not necessarily Michoacdn, Hidalgo, Puebla and Tlaxcala, like I do in this thesis. Other relevant studies, like the Strategyfor the Southern States Developmentby the World Bank (2003) have also used a similar choice than the above authors. 1.2 Why Does Development in Southern Mexico Matters? This thesis has a primary interest in investigating the spatial structure of key economic and social variables in the municipios of Southern Mexico. By performing a simple inspection of the levels of economic underdevelopment and poverty rates of Mexico, it seems clear that there is a distinct pattern of underdevelopment in the South that clearly sets it apart from the rest of the country. For example, the maps of the percentage of population living under poverty conditions (that is, with insufficient income to access basic food and services) by municipio already suggest the formation of a cluster of underdevelopment in Southern Mexico (see Figures 3 and 4). As can be seen in Figures 3 and 4, most of the municipios with a poverty rate above 70% lie in the states of Chiapas, Guerrero and Oaxaca. Even though there are other belts of high poverty rates as one goes north, in the states of Chihuahua, Durango, and Sinaloa, the distinctiveness of the South is clear. 11 Figure 3. Percentage of Population Living in Poverty Conditions by Municipio N % of Population In Poverty 0~ L~] EJ m m 0.0000 - 0.5158 0.5158 - 0.6413 0.6413 - 0.7560 0.7560 - 0.8507 0.8507 - 0.9735 'I 4' -I Source: National Council for Social Development Policy, 12 2010, Mexico. Figure 4. Percentage of Populating Living in Poverty Conditions in Southern Mexico by Municipio, 2010 % of Population in Poverty [2] 0.000 - 0.510 [2 0.510 - 0.640 0.640 - 0.750 m m 0.750 - 0.850 0.850 - 0.974 Source: National Council for Social Development Policy, 2010, Mexico. In this thesis, I give primary interest to this type of economic variables; however, it is relevant to note that my research is conducted in the context of increased governance and security concerns in the region (Horton, 2014 and Illades, 2014). Since 2012, it has become increasingly clear that the pervasive levels of poverty and social inequality in states such as Chiapas, Guerrero, Hidalgo, Michoacan, Oaxaca, Puebla and Tlaxcala, can have unsettling consequences for the governance and political order of the region (Illades, 2014). These concerns echo research on how high levels of social inequality and polarization can be a triggering factor of conflict and violence (Esteban, 2011). As Mexico has become an increasingly globalized and diversified economy, the inequality between individuals and regions has tended to increase (Messmacher, 2000). In 1994, after the NAFTA came into effect in January 1, the Zapatista Army of National Liberation public appearance in Chiapas brought attention, once more, to the high levels of poverty and inequality in Southern Mexico. This prompted, among other things, a significant increase in the level of public resources invested in the state. For example, only between 1994 and 2000, there was a real annual mean increase of 5.4% in the resources the federal government destined to public services provision in Chiapas, almost 5 times the 13 increase of the total resources for all the country during the same period (Davila, Kessel y Levy, 2002: 206). However, these resources came, once more, in the form of public services, and not long-term infrastructure or human capital investments. In the last decade, and especially given the increase of violence and crime trends linked to drug cartels in the Southern region, concerns regarding the links between economic inequality and social polarization with governance and public order have resurfaced. Even though during in 2008-2012, the early phase of Mexico's security crisis, much of the crime and violence tended to concentrate in states in the North, like Chihuahua or Tamaulipas, during the past couple of years new non-state armed actors, like vigilante groups and guerillas, have started to emerge in Southern Mexico. These groups are quite distinct from drug cartels, which resemble quasi-professional and mercenary organizations, and pose a unique challenge to local and federal authorities, as they are composed by civilians and are deeply rooted in local conceptions of order and authority (Horton, 2014). As can be seen in Figure 5, the states of Southern Mexico have relatively low rates of homicide (per 1o,ooo inhabitants) as compared to states like Chihuahua, Coahuila and Sinaloa. However, states like Guerrero and Michoacan are rapidly becoming the focus of concern both for the federal and state governments, and civil society and human rights organizations (Horton, 2014). Figure 5. Homicides per 10,000 inhabitants by State, 2012 Hoide Rate S2.0 - 7.8 J7.8 -11.8 m11.8 -18.4 m18.4 -36.8 -36.8 -71.9 4~ Source: National Institute of Statistics and Geography, Mexico. 14 After the surge of vigilante and militia groups in the Southern state of Michoacatn during 2013-2014, the federal government announced that I would invest $3,400 million dollars during 2014 in economic and social development initiatives in the states (BBC, 2014). In the case of Guerrero, where drug cartel violence and crime has also increased since 2012, the federal government also announced the "Plan Nuevo Guerrero" which consisted mostly on fiscal incentives, temporary employment programs and a fund for firms in risk of going bankrupt (Sanchez, 2014). Economists and analysts in Mexico, however, regarded these plans as generally insufficient and did not expect any major changes as a consequence of their implementation (Montalvo, 2014). This renewed concern in the governance and public order situation in Southern Mexico and its links to economic growth relevant has, then, prompted interest in considering regional development initiatives as an alternative. However, this is true not only because regional development might affect violence and crime trends in Southern states, but also because it might initiate a feedback loop that could potentially worsen the situation, as research has suggested that the poor quality of political and social institutions could further hampers the prospects of economic growth (Esquivel, 2002: 2). My thesis is inscribed in this landscape of awakening social and political conflict in Southern Mexico and it aims to provide elementary evidence for a more nuanced understanding and policy making for the region. In particular, by providing evidence that suggests patterns of spatial autocorrelation and heterogeneity in Southern Mexico I aim to contribute to a more nuanced and informed understanding of the region. Further, this first steps might lay the ground for more advanced spatial analytics considering the potential network and feedback effects that key variables like income per capita, poverty rates and human capital endowments can have between neighboring states and municipios in Mexico. This in itself should be part of a theoretical research agenda linked to endogenous growth theory and institutional economics to understand the process of regional development. 1.3 Structure In the remainder of this thesis, I argue first the importance of locality and space for development policy, especially in the context of Southern Mexico. Subsequently, I review the history and evolution of regional inequality in Mexico, with emphasis, once more, on Southern Mexico, and finally I present a basic analysis of spatial autocorrelation at the state and municipio level in this region. In particular, in Chapter 2, 1 discuss the importance of thinking spatially about regional development. This section aims to link the elementary spatial autocorrelation analysis at the municipio level I present in Chapter 4 to the increased awareness of place and locality in the development literature (see, especially, Barca, McCann, and RodriguezPose, 2011). In particular, I emphasize how spatial analysis of regional development can help unearth potential network or feedback effects between neighboring municipios, and it can help to draw a more nuanced picture of the diversity and patterns within large 15 spatial units, like regions or states. In this chapter, I also introduce the main methodological tools I utilize: measures of regional income per capita convergence and inequality, as well as spatial (both global and local) autocorrelation measures to inspect municipio data for spatial structure. In Chapter 3, I review the literature and historical evolution of regional (income per capita) economic convergence and inequality in Mexico at the state level for the last seven decades (from 1940 to 2010). In particular, I review the evidence in the literature regarding the state of "permanent underdevelopment" for Southern Mexico as well as some of its possible causes, including lack of infrastructure, human capital and investment (Esquivel, 1999; Esquivel et al., 2002 and Gonzalez-Rivas, 2010). Although, admittedly, convergence as measured by income per capita does not draw the entire picture of economic and social welfare across regions in Mexico, this has been a common measure in the literature both in Mexico and in other countries, and it will allow me to give a general look to the evolution of this process in the South. In particular, income inequalities seem to be posed to be better addressed by regional development policies, given that most of the dispersion in this regard is given by differences between Northern and Center States, an states in the South (and not by within state inequality). On the contrary, social welfare inequalities, including access to post-elementary education or infant mortality rate, might be better addresses by a more tailored approach (Esquivel, 2002). In the final chapter, I present my conclusions, the scope and limitations of my research, and some future research areas related to my analysis. The evidence I present in this thesis does not allow for an in depth discussion of this, but I believe it is relevant in order to outline a future research and policy agenda. I would like to focus on the potential tradeoff between economic growth for the whole country and a regional development policy for the South. However, I also would like to emphasize the importance of thinking on a more tailored approach to regional development in Southern Mexico: particularly, by recognizing the fact that within the states of Southern Mexico, municipios face different challenge regarding economic development and/or social welfare. Previous research has found that there are areas of both economic and social development that might be better addressed by these two different types of policy focus (Esquivel, 2007). 16 2. The Case for Place-Based Social and Economic Development Why does place matter in regional development policy in Southern Mexico? How does inspecting the spatial structure of economic and social indicators at the state and municipio level contribute to a better understanding and (potentially) implementation of development policies? The emphasis on space and locality in this thesis echoes the increasing awareness both in the academic and policy development community. During the last decade, the importance of bringing space and localities back in development policy has become increasingly recognized by international organizations like the World Bank and the Organization for Economic Cooperation and Development, as exemplified by the publications of such reports as Reshaping Economic Geography(2009) and How Regions Grow (2009), respectively. As discussed by Barca, McCann and Rodriguez-Pose (2011), this has been largely as a result of the perception that conventional regional development policy, as implemented in the European Union and other regions, focused only in large infrastructure projects and state aid, has had, to say the least, mixed results. On the contrary, the current emphasis on locality urges to rethink how development intervention should focus not only on this kind of interventions, but instead emphasize the role of social inclusion and the political and institutional diversity within regions (Barca, McCann, and Rodriguez-Pose, 2011: 1). This emphasis is increasingly relevant in a context in which the increased integration of economies in the international sphere has also resulted in increased inequalities within national economies (Esteban, 2000) In this context, bringing space back into regional development policy for Southern Mexico, especially at the municipio level, matters for two main reasons. On the one hand, inspecting for spatial structure in Mexico's municipios might unveil evidence that indicates interactive or feedback effects between neighboring spatial units (Gleditsch and Ward, 2007: 7, and Vaya, 2004). In this thesis context, this could be the case between neighboring states or municipios. This is especially relevant in the context of economic development, as there are both theoretical and empirical grounds to investigate the presence of externality effects of certain productive factors, like human capital or infrastructure (Vaya et al., 2004) Although later on in this thesis I do not present a fully specified regression model to better understand the effect that, for example, the GDP per capita or poverty rates between neighboring municipios may have between each other, the spatial autocorrelation measures presented are a first (albeit elementary) step in this direction. A second reason why spatial analysis matters in the context of regional development is that a more detailed understanding of the spatial patterns in the municipios of Southern Mexico can help produce a more nuanced understanding of the types of challenges some 17 municipios are facing, but also how are they patterned (or not) across space, and the development policies that might prove more effective. This latter concern, as I have pointed out, relates to the literature on regional development emphasizing the role of local production factors, institutions and actors, especially (but not exclusively) in the context of the European Union cohesion and integration process (Barca, McCann, Rodriguez-Pose, 2011). This parallel is even more relevant if one thinks about regional development in Mexico in the context of its integration to the North American region after the signing of NAFTA (Messmacher, 2000). Bringing space back in for regional development policy can also help to distinguish between localities facing conditions of "relative underdevelopment" and those facing "permanent underdevelopment" (Farole, Rodriguez-Pose, and Storper, 2011: 10971098). That is, municipios or localities that struggle to initiate or sustain economic and social development, but do not face critical shortages of infrastructure or human capital, and municipios that have demonstrated to be chronically incapable of making productive use of their own resources and are in clear need of critical investments to prevent critical economic or social collapse. However, bringing space back into regional development policy implies a specific methodological challenge in terms of looking in more detail the localities that compose economic regions. As Farole, Rodriguez-Pose and Storper put it (2011: 1098): "On one hand, it would require a more precise definition of the probabilities that a place will, or will not, move up the technology frontier/product space and over what type of time horizon. On the other hand, it would require identifying whether active intervention could improve these probabilities, and precisely how it would improve capacities to move up the technological frontier/product space, and which such capacities are amenable to improvement with intervention." The problem with thinking Southern Mexico as a homogenous region, at the policy level, is twofold. On the one hand, it can dismiss the fact that certain economic and social characteristics of municipios may be part of a spatial network or feedback structure. By ignoring this, potential positive or negative feedback loops could be ignored and lead to suboptimal scenarios. In particular, developing a spatial regression model to understand the effect of GDP per capita, poverty rates or human capital endowment between neighboring municipios, could shed light on how to better allocate development resources in order to create synergies among networked spatial units, and take advantage of spillovers or mitigate negative externalities (Gleditsch and Ward, 2007: 7) For example, when inspecting Figure 6, showing the Human Development Index 2010 (H DI) by municipio, one can see that there is a cluster of deep underdevelopment in the 18 inner territory of Guerrero, Oaxaca and Puebla (in light blue) where the HDI is below (approximately) 0.5, whereas certain municipios (in dark blue) in the coasts of Guerrero, in the belt composed by Hidalgo and Puebla surrounding the industrial clusters of Mexico City and State of Mexico, and the border of Chiapas and Oaxaca, have relatively high HDI scores (above 0.7). Although this visual inspection does not confirm the existence of spatial clusters, it illustrates the fact that even within Southern Mexico there is a distinct system of municipios with challenges of their own and that it makes sense to think beyond a "one size fits all" development policy for the region. Figure 6. Human Development Index in Southern Mexico, 2010 Human Development Index, 2010 [1] 0.000 - 0.580 [-] 0.580 - 0.630 m 0.630 - 0.660 0.660 - 0.710 0.710 - 0.850 m Source: United Nations Development Programme, Human Development Index, 2010. On the other hand, ignoring the spatial dimension of development, could also lead to miss the opportunity to identify the appropriate types of policies for municipios or localities at different steps of the "development ladder" or with specific economic or social development needs. Even if conventional measures of economic development, like 19 GDP per capita, may group municipios as part of the same "underdeveloped" category, the particular causes may be more heterogeneous than expected. For example, when looking at Figures 7 and 8, showing the percentage of population without access to health services, and of households without access to water (both by municipio), it is not clear that the same municipios are lacking in both measures. In the case of the population without access to health, Michoacain and Guerrero (in the upper left) appear to be doing relatively well as compared to Oaxaca (in the lower left corner). On the other hand, when considering households without access to water, Michoacan has clearly a significant number of municipios that are doing fairly bad (above 50% of the households lack access to water). This is just an illustration of the fact that whereas some municipios might be in need of investment in basic social services, other might need resources focused on urban or economic infrastructure. Figure 7. Percentage of Population without access to Health Services by Municipio, 2010 % of Population without Health Services [ 0.00 - 73.36 E] 73.36 - 84.61 84.61 - 91.45 91.45 - 95.83 95.83 - 100.00 Source: National Council for Social Development Policy, 20 2010, Mexico. Figure 8. Percentage of Households without Access to Water by Municipio, 2010 % of Households without Water Access E~] 0 - 11 11-21 21-33 33 - 52 f52 - 100 [ - K -~ Li Source: National Council for Social Development Policy, 2010, Mexico. The distinction above is not only relevant at the theoretical level: that is, the distinction between municipios or localities with real medium term potential to move up the technology or product space and those that are in lack of it. It can also have a significant bearing in the way regional policy development is designed and implemented. Economists, policy professionals and development scholars are aware of this scenario (Esquivel, 2007), but explicit spatial analysis and modeling has not yet been fully integrated into development proposals. For example, as part of a preliminary proposal for a regional development policy in the South, Esquivel (2007) envisioned a Regional Fund for Southeastern Mexico that would include two distinct components: a regional productive infrastructure fund, presumably for municipios or localities with reasonable potential to move up the technology product space, and a social equity fund that would address the needs of municipios facing 21 conditions similar to what Farole, Rodriguez-Pose and Storper (2011) refer to "permanent underdevelopment". That is, municipios that, given their critical insufficiencies in basic levels of economic and social wealth, lacks the capacity to make productive use of their own resource endowment. In fact, Esquivel (2007) developed an extremely insightful global inequality index, comprising measures for health, housing, education and income, at the municipio level, in order to distinguish different levels of needs. In Esquivel's (2007) own words: "The Regional Productive Infrastructure Fund should allocate its resources to invest, finance and co-finance productive infrastructure in the region (mainly in the form of airport, ports, communications and other infrastructure destined for tourism). All of the funded projects should be fundamentally regional. [...] On the other hand, the Social Equity Fund should have as its main objective the investment in areas related to education, health and housing with the goal of reducing inequality within this region." (Esquivel, 2007: 35) However, the issue is not only distinguishing the specific needs of these two broad types of municipios in South Mexico. It is also a matter of diagnosing the capacity of local authorities to actively engage in the regional development process. This renewed emphasis on the local is of critical importance to understand the role that spatial analysis can and should play in development policy. Thinking about space in regional development policy means thinking not only about its geographic or physical characteristics, but also about the institutional fabric in it, their local context. As argued by Barca, McCann and Rodriguez-Pose (2011: 6), there is an increased awareness of the fact that conventional regional development policy, solely focused on large infrastructure effects, can have either limited or unintended consequences. The importance of producing development policies with increased awareness of local is further supported given the fact that policies intended to alleviate regional disparities sometimes had resulted in the opposite effect: "An increasing body of research has tended to highlight that, even if the aggregate impact of infrastructure policies has sometimes been positive, they have often led to greater economic agglomeration, regional polarization and to an increasing economic marginalization of many peripheral regions where significant infrastructure investment hast taken place..." (Barca, McCann and Rodriguez-Pose, 2011: 6-7) Although this thesis provides only preliminary evidence to think more carefully about the role of space and locality in regional development, it aligns with a broader agenda aimed to refocus on the importance of institutions and localities. In particular, it aims to ask deeper questions about how regional development can look like for a country as 22 diverse as Mexico were surely not every region and urban growth engine can replicate Mexico's City or Monterrey's development experience (Elizondo and Krugman, 1992). This, not only because it seems highly unlikely, but because its potential pernicious effects regarding the efficient allocation of population flows and development funds: "Obviously this also raises the question of whether the existence of megacities at the top of a national urban hierarchy provides any distinct and systematic advantages for the nation as a whole over a national urban hierarchy with several intermediate or smallersized cities at the top." (Barca, McCann and Rodriguez-Pose, 2011: 12, and Elizondo and Krugman, 1992). Furthermore, as the Mexican economy becomes more sophisticated and diversified, the relevance of locality increases. In terms of economic development, scholars have already noted the relevance not only of traditional factors, like infrastructure or the size of the internal market, but also new factors, like mobile and Internet penetration (Esquivel, 2002). Having a more nuanced understanding of the local characteristics of municpios in Mexico will also be of increased relevance in the near future. Investigating the spatial structure among is one step ahead in this research and policy agenda. Finally, it is relevant to note that research suggests that development policy, for example poverty alleviation policies, show strong advantages in terms both of their effectiveness and administrative costs. In particular, Barker and Gorsh (1994) have argue for the case of Mexico and other countries that poverty alleviation is more effective when geographical targeting, that is, the allocation of resources according to place (for example, a municipio), is at the core of policy efforts. This, as compared to schemes involving individual or no targeting at all: "A comprehensive comparison of targeted programs in Latin America indicates that in actual program practice, geographic targeting works as well as other targeting mechanisms (see Gorsh, 1994). The median share of benefits going to the poorest 40 percent of households is 72 percent for geographic targeting, 71 percent for those using self-targeting mechanisms, and 73 percent for programs with individual assessment mechanisms (means test, nutritional status or risk)." (Baker and Grosh, 1994: 1) 2.1 Methodology In Chapter 3 and 4 1 present data referred to regional income per capita convergence and inequality, and spatial dependence, respectively. In Chapter 3, the regional income convergence and inequality measures are adaptations of standard measures for the study of non-spatial convergence and inequality metrics. In particular, I use measures of beta convergence, and a regional Theil Index to measure income inequality. These 23 measures are descriptive and only update for the year 2010 previous inequality research done originally by Sastre-Gutierrez and Rey (2010) as part of their pioneering work in spatial econometrics for Mexico for the period prior to 2000. In the past, research done on regional inequality in Mexico has been conducted using economic convergence measures that account for the disparities across the 32 states in terms of income per capita. In particular, measures of beta convergence have been used as way to investigate the way poor and rich states have either converged towards a common level of income or not. During the past couple of decades, economists have developed research trying to understand the existence of trends towards economic convergence both within nations and across nations (Barro and Sala-i-Martin, 1991 and Krugman, 1991). This strand of literature has tried to understand where processes of economic convergence have occurred, at what pace and under what circumstances. Indeed, some authors, like Barro and Sala-i-Martin (1991) have found that processes of economic convergence have occurred for the states of the United States and across countries in Western Europe, even though in most cases this has happened at slow rates. In the economic literature, economic convergence is generally understood to have one of three meanings: absolute convergence, relative convergence and club convergence (Esquivel, 1999: 3-4). In the case of absolute convergence, the income per capita of an economy converges with others regardless of their initial conditions. In the case of conditional convergence, the income per capita of an economy converges with that of other economies that share structural conditions (for example, their level of technology development) regardless their initial conditions. Finally, in the case of club convergence, the income per capita of an economy converges with that of others that share the same initial conditions. In this section of the thesis, I aim to present an elementary analysis of how states in Southern Mexico have developed in terms of economic convergence. Later on, in Chapter 4, in order to investigate evidence of spatial structure at the state and especially at the municipio level in Southern Mexico, I make use of standard global and local measures of spatial autocorrelation. It is important to note that I did not develop an actual spatial regression model to develop the evidence of spatial autocorrelation that I found at the municipio level. This means that the global and local autocorrelation measures I present are only a preliminary exercise in the investigation of possible network or externality effects across key economic and social development variables between neighboring municipios in Southern Mexico. As I mentioned, I use both global and local spatial association measures. But, what is the difference between these two measures? Whereas Global Moran's I spatial autocorrelation measures can identify the overall clustering across spatial units, Local 24 Indicators of Spatial Association or LISA (Anselin, 1995) measures: "... may be interpreted as indicators of local pockets of nonstationarity, or hot spots... and they may be used to asses the influence of individual locations on the magnitude of the global statistic to identify 'outliers'" (Anselin, 1995: 93). In other words, while the global Moran's I can serve the purpose of identifying the process of clustering, this measures does not help in identifying where the clusters (if any) are. On the contrary, LISA measures do allow us to identify clusters of positive spatial autocorrelation (for example, municipios with high income per capita next to other high income per capita municipios) as well as those with negative spatial autocorrelation (for example, municipios with low poverty levels next to municipios of high poverty levels). The study of these spatial metrics in the context of economic and social development in Southern Mexico is especially relevant when it is made at the municipio level, as it allows to gain scholarly and substantive insight on how the challenges facing municipios in the states of Southern Mexico might have a spatial dimension to them. In particular, I implement the aforementioned spatial association measures for economic and social welfare data obtained from sources like the National Institute of Statistics and Geography, the United Nations Development Program Office in Mexico, and the State Competitiveness Index developed by the think tank Instituto Mexico para la Competitividad. This will provide me with a first glance at the regional and state level perspective of the relevance of space in thinking and implementing economic and social development of Southern Mexico. 25 26 3. An Overview of Regional Development and Inequality in Mexico Since the 1940s, states in Southern Mexico, Chiapas, Guerrero, Hidalgo, Michoacan, Puebla and Tlaxcala, have consistently ranked at the bottom of income per capita measures, especially when compared to the Center and North of the country. This is also true for other measures of social welfare, like its poverty, inequality and access to public services levels (Esquivel, 1999). Whereas states in the middle and upper brackets have moved up and down the ladder of income per capita (and other key social welfare measures), the states of Southern Mexico have remained at the bottom for the last seven decades (Esquivel, 1999). Although it has become common to think about Southern Mexico as a land of underdevelopment and inequality, there is no clear explanation of why this is so or its relation to economic development's spatial structure. After closer inspection, it is not trivial that the same states in Mexico have remained at the bottom of the economic welfare ladder for decades, and that they seem to be next to each other, forming a geographical cluster. In the past decades, a number of regional development policies for the region had been considered in order for them to "catch-up" with the more industrialized North and Center of Mexico. In fact, as shown in Figure 9, the federal resources destined to states in the South, like Chiapas and Guerrero, known for the presence of guerrillas and other non-state armed actors, have increased after 1994 (Diaz-Cayeros, 1995 and Davila, Kessel, Levy, 2002).1 However, most analysis and policy proposals on this regional development overlook the geographic or spatial component of underdevelopment in Southern Mexico, and, in the same vain, assume that Southern Mexico is a fairly cohesive region within itself. This has very significant implications, as I have discussed before, on the way policies are designed and its potential effectiveness. 1States in Mexico receive both "partipaciones federales" and "aportaciones federales". By federal transferes I refer only to participaciones federales, as Gonzalez-Rivas, 2009 and Diaz-Cayeros, 1995, as they tend to be more substantial than participaciones federal. Additionally, the participaciones federal can be used however the state governments decide to, as opposed to the aportaciones federal, where the resources are allocated directly by the federal government and only given to the state government to administer. 27 Figure 9. Federal Transfers for Southern Mexico Federal Transfers in Million pesos (1995 prices) 100000 - 0000 12 - Other States South 60000 CU 40000 U- 20000 Il I 1990 I - -I-1995 2000 I 2010 Source: National Institute of Statistics and Geography, Mexico. Despite the high level of regional inequalities, the structure of federal transfer to state governments is highly regressive (see Figure i), as indicated by the positive slope relating state GDP per capita (in the Y axis) and federal transfers per capita (in the X axis). In other words, states with higher levels of GDP per capita receive more federal transfers per capita than states with lower levels of GDP per capita. For example, Mexico and Nuevo Leon, two of the most wealthy and industrialized states in Mexico receive proportionally more (relative to their GDP per capita) than states with lower income per capita. This is particularly true for all the states that compose Southern Mexico, as can be seen in Figure 10, including Chiapas, Guerrero, Hidalgo, Michoacan, Oaxaca, Puebla and Tlaxcala. Previous research on regional development policy in Mexico confirms this trend in federal transfers at least since the 1990s (Diaz-Cayeros, 1995). 28 Figure 10. Federal Transfer to State Governments (in Mexican pesos), 2012 Federal Transfers to Slate Govenments. 2012 Federal Transfs per capta Source: National Institute of Statistics and Geography, Mexico. As I mentioned in the Introduction, the issue of regional inequality has become increasingly relevant given the increasing levels of crime, violence and the emergence a variety of non-state armed actors (Davis, 2009), like the guerrilla groups in Chiapas (including the Ejercito de Zapatista de Liberacion Nacional) and Guerrero, or the vigilante and organized crime in Michoacin's Tierra Caliente (Asfura-Helm and Espach, 2013). This situation is well acknowledged not only in policy and media circles, but also in academic research on economic geography and social conflict in Mexico (Rey and Sastre Gutierrez, 2010: 278). All these groups vary significantly in origins, organizational structure and purposes, but it is believed that the chronic or permanent underdevelopment of Southern Mexico has served as a breeding ground to the lack of governance and unrest that it is being experience now. 29 3.1 A Review of Research on Regional Income Per Capita Convergence and Inequality in Mexico Even though during the first half of the 2 0 th century the federal government had an active role shaping industrial clusters in Central and North Mexico as part of its import substitution growth strategy, during the second half of the century no major regional development initiative took place (Friedman, Gardels, Pennink, 2009). The economic crises in 1982 and 1994 prompted a number of macroeconomic disciplinary measures that, among other things, resulted in a reduced level of public investment, especially regarding regional development (Ros Bosch, 2014). However, after years of lagging economic growth rates, increased attention has been paid to the possibility of regional development plans that would revert historical economic inequalities among states and regions in the country (Casar and Ros, 2004). In particular, it has been argued that the resources that the federal government allocates for poverty alleviation and public services will not solve the root of the problem: the lagging productivity and GDP per capita growth of Southern Mexico. On the contrary, a "New Deal with the South" has been proposed as an alternative in order to revert historical inequalities in states like Guerrero, Michoacan, Oaxaca and Chiapas (Casar y Ros, 2004). In the case of literature dealing with regional convergence and inequality in Mexico, it is worth noting that there has been significant research made on the evolution of these issues after the implementation of the NAFTA in 1994. However, as I detail below, most of the literature has used an economic convergence framework which does not necessarily include spatial effects that might be derived as a consequence of patterns of the spatial clustering and heterogeneity of economic growth and productivity in the states of Mexico. In what follows, I present a summary of the research done of regional inequality in Mexico with a special focus on understanding this process for states in Southern Mexico. In doing so, I present measures of beta economic convergence that had been developed previously by Esquivel (1999) in its study of economic convergence across states in 1940-1995, and a regional inequality Theil Index developed by Sastre-Gutierrez and Rey (2010). These measures have been explained in more detail previously. In Figure 11, the state beta absolute convergence measures for the period 1940 to 1995 are presented. In the horizontal axis, we have the log of the state GDP per capita in 1940, and in the vertical axis we have the rate of growth of the state GDP for 1940 to 1995. I obtained the data presented from Esquivel (1999) and is based on a first of its kind and meticulously reconstructed series by the author for this period. 30 As can be observed, the linear fit of the data has a negative slope. In other words, states with a lower state GDP per capita in 1940, like Oaxaca, Guerrero and Tabasco, experienced a higher growth rate for the period 1940 to 1995 than richer states, like the Federal District (Mexico City), Durango or Queretaro. This is, in fact, is what absolute beta convergence means: the poorer states or regions will tend to grow more rapidly than the more well-off states or regions and eventually converge to a common income per capita level. However, as Esquivel (1999) has demonstrated, this eventual catching up process did not occur in the long run and in fact seems to have stopped by the 1960s. Figure 11. State Beta Absolute Convergence, 1940- State Beta Absolute Convergence, 1940-95 16 1995 Um o0 Source: Esquivel (1999) and Sastre-Gutierrez and Rey (2010) In Figure 12, the state beta absolute convergence for the period 1980 to 2010 is presented. The regression line for the data has a negative slope, as in the previous period of 1940 to 1995. This suggests that the process of economic convergence (state 31 - ..... .. .......... ........ .. .. ....... - GDP income per capita) continued during this period, although at much lower state GDP per capita growth measures, as can be seen by inspecting the vertical axis scale and compare it to the one corresponding to the period 1940 to 1995. It is important to note that the data I use here has not been corrected in order to account for the extraordinary oil financial resources for the state of Campeche. This explains why in both graphs Campeche appears as an outlier. Figure 12. State Beta Absolute Convergence, 1980-2010 Stea Beta About Convergence, 1NO-2010 Af Uof urn GOPI Mq ermIWl Source: Esquivel (1999) and Sastre-Gutierrez and Rey (2010), and Institute of Statistics and Geography, Meixco. Interestingly, even as the process of economic convergence seems to have stagnated since at least the 198os, the federal arrangement in Mexico does not seem to account for this, as previous research (Gonzalez-Riva, 2009 and Diaz-Cayeros, 1995) has shown the highly regressive character of federal transfers to Mexico to states. The only exception to this trends seems to bee the occasions in which social unrest and political governance challenges have incentivized the investment in states like Chiapas, where guerrilla groups emerged in protest, among other issues, to the signing of the 32 NAFTA in 1994 and the profound inequalities of the state. In fact, as a result of the uprising of guerrilla groups in Chiapas and other non-state armed actors in Southern Mexico, the federal transfers to states in the region increased, as reported by Davila, Kessel and Levy (2002: 206). However, the measures of income per capita convergence present only one side of the story. It is also important to understand how the inequality of income per capita has evolved for the states in Mexico. The Theil Index is a measure overall inequality (in this case, state GDP per capita) that can be used to study regional inequality. More specifically, the Theil index (see Figure 13) is a weighted average of inequality within subgroups, plus the inequality among the subgroups, and is given by the following equations (Rey and Saste-Gutierrez, 2010: 284-285): Figure 13. Regional Theil Index stlog(ns ) T= i=1 = r /' Lr Y Source: Sastre-Gutierrez and Rey (2010) In these equations, Tt is the Theil index at year t, n is the number of regions (in this case, n = 32 states), and yi is the per capita income in region i in period t (in 1995 Mexican pesos). The regional inequality Theil index ranges from the interval [o, log(n)] where o means perfect equality and log(n) corresponds to income concentration in one single region or state (perfect inequality). As mentioned, one of the advantages of the Theil index is its decomposability in order to inspect inequality within regions and across regions. Figure 14 shows the regional inequality Theil Index as calculated by (Esquivel, 1999). This historic series from 1940 to 1995 was carefully reconstructed by Esquivel (1999) and has been used in further research on regional convergence in Mexico with updated data to 2000 (Gonzalez-Rivas, 2010). As can be seen in the plot, the Theil index dropped sharply from 1940 to 1995, as originally shown by Esquivel (1999). As already noted, further research by Esquivel (1999) has demonstrated that this overall process of regional convergence was more accelerated from 1940 to 1960, while stagnating in the following decade. 33 Figure 14. Regional Inequality Theil Index, 1940-1995 (1995 Mexican Pesos) 0.22 0.20 0.18 0.16 0.14 0.12 0.10 0.08 1940 1950 1960 1970 1980 1990 2000 Source: Esquivel (1999) In Figure 14 and Figure 15, the Theil indexes for the periods from 1940 to 2010 (for this thesis, I only updated the data for 2005 and 2010) are shown in the blue line. The green line and the red dotted line represent the between groups (i.e. regions or groups of states) and within group income per capita inequality.2 As can be seen in Figure 15, no major process of overall economic convergence (state GDP income per capita) occurred during the period 1980 to 2010. In a closer look at the index, it seems that there was a slight increase of overall inequality across Mexican states during the 1980s and 1990s, but this process seem to have reversed in the following years with no major fluctuations. In the case of between and within group inequality (the green and red dotted lines, respectively), the former was higher during the period of the 196os to the 1970s, which means that the difference in income per capita between regions in Mexico (i.e. South and North) was higher than the inequality within any given region. However, this changed in the subsequent decades and within region inequality was the driver of overall inequality. In other words, after the 1970s, it was not the inequality between regions, but rather between states that compose a region that prevailed. This is indeed an intriguing finding and merits further exploration. allocation of all 32 states in Mexico to different geographic regions to perform this calculations follows the max-p scheme presented by Sastre-Gutierrez and Rey (2010: 288). 2 The 34 Figure 15. Regional Inequality Theil Index, 1940-2010 0 25 . 1 Thiel Index, 1940-2010 Regional 1 1 1 2000 2010 0.20 0 15 0 10 1 0,05 0.00 19'40 1950 1960 990 1980 1970 Source: Esquivel (1999) Figure 16. Regional Inequality Theil Index, 1980-2005 Regional Thiel Index, 1980-2010 010 0 08 0 06 0041 002 1980 1985 1990 1995 2000 2005 1 2010 Source: Esquivel (1999), National Institute of Statistics and Geography, Mexico 35 After the period of economic liberalization during the 198os and the 199os, there was an increased interest in investigating what would be the consequence of this process (in particular, after the signing of the NAFTA in 1994) in terms of the economic convergence of states in Mexico. This was particularly relevant not only because states in Mexico diverged in their access to North America's market, both in terms of distance and infrastructure, but also because the different human capital and innovation factors varied significantly across regions in Mexico. Previous research on regional inequality in Mexico has found evidence of convergence in the income per capita across states during the second half of the 2 0 th century; specifically, during the period 1940 and 1995 (Esquivel, 1999). In particular, Esquivel (1999) has found that the rate of income per capita convergence during this period occurred at an average rate of 1.1% per year; that is, relatively low by the 2% international standard. As Esquivel (1999: 32) himself puts it: "This rate of absolute convergence is one of the lowest that has been estimated at the regional level within a country and it is only comparable to that of other country with known regional development issues, Italy." However, according to Esquivel (1999) this larger trend can be divided in two periods: from 1940 to 1960, and from 1960 onwards. Between 1940 and 1960, the process of absolute convergence was fast and was coupled by a substantial reduction of income per capita disparities between states. In the second phase that Esquivel (1999: 32) studies, from 1960 to 1995, he found no evidence of economic convergence across states. 3.2. Causes of Regional Inequality in Mexico and Policies to Combat It In the literature, there are various factors that have been advanced to explain the process of economic convergence across states in Mexico, but they all generally tend to point out to similar factors, like human capital or infrastructure, which states like Chiapas, Guerrero, Michoacan, Oaxaca, Puebla, despite their rich endowment of natural resources and cheap labor, tend to lack. In his study of economic convergence during 1940-1995, Esquivel (1999: 33) pointed out two additional possible factors for the slow economic convergence in Mexico: the low sensitivity of inter-state migration to income difference, and rise of inequality in the access to post-elementary education. According to Esquivel et al. (2002), labor productivity was also a key driver in the economic convergence process, as measured in terms of output per capita by state, during the period from 1960 to 1990. The divergence in terms of employment and sector participation rates also help explains the fact that the process of convergence across states stagnated during the following decades up to the 1980s. Again, according to Esquivel et al. (2002), labor productivity divergences across states in Mexico were a 36 critical factor in the regional divergence across states experienced during the 1990s. This is particular true given the increasing relevance of variables such as postelementary education and infrastructure in the determination of output per capita and labor productivity in this latter period (Esquivel et al., 2002: 1). It is important to note that the factors that seem to be the most relevant in explaining the process of economic (income per capita) convergence across states seem to be evolving as the Mexican economy is becoming increasingly more sophisticated. Whereas in the 1940s or 1960s factors such as elementary or secondary education, or highway infrastructure, might have been more relevant, research done for recent decades points out to other factors like mobile phone penetration and access to the Internet (Aroca, 2005). In both scenarios, however, states like Guerrero or Chiapas and other states in the South, seem to be lagging as compared to states in the Center and North of Mexico. According to Messmacher (2000) and Esquivel et al. (2002: 5) the opening of the Mexican economy during the 198os and 199os, and in particular the signing of NAFTA, appears to have favored certain economic sectors, like manufacturing and advanced industries, and the states and regions endowed with more advanced human capital and infrastructure to support these sectors. In fact, according to Esquivel et al. (2002: 5), it was found that after the signing of NAFTA, the sectors that outperformed the rest during the 1990s were manufacturing production and transport and communications. Not surprisingly, these are the sectors that tend to be located in Northern states and in the manufacturing belt in Mexico City and other states like Jalisco and Veracruz. As noted above, several factors have been advanced to explain the process of economic convergence across states in Mexico during the 1940s and 196os and, in the following decades, its relative stagnation. In particular, labor productivity and the industry mix of each state have been proposed as explanations for this (Esquivel, 1999, and Messmacher, 2000). In the former case, labor productivity, the theory would imply that the different endowments of access to education and other skills formation factors would have made the productivity in states from the South particularly low, regardless of the industry mix in these states. The latter explanation, related to the industry mix, would ascertain that even if there was a generally similar level of labor productivity across states, the particular industry mix of each state (for example, the percentage of advanced manufacturing or traditional agriculture) would be the main factor explaining why poorer states did not catch up richer states during the decades following the 196os. However, it is clear that even though these two factors are of critical importance, they interact with other regional economic characteristics in the overall economic process. For example, a World Bank report on the states of Chiapas, Guerrero and Oaxaca found that the reason behind the lagging economic growth of these economies was their extremely low productivity: "the value of goods and services produced per person in the South [in the three states mentioned above] is less than half as much as in the rest of 37 Mexico" (Hall and Humphrey, 2003: 3). In turn, this low productivity was linked to three main factors: the high cost of doing business associated to poor infrastructure and land tenure and violence, the disinvestment in the comparative advantages of the region and the focus on subsistence agriculture, and the inefficient government activity, meaning a regressive federal transfer scheme, biased pricing policies, inefficient use of public administration resources, poor education and an inefficient judicial system (Hall and Humphrey, 2003: 3). As Esquivel et al. (2002: 5) have pointed out, even in the case where the industry mix is a much more significant explanation of the process of economic convergence across states in Mexico, the question of the causes of this industrial specialization by state remains. How did it occurred that some states were able to develop advanced manufacturing industries and other not? More importantly, a policy question regarding how to take advantage of industrial specialization or to build the capacities to autonomously do so is also apparent. As Davila, Kessel and Levy (2002) have argued, it is those states lacking human capital, adequate infrastructure (specially connectivity infrastructure, like highways and ports) and with a higher proportion of public sector employment that seem to be lagging behind the more industrialized ones. These authors also have remarked the role that distance to large economic markets (like the one in Mexico City or in the United States) and the initial concentration of manufacturing and population density in the larger process of economic convergence. More prominently, according to Davila, Kessel and Levy, the effect of transportation costs and economies scales, through the agglomeration in large urban centers, seems to be a distinctively influential factor in the economic laggardness of states in the South. In any case, the study of regional inequality in Mexico, and in particular the low rate of convergence during the first of the 2 0 th century, and its stagnation during the last decades, have inspired several policy frameworks directed towards the regional development of the states in the South; in particular, Chiapas, Guerrero, Michoacan and Oaxaca. As noted in the Introduction, the urge to rethink the future of this region of Mexico has increased as it has becoming increasingly clear that the low public investment in the region and its high levels of economic and social inequality might be significant precursors of the violence and crime that states like Chiapas, Guerrero and Michoacan had undergone particularly since 1994, with the emergence of the Ejercito Zapatisa de Liberacion Nacional and other non-state armed groups. 38 4. The Spatial Structure of Economic and Social Development in Southern Mexico 4.1 Spatial Structure at the State Level As noted before, there has been for long a presumption of Southern Mexico as a geographical and economic cluster of underdevelopment. However, regional convergence and inequality literature has only recently incorporated spatial analysis tools in order to understand this phenomenon. In this section, I present basic spatial autocorrelation analysis to explore the spatial structure of GDP per capita by state in Mexico. After presenting this analysis and discussing it, I present data for spatial structure at the municipio level. Figure 17 presents box maps (Anselin, 2008) for the state GDP per capita (in 1995 pesos) for the decades of 1940, 1980, 2000, and 2010. Each of the following box map presents the state GDP per capita organized by percentiles in the following intervals: o25%, 25-50%, 50-75%, 75-100%. The color scale goes from deep red (for upper outliers, those states with a state GDP per capita 1.5 times higher than the interquartile range for any given year values) to deep blue (for lower outliers). As can be seen in each one of the box maps in Figure 17, the states of Chiapas, Guerrero, Oaxaca, and Puebla (all in varying shades of blue) have consistently remained under the did Chiapas was 2 5 th percentile of GDP per capita between 1940 and 2010. Only in 1980 part of the 2 5 th to 5 0 th percentile interval, and Hidalgo and Tlaxcala have moved up to the 2 5 th- 5 0 th percentile since the 1990 and 2000, respectively. In other words, for the last 70 years, the states in Southern Mexico have remained at the very bottom of the state GDP income per capita ladder. 39 Figure 17. State GDP per capita, 1940, 1980, 2000, 2010 (1995 Pesos) Hinge=1.5: 1980 0 Hinge=15:1940 Hger1i 5 180 Mrgen?.5 1940 Lower OUNKi (0) Lower 0~he (0) *25%54) E0 2.%-0%() .5% - 5%) *Uppwer l Wor. *751911 Hinge=1.5: 2000 MagW-1 5: 2000 50% () %.7..5%. Hinge=1.5:2010 0 - E3 = Hvge-.15 2010 Lower ouiner io) * 20148%(6 E so% (8) ) E Lower ovrer (0) '25%(8) 25% *LeW outle 1 - SUppae 1 Source: Equivel (1999) and National Institute of Statistics and Geography, Mexico The above is in itself is a relevant fact (that the same states have remained the poorest across decades), but, as I will now argue, another trait is no less relevant: the fact that these states have tended to be located in Southern Mexico and as part of what it appears to be a geographical cluster. But, in order to confirm this, I will need to provide analytical evidence. Interestingly enough, the case of the Center and North of Mexico, with states that include industrial clusters in the Federal District (Mexico City), the State of Mexico in the Center and Nuevo Leon in the North, seem to be part of a different spatial cluster with overall higher levels of state GDP per capita. As previous research on regional 40 convergence in Mexico has shown (Aroca, Bosch, and Maloney, 2003), spatial analysis does not seem to support the conclusion that there is something like a North or Center cluster for states in Mexico. However, as I argue, there is evidence that suggests some type of spatial structure in the Southern states in Mexico, both at the state level, and within the state (at the municipio level). As it was discussed in Chapter 2, spatial structure can be inspected either by using measures of global or local spatial autocorrelation of a variable with itself in space. The Global I Moran's for 1940 and 1980 are o.11, o.o8, and 0.14 for both 1990 and 2012 (the Global Moran's I scatterplots are in Appendix B). The permutation tests were conduced with 99 permutations and utilizing queen contiguity spatial weights matrix. These global spatial autocorrelation measures, which can vary from -1 to +1 (and were values close to o indicate low or no spatial autocorrelation), provide information of the overall clusteringfor a given dataset, although they do not indicate where the actual clusters are located. In the case of states in Mexico, these global spatial autocorrelation measures are rather low, which suggest low overall spatial auto regression. However, inspecting this in more detail, at the local level, will provide some evidence on the spatial structure of GDP per capita in the South. Figures 18-21 show Local Moran's I maps for the GDP per capita of Mexico by state for 1940 1980, 1990, and 2010.3 This analysis has been done with data from Esquivel (1999) and Garcia-Verdu (2005) and I updated for the year 2012. As it was also discussed in Chapter 2, local spatial autocorrelation measures provide information of where spatial clusters can be identified with some degree of statistically significance. In particular, in Figures 18-21, only spatial clusters deemed statistically significant following a permutation test approach (and thus, with a pseudo p-value lower than o.05) are shown. The spatial clusters can show either positive autocorrelation (of GDP per capita with itself in space) in deep red, for states with high GDP per capita next to other "rich" states, or (in deep blue) for states with low GDP per capita next to other "poor" states. The maps can also show clusters of negative autocorrelation (in light red or blue, respectively): that is, "rich" states that are next to "poor" states or vice versa. After inspecting Figures 18-21, two things are relevant to notice. First, the states of Chiapas, Guerrero, Oaxaca, and Puebla are consistently spatially auto correlated for a period that spans from 1940 to 2010. These states pertain to a cluster of low GDP per capita for a half-century period. Before, the box map had already suggested this pattern, but on a purely visual basis. On the contrary, local autocorrelation measures allows to pointed out in Esquivel (1999) and Instituto Mexicano para la Competitividad (2014), the GDP per capita in this analysis (except for year 2012) has been corrected to account for oil production in the states of Campeche, Chiapas and Oaxaca. This is due to the fact that oil revenues do not necessarily go back to their state of origin and my overestimate the GDP per capita measures. 3As 41 confirm that this spatial pattern is statistically significant, having as a null hypothesis the random spatial structure of GDP per capita across states. It should also be noticed that the other states in Southern Mexico (Hidalgo, Michoacain, Tlaxcala, and, with some exceptions, Puebla) do not seem to be part of this spatial cluster, and that others, like Veracruz, do. This is interesting in itself and merits further research. A second thing to notice in Figures 17-20 is that there does not seem to be other spatial cluster that is consistent across this period: only the states of Baja California Sur and Jalisco are part of High-High GDP per capita ad High-Low clusters, but not for all years. This is important because in policy circles there is a presumption that the Center and North of Mexico, jointly or separately, are part of an industrial and high GDP per capita corridor. Figure 18. Local Moran's I of GDP per capita, 1940 # LISA Cluster Map: mappibepc4O12_base95_queen, 1_1940 (9... LISA Chister Map: imp_pbe Not Significant (25) High-High (1) Low-Low (6) Low-High (0) High-Low (0) Source: Esquivel (1999). 42 0 = Figure 19. Local Moran's I of GDP per capita, 1980 O - LISA Cluster Map: mappibepc4O12 base95_queen, 11980 (9... 4 LISA Chister Map: mappbe [] Not Significant (27) High-High (0) Low-Low (4) Low-High (0) Hig*-Low (1) I Source: Esquivel (1999). Figure 20. Local Moran's I of GDP per capita, 1990 LISA Cluster Map: mappibepc4O12_base95_queen, 1_1990 (9... - I LISA Chister Map: map_pbe Not Significant (26) High-High (0) Low-Low (4) Low-High (1) High-Low (1) 'V Source: Esquivel (1999). 43 Figure 21. Local Moran's I of GDP per capita, 2012 LISA Cluster Map: mappibepc4O12_base95_queen, L2010 (9... LSA Chuster Map: msppibX 3 Not Significant (27) High-Hi* (0) Low-Low (S) Low-HIgh (0) High-Low (0) Source: Indice de Competividad Estatal, Instituto Mexicano para la Competitivdad, 2014. It should be recognized that even though state GDP per capita is itself an illuminating measure of the overall economic situation of states in Southern Mexico it gives us only a partial vision of the more general economic and social welfare in the region. In the following section, I will present other social and economic variables at the municipio level, in order to understand the underlying spatial structure within the Southern Mexico region. 44 4.3 The Spatial Structure at the Municipio Level Figure 22. States in Mexico's South WW K~-' &.* L ' I rW I9I 2 riUATE-A CNOACAW r MI- Mexico's South: Chiapas, Guerrero, Hidalgo, Oaxaca, Puebla and Tlaxcala In this section, I take a look at economic and social development data at the municipio level. Two things should be noticed in this regard. The first, the spatial structure of key economic and social data at the municipio level in the Southern states. As stated in Chapter 2, this is a relevant theoretical and policy issue given that spatial autocorrelation analysis can provide elementary evidence for further research on network or feedback effects between municipios. For example, there might be a feedback loop between spatial clusters of high GDP per capita municipios, or between municipios with high levels of infrastructure or human capital. This knowledge can inform policy and investment decisions for the development of Southern Mexico. The second issue I discuss in this section is the degree of, on the one hand, certain characteristic underdevelopment features of the municipios in Southern Mexico. On the other hand, however, I emphasize some measures of heterogeneity among municipios within states of Southern Mexico. This is critical to understand and increase the awareness of the heterogeneity within states in Southern Mexico, and it is a first exploratory analysis of how to think about development in the context of economic and social diversity. As can be seen in Figure 23 (the same than Figure 3 in Chapter i), which shows the percentage of population living under poverty by municipio, there is a clear cluster of very poor municipios (with poverty rates above 70%) in Southern Mexico, especially for the states of Chiapas, MichoacAn and Oaxaca. At the same time, it is clear that the north seems much less homogenous, and it may even exhibit an opposite spatial clustering of 45 well-off municipios in the North, especially in the states of Coahuila, Chihuahua, Nuevo Leon and Sonora. Figure 23. Percentage of Population Living in Poverty Conditions by Municipio % of Poplation In Poverty 0.0000 - 0.5158 [ 0.5158 - 0.6413 0.6413 - 0.7560 0.7560 - 0.8507 0.8so7 - 0.9735 Source: National Council for Social Development Policy, 2010, Mexico. By looking at other data, this pattern of underdevelopment in Southern Mexico is also clear. For example, in Figures 24 and 25, 1 show the percentage of municipios from Southern Mexico by percentile of GDP per capita and the Human Development Index (for all the municipios in Mexico). As can be seen, there is a very consistent pattern: the municipios in Southern Mexico compose a large share (approximately 90% and 80%, respectively) of the lowest percentile for both GDP per capita and HDI. The share corresponding to municipios in Southern Mexico in both measures systematically decreases as the percentile increases. This another way of point out that the municipios in Southern Mexico tend to be overrepresented in the lowest levels of income per capita and HDI. This pattern, as well as the spatial clustering apparent in Figure 23, is some of 46 the reasons why there is a presumption that Southern Mexico constitutes an homogenous spatial cluster of underdevelopment. Figure 24. Percentage of Municipios from Southern Mexico by Income Per Capita Percentile Iz b 1 1 S 0 .4 9 $4 Source: United Nations Development Programme, Human Development Index, 2010. Figure 25. Percentage of Municipios from Southern Mexico by HDI Percentile Source: United Nations Development Programme, Human Development Index, 2010. 47 In fact, as I mentioned previously, I would like to emphasize that despite this apparent cohesiveness within the states of Chiapas, Guerrero, Hidalgo, MichoacAn, Oaxaca, Puebla and Tlaxcala, there is some degree of diversity within these states. In Table 2, 1 show the Theil Index (TI) calculated for each state. This is the same measure I have used previously in Chapter 3 to analyze the degree of inequality between regions in Mexico. In Table 2, the Theil Index presented measures the degree of inequality for each state between municipios, again from a scale of o to log(n), where n is the total number of municipios. Unfortunately, the Theil Index is not directly comparable between states, as it scale depends on the number of municipios in each state. Table 2. Theil Index GDP per capita by State, 2010 1 r- j 0.089 118 0.070 81 0.084 0.04 84 113 0.13 570 0.07 0.04 217 60 In Table 2, I show the Theil Index calculated for each state. This is the same measure I have used previously in Chapter 3 to analyze the degree of inequality between regions in Mexico. In Table 2, the Theil Index presented measures the degree of inequality for each state between municipios, again from a scale of o to log(n), where n is the total number of municipios. Unfortunately, the Theil Index is not comparable between states, as it scale depends on the number of municipios in each state. In Figures 26-27, 1 present mosaic plots for the distribution of the municipios in the states of Southern Mexico by GDP per capita percentile, and by level of HDI. This is a way to look at the distribution within each state in terms of the income per capita and their HDI. For example, by inspecting Figure 26, it is clear that Oaxaca has a very large proportion of its municipios in the lower 3 percentiles of GDP per capita. On the other hand, Tlaxcala has a more homogenously distribution of its municipios across the income per capita percentiles, as Hidalgo does. Chiapas, too, has a relatively more homogenous distribution of its municipios in this regard, as compared to Oaxaca. 48 Figure 26. Municipios in Southern Mexico by GDP per capita percentiles, 2010 Municiplos by Income Percentle Chiaps Own.rro -iag Th4ad~ Ow.X; -ahacn -E -i -I - II 0. -m - M - states Source: United Nations Development Programme, Human Development Index, 2010. In Figure 27, it is also clear that there is a stark contrast between Chiapas and Oaxaca, on the one hand, and Tlaxcala, in terms of the distribution of their municipios by HDI level. Both Chiapas and Oaxaca have a very significant number of their municipios in the Low HDI level, whereas most of Tlaxcala's municipios belong to the Very High level. This is yet another indication that it is not accurate to think about the municipios in Southern Mexico as part of an homogenous region, even if, as I show later, they may be spatially patterned. 49 Figure 27. Muncipios in Southern Mexico by HDI level, 2010 Municiplos by HDI Percentile Chaps Goofr=dr Hdfto IdCh a Omxce Pdfbs Tbxcem A 0~ I 5 Staes Source: United Nations Development Programme, Human Development Index, 2010. Finally, another way to look at the diversity within the municipios of Southern Mexico, I look at the coefficients of variation for key economic and social development variables. The coefficient of variation is simply the standard deviation of the municipios data within a state divided by its mean. This is a way to standardize the data in order to be able to compare it across states. In Figure 28, 1 show the coefficient of variation for the HDI by state. The states from Southern Mexico are signaled with a vertical red line. As can be seen, all of these states, except for Tlaxcala, have higher coefficients of variation as compared to the mean coefficient of variation for all states in Mexico (dashed blue line). Among the states in Southern Mexico, Oaxaca and Hidalgo have the highest coefficient of variation for their HDI. Substantively, this means that these are the states with the highest degree of within variation at the municipio level. 50 Figure 28. Coefficient of Variation of HDI by State, Coefficient of Vartion of HDI by State, 2010 I I I I I I I I I III I I I - I I I I I - 006 - 0-02 I I I I I I I I I I I I I I I I I I I I I I I L~ IIIIIII I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I III I I II I. I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I F- I~L I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I II I I I I I I I ITF I I I 0-04 I I I I I I I I I I I I I I I I I I I I II 010 1 III I I I I I - I I I I I I I i I I I I ''''III I I I I I - 0- 12 I I I I I I I I I I I I I I I I I I I I I I I - 0 14 0,08 2010 mzg<.-jmxLL-w< Source: By the author. In Figure 28, 1 present a scatterplot of the coefficient of variation for the Education and Income HDI components, in the X and Y axis, respectively. The states in Southern Mexico are marked in blue. The scatter plot is interesting because it shows that most of the states in Southern Mexico rank high in both measures of diversity. In other words, the municipios in the states of Chiapas, Guerrero, Hidalgo, Michoacan, Puebla, Oaxaca, and Tlaxcala vary significantly between each other. 51 Figure 29. Coefficient of Variation of HDI vs. Coefficient of Variation of Income Per Capita by State, 2010 0-12 Coefficient of Varition of Education HDI vs. Income HDI * 0, 10- 0 0 em. * OB -- U U U'$ e 0-06 as 0 - Oil 0 - 0-04 0,02!0.o0 I 0 M 0-05 0 15 0.10 0-20 0-25 Source: By author. Finally, in Figures 30-32, 1 present the coefficient of variation for other relevant variables at the municipio level, including the Infant Mortality Rate, the poverty rate (percentage of the population living in poverty conditions), and the percentage of households without sewage. In Figure 30, it can be seen that most states in Southern Mexico have a high degree of variation in their Infant Mortality Rate, but this is not true for the poverty rate or the access to sewage rate. In the case of the poverty rate, this makes sense, as we have seen that a large proportion of the municipios in Southern Mexico are consistently poor. In the case of the sewage, this is more surprising: it can be interpreted as that most municipios in Southern Mexico have a significant percentage of households without sewage. 52 Figure 30. Coefficient of Variation of IMR by State, I I I I I I I I I I I I 061I 0-5 I I I I I I I I I I I II '04 I I I I III I I I I I II I I I I I I I II I I II I I II I II 1 I I IIIIIII II I I II I II I I I I I I III II I I I I I II I II I 03 2 -4 4- I II III II II I II II II 021 II SI I III - 01 <-a m I ::Izkn I I - I - - -I -I - I I I I II I IIIIIIII I II II I -wo ?MM9886 68524%9*220649 1 ~ - - II jIll m.zxmx jjj I e<00-axU u e-W-x I I 44-4-44- -I -4.4. - I III I I I I II I I II - IIII I I I < Source: By author. Figure 31. Coefficient of Variation of Poverty Rate by State, 2010 Coefficient of Vartion of Poverty Rate by State, 2010 05 I II I II I I I I I I 1- " I i I I I I I I I I i I I I I I I I I I I I I I I I I I I I I I I I I I i I I I I II I I - " I - Ill I I 0-4 I I I I I II I I I 03 II I I I I - I I I I I II - 0-1 I I I - II- - I I - / 0.2 Source: By author. 53 I - III oefficient of Vartion of Infant Mortality Rate by State, 2010 - 0 2010 I Figure 32. Coefficient of Variation of Percentage of Households Without Drinking Water by State, 2010 % Houseolds Without Drinking Water Coefficient of Vartion by State, 2010 P 'l 5 - I I I I I I I I I I I I I I I I I I I I I I I I I I I I I 3 2 I I 4 - I I I I I I I III 6 I I I I I I I I I I I I I I I I I II I I I I I I I I I I I I I I I I I I I I I II 0 iii II I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I II I I I I I I I I I I I I I I I I I I I I I I I I II I II II I I I I Il I I I I I I I I I I I I II Iii liii I I I I I II I II I I II I I I I I I I I I I I I I I I I I I I I I I - III I II I IiI I I II I I I I I I I I I II I I I & I MX Source: By author. In Table 3, the Global Moran I's for key economic and social development variables for the states in Southern Mexico are presented. The measures suggest strong spatial positive autocorrelation in all variables considered, except in the Infant Mortality. In other words, municipios in Southern Mexico tend to be clustered with municipios with similar values (high or low) overall (the Global Moran's I scatter plots are in Appendix D). Table 3. Global Moran's I of Economic and Social Development Variables 0.53 0.53 030 0.15 54 046 0.59 The Global Moran's I indicate that there is strong evidence of spatial structure between municipios in Southern Mexico for key economic and social development variables. Once more, I would like to state that this is only a preliminary step in the process of modeling what is the effect for each of these variables of a municipio being neighbor of another municipios with similar values. In other words, a spatial regression model, which I do not develop here, should be able to investigate with more detail the possible network or feedback effects of this spatial structure. However, as I mentioned before, the global spatial autocorrelation measures do not indicate where this spatial clusters are located. In order to do so, it is necessary to perform local autocorrelation measurements. In Figures 31-36, I present local Moran's I maps for the municipios in the states of Chiapas, Guerrero, Hidalgo, Michoacan, Oaxaca, Puebla and Tlaxcala. In this series of maps, the clusters of municipios with high levels of a given value (i.e., GDP per capita, Infant Mortality Rate, percentage of the population living in poverty conditions) next to other municipios with high levels of the same value are colored in deep red. On the contrary, those cluster with municipios with a low value next to municipios with a low value for the same variables are colored in deep blue. The clusters colored both in light red and light blue represent clusters negative spatial autocorrelation: low and high values, or high and low values, respectively. It is important to highlight that these are clusters deemed statistically significant after performing a permutation test (the significance maps for each one of the maps in Figures 31-36 are in Appendix D). The non-colored municipios represent non statistically significant clusters. The first thing to notice about maps in Figures 33-37 is that there are a significant number of positive autocorrelation clusters in the municipios of Southern Mexico for almost all variables, as was already suggested by the Global Moran's I reviewed. This is, again, of theoretical and substantive relevance in order to highlight the fact that, despite the diversity I discussed in the previous section, there is evidence of possible network or feedback effects between these municipios. Second, it is relevant to notice that even if some of the spatial clusters correspond: for example, the spatial clusters of high GDP per capita in Figure 33 and low levels of the percentage of the population without elementary education, some others are less "intuitive". For example, the municipios with high levels of households without sewage form a corridor that traverses the states of Guerrero and Michoacan, irrespective of the levels of GDP per capita. This is yet another indication of the relevance of thinking in detail the local characteristics of municipios in Southern Mexico. In Figure 31, the spatial clusters of low GDP per capita are mostly present in the states of Oaxaca and Chiapas. It is interest to note that, on the contrary, the splashes of deep red, indicating clusters of high GDP per capita municipios, are surrounding the 55 industrial heart formed by the Federal District and the State of Mexico. In the case of Michoacin, there does not seem to be any statistically significant spatial clusters. Figure 33. Local Moran's I of GDP per capita, 2010 LSA Cluster Map: chphgogromicoaxpueta, Iincome pcl (99 p... LSA Cister Map: chphgogt LI Not Significant (838) High-H* (157) Low-Low (225) Low-High (15) ,~ High-Low (8) 4' Source: National Council for Social Development Policy, 2010, Mexico. In Figure 34, the spatial clusters of low HDI are mostly concentrated in the states of Oaxaca and Chiapas. This is not surprising given the data I presented in the previous section, however the spatial structure that it suggests shed light on the importance of developing a place-based policy for the development of these particular municipios. Again, as in the case of the GDP per capita spatial clusters, the municipios with high HDI appear to form a belt around the Federal District and the State of Mexico. 56 Figure 34. Local Moran's I of HDI, 2010 USA Cluster Map: chphgogromicoaxpueta, Iidh (99 perm) - M= LSA Ckster Map: chphgogi Not Significant (837) Hot-High (193) Low-Low (183) Low-High (14) High-Low (16) 7 Source: National Council for Social Development Policy, 2010, Mexico. In Figure 35, the spatial clusters of the Gini coefficient is extremely interesting. For once, the high-high clusters seems to be mostly concentrated in Michoacin. As mentioned in Chapter 1, this state has currently experienced an increasing level of violence and crime associated with vigilante groups and other non-state armed actors. Contrary to the drug cartels from northern Mexico, these vigilante groups are more prone to be constituted by citizens with no previous crime records and with deep social ties with the local community. The emergence of these groups has created not only security but also governance challenges. For example, in the Summer of 2014, the governor of Michoacatn renounced after a deep political and security crisis that culminated in with the creation of a Development and Security Commission for Michoacin that was heavily criticized given that it challenged the limits of the federal authority vis a vis the state government. As I briefly discuss in the conclusion, his indicates a research agenda focused on the effects of economic polarization, spatial clustering and governance in Southern Mexico. 57 Figure 35. Local Moran's I of Gini Index, 2010 LISA Cluster Map: chphgogromicoaxpueta, Lgini_10 (99 perm) LISA Ckaster Map: chphgoge SNot Significant (857) *Higft-High (190) Low-Low (137) Low-High (31) High-Low (28) Source: National Council for Social Development Policy, 2010, Mexico. In Figure 36, the spatial clusters of low Infant Mortality Rates are mostly concentrated in the state of Chiapas. Only a number of municipios with high rates are spread over the state of Oaxaca. It is interesting to note that, despite the fact that some of these municipios also pertain to clusters of low levels of both GDP per capita and HDI, they seem to be performing rather well in this indicator of social development. Once more, this contributes to the argument regarding the importance of looking closer at the municipios that compose Southern Mexico. 58 LISA Cluster Map: chphgogromicoaxpueta, Ijmr (99 perm) - Figure 36. Local Moran's I of Infant Mortality Rate, 2010 USA Chuster Map: chphgogi Not S fcant (990) * High-.gt (67) Low-Low (128) LOW-*igh (36) High-Low (22) Source: National Council for Social Development Policy, 2010, Mexico. In Figures 37 and 38, the spatial clusters of municipios with a high percentage of households without sewage and the percentage of population without elementary education seem to be clustered in the states of Chiapas and Michoacin. This pattern is clearer in the case of the percentage of households without sewage. In the case of the percentage of population without elementary education, there is also a cluster of high values in the state of Chiapas, were the school infrastructure and human capital is known to be of low quality. Again, in the case of the percentage of population without elementary education, it is important to note a belt of municipios with low values surrounding Mexico City and the State of Mexico. 59 Figure 37. Local Moran's I of % of Households without Sewage, 4 2010 LISA Cluster Map: chphgogromicoaxpuetia, Lporcyvivsn (99 pe... LISA Cksster Map: chphgog Not Signicant (865) Hig*-High (150) Low-Low (189) Low-*igh (27) High-Low (12) Source: National Council for Social Development Policy, 2010, Mexico. Figure 38. Local Moran's I of % of Population without Elementary Education, 2010 LISA Cluster Map: chphgogromicoaxpueta, Lporcbas_5 (99 pe... LISA CWS.ter Map: chphgop Not Signfcent (856) High-High (199) Low-Low (165) Low-high (15) High-Low (8) I. Source: National Council for Social Development Policy, 2010, Mexico. 60 The global and local spatial autocorrelation measures I have presented are a first step in understanding the spatial processes that govern the economic and social development of Southern Mexico. It has only been during the last years that regional scientists and geographers have started to incorporate the spatial dimension in a systematic way in the study of regional inequality and development, as in the pioneering work of SastreGutierrez and Rey (2010). However, the data at the municipio level has only recently has been incorporated in development scholarship in order to gain a more nuanced understanding of the patterns of homogeneity and heterogeneity in the region. 61 Conclusion In Chapter 1I provided a definition of Southern Mexico as composed by the states of Chiapas, Guerrero, Hidalgo, Michoacin, Oaxaca, Puebla and Tlaxcala, and I justified the choice of this regional scheme by its GDP per capita consistency (Sastre-Gutierrez and Rey, 2010). Additionally, in Chapter 1, 1 framed the relevance of studying development in Southern Mexico as a matter not only of economic and social interest, but also of increasing political and governance urgency. As noted before, the challenges Southern Mexico is facing are not new at all, however, as the economy becomes increasingly globalized and diversified, inequality across regions and individuals within them can contribute to create the conditions that may spark social or political conflict. The emergence of guerilla and vigilante groups, as well as other social conflicts in which the federal government has repressed the population in order to reestablish control, are only indicators of this. In Chapter 2, I provided a theoretical and policy rationale to focus on the spatial analysis of key economic and social development data at the municipio level in Southern Mexico. As I have argued, analyzing the spatial structure or patterns in these municipios may shed light on potential network or feedback effects between municipios in the region. Additionally, gaining a more granular understanding of the heterogeneity within the region can help design more effective policy interventions based on the local characteristics of spatial clusters. In Chapter 3, I presented a brief overview of the research done on regional development and inequality. In particular, I focused on the persistent underdevelopment of Southern Mexico and some of its possible causes, including lack of infrastructure, human capital and access to markets. As I argued, as the economy of Mexico globalizes and diversifies, it is unclear that the pattern of underdevelopment for Southern Mexico will change. In order to correct this, both federal, state and Michoacan's governments need to act effectively to promote development in the localities that can sustain their own growth and intervene in those municipios that are facing critical economic and social challenges that may evolve in governance instability. In Chapter 4, I presented global and local spatial autocorrelation measures for municipios in Southern Mexico. Additionally, I presented measures to understand the degree of diversity between municipios in Southern Mexico. This analysis was developed taking into account that development policy should take seriously the heterogeneity within Southern Mexico and, more specifically, the diverging challenges that the region faces. Even if municipios in Southern Mexico are consistently facing economic and social development challenges, they may have different causes or interactions. However, identifying spatial patterns in these municipios can also help designing and 62 implementing more effective policy interventions, if a more detailed study of the potential network or externality effects of some of the variable studies is implemented. Finally, I would like to discuss the scope and limitations of my study, as well as some topics for a future research agenda on the development of Southern Mexico. Regarding the scope and limitation of my study, it is important to highlight that the spatial autocorrelation measures presented constitutes only elementary evidence for a more complete spatial regression model that can help understand the feedback effects between municipios in the region. Further, some of the regional inequality and spatial analysis presented here is descriptive; even though the spatial autocorrelation analysis presented has been tested for statistical significance with permutation tests (Anselin, 2008). However, SastreGutierez and Rey (2010) have also implemented non-parametric approaches to develop inferential statistics for regional inequality measures, for example, by testing the statistical significance of the difference between Theil indexes in different time periods. This is one of many possibilities that signal the increased analytical relevance that spatial analysis may have in the context of regional development and inequality studies (for example, see Vaya et al., 2004). In terms of a future research agenda, I would highlight not only a more sophisticated implementation of spatial analysis tool for state and municipio data in Mexico, but also the development of a broader agenda focusing on the institutional and fiscal design and implementation of a regional development policy for Southern Mexico. As pointed out by Ros (2013), the federal government in the past decades has foregone its role as a major catalyzer of growth in Southern Mexico. However, the increasing security and governance challenges in the region are changing this situation. Given that Mexico has only recently experienced political party competition at all levels of government (federal, state and municipal) the implications of regional development of the fiscal and governance structures that are needed for its implementation is still a relatively overlooked research and policy area. As discussed in Chapter 3, Esquivel (2007) has already developed a preliminary agenda to design a regional development fund that would focus on the diversity of the municipios in Southern Mexico. In fact, Esquivel (2007) developed also developed a regional fund proposal for distinct infrastructure and social equity funds, echoing the need to think economic growth and regional equality as separate policy objectives of Farole, Rodriguez-Pose and Storper (2011). Moreover, Esquivel (2007) developed a basic ranking system to identify municipios in Southern Mexico were intervention is critical given the high levels of income, public services and social inequality. An alternative approach, for example, could make use of clustering techniques, now more common for policy studies, to identify groups of municipios with similar characteristics and challenges that can be targeted collectively by policy interventions. 63 It is also important to note that further research needs to be done regarding regional inequality focusing on income and social polarization as a predictor of social and political conflict (Esteban and Ray, 2011). As discussed by these two authors, polarization (defines as a state of increasing divergence between internally homogenous groups, as measured by income or ethnicity, for example) is a distinct and inherently relevant feature of the social world. Furthermore, it has already demonstrated to be a relevant predictor of political and social conflict. Given the ongoing security and governance crisis that Mexico is experiencing in part of its territory, this is indeed a very promising research and policy agenda that can be augmented when coupled with the kind of spatial analysis I have outlined here. 64 REFERENCES Anselin, Luc, 2008. SpatialEconometrics: Methods and Models, Springer. Anselin, Luc and Sergio Rey, 2014. Modern SpatialExconometrics in Practice:A Guide to GeoDa, GeoDa Space and PySAL, GeoDa Press. Arbia, Giuseppe, 2014. A Primerfor SpatialEconometrics: With Applications in R, Palgrave MacMillan. Aroca, Patricio, Mariano Bosch and William Maloney, 2003. "Is NAFTA Polarizing Mexico?: Spatial Dimensions of Mexico's Post-Liberalization Growth", World Bank, Povery and Economic Unit. Aroca, Patricio, Mariano Bosch and William Maloney, 2005. "Spatial Dimensions of Trade Liberalization and Economic Convergence: Mexico, 1985-2002", The World Bank Economic Review, 19, 3, pp. 345-378. Barca, Fabrizio, Philip McCann, and Andrez Rodriguez-Pose, 2011. "The Case for Regional Development Intervention: Place-Based versus Place-Neutral Approaches", Journalof Regional Sciences, 52, 1, pp. 134-152. Barro, Robert, 1990. "Government Spending in a Simple Model of Endogenous Growth", Journalof PoliticalEconomy, 98, 5, 103:125. Barro, Robert and Xavier Sala-i-Martin, 1991. "Convergence across States and Regions", Brookings Paperson EconomicActivity, 1, pp. 107-182. Bivand, Roger, Edzer Pebesma and Virgilio Gomez-Rubio, DataAnalysis with R, Springer. 2013. Applied Spatial Casar, Jose I. and Jaime Ros, 2004. "iPor que no crecemos?", Nexos, 1 October, Retrieved from: http://www.nexos.com.mx/?p=11271. Davila, Enrique, Georgina Kessel y Santiago Levy, 2002. "El sur tambi6n existe: un ensayo sobre el desarrollo regional de Mexico", EconomiaMexicana: Nueva Epoca, 11, 2, pp. 205-261. Davis, Diane, 2009. "Non-State Armed Actors, New Imagined Communities, and Shifting Patterns of Sovereignt and Insecurity in the Modern World", Contemporary Social Policy, vol. 30 no. 2, pp. 221-245. Diaz-Cayeros, Alberto, 1995. Desarrolloecon6mico e inequidadregionalHacia un nuevo pactofederalen Mexico. Elizondo, Raul Livas, and Paul Krugman, 1992. "Trade Policy and Third World Metropolis", National Bureau of Economic Research Paper, no. 4238. 65 Esquivel, Gerardo, 1999. "Convergencia Regional en Mexico, 1940-1995". Centro de Estudios Economicos, no. 9, Working Paper. Esquivel, Gerardo, et al., 2002. "Why Did Nafta Not Reach the South?", Draft, Retrieved from: http://web.worldbank.org/archive/websiteoo894A/WEB/PDF/ESQUIVEL.PDF. Esquivel, Gerardo, 2007. "El reto de la equidad en el Sur-Sureste de Mexico", El Colegio de Mexico, Draft. Esteban, Joan, 2000. "Regional Convergence in Europe and the Industry-Mix: a Shift Share Analysis", Institut d'Analisis Economica, Retrieved from: http://esteban.iaecsic.org/pdf/ShiftShare.pdf. Esteban, Joan and Debraj Ray, 2011. "Linking Conflict to Inequality and Polarization", American Economic Review, pp. 1345-1374. Farole, Thomas, Andres Rodriguez Pose, and Michael Storper, 2011. "Cohesion Policy in the European Union: Growth, Geography, Institutions", Journalof Common Market Studies, vol. 49, no. 5, pp. 1089-1111. Friedman, John, Nathan Gardels, Adrian Pennink, 2009. "The Politics of Space: Five Centuries of Regional Development in Mexico", InternationalJournalof Urban and RegionalResearch, vol. 4, no. 3, pp. 319-349Gamboa, Rafael and Miguel Messmacher, 2003. "Desigualdad regional y gasto publico en Mexico", Banco Interamericano de Desarrollo, Documento de Divulgacion, no. 21. Retrieved from: http://publications.iadb.org/bitstream/handle/11319/1235/Desigualdad_%2oregional_ y-gasto p%C3%BAblicoenM%C3%Axico.pdf?sequence=1. Garcia Verdu, Rodrigo, 2005. "Income, Morality and Literacy Distribuion Dynamics across States in Mexico: 1940-2000", Cuadernosde Economia, vol. 42, pp. 165-192. Gonzalez Rivas, Marcela, 2011. "Trade Openness, Infrastructure and the Wellbeing of Mexico's South", Mexican Studies, vol. 27, no. 2, pp. 407-429. Gleditsch, Kristian, and Michael D. Ward, 2007. An IntroductionTo Spatial RegressionModels in the SocialSciences, Sage. Guash, J. Luis and Marianne Fay, 2003. "Economic Activity, Agglomerations, and Logistics in the Mexican Southern States", Draft. Hall, Gillette and Christopher Humphrey, 2003. "Mexico: Southern States Development Strategy. Volume 1: Synthesis Report", World Bank Working Paper, Draft.. 66 Horton, Gillian, 2014. "Conflict in Michoacain: Vigilante Groups Present Challenges and Opportunities for Mexican Government", Woodrow Wilson CenterMexico Institute. Retrieved from: http://www.wilsoncenter.org/sites/default/files/hortonMichoacan.pdf. Illades, Esteban, 2014. "Guerrero: La violencia circular", Nexos, Retrieved from: http://www.nexos.com.mx/?p=23092. Instituto Mexicano para la Competitividad 2014, 2014. Indice de Competitividad Estatal. Krugman, Paul, 1979. "Increasing Returns and Economic Geography", The Journalof PoliticalEconomy, 99, 3, pp. 438-499. LeSage, James and Robert Kelley, Chapman and Hall. 2009. Introductionto SpatialEconometrics, Martin, Ron and Peter Sunley, 1998. "Slow Convergence? The New Endogenous Growth Theory and Regional Development", Economic Geography, vol. 74, no. 3, pp201-277. Messmacher, Miguel, 2000. "Desigualdad regional en Mexico: El efecto del TLCAN y otras reformas estructurales". Documento de Investigacion 2000-2004, Banco de Mexico. Rey, Sergio and Mark Janikas, 2004. "Space-Time Analysis of Regional Systems", Draft. Retrieved from: http://econwpa.repec.org/eps/urb/papers/o406/o4o6o01.pdf Ros Bosch, Jaime, 2013. "Algunas tesis equivocadas sobre el crecimiento economico de Mexico", Draft. Sastre-Gutierrez and Rey, 2010. "Interregional Inequality Dynamics in Mexico", SpatialEconomic Analysis, vol. 5, no. 3, pp. 277-298. Sastre Gutierrez, Myrna and Segio Rey, 2008. "Polarizacion espacial y dinimicas de la desigualdad interregional en Mexico", Problemasdel Desarrollo:Revista Latinoamericanade Economia, vol. 39, no. 155, pp. 181-201. Scott, Allen J. and Michael Storper, 1990. "Regional Development Reconsidered", Lewis Center for Regional Policy Studies, University of California, Los Angeles, Working Paper, no.1. Retrieved from: http://16 4 .67 .121.2 7 /files/LewisCenter/publications/workingpapers/WP1.pdf Vaya, Esther, et al., 2004. "Growth and Externalities Across Economies: An Empirical Analysis Using Spatial Econometrics", in Anselin, Luc, R. J. G. M. Florax, and S.J. Rey, 2004. Advances in SpatialEconometrics:Methodology, Tools and Applications, Springer, pp. 433-456. 67 Villareal, M. Angeles, 2010. "NAFTA and the Mexican Economy", Congressional Research Service. Ward, Michael and Kristian Skrede, 2008. SpatialRegression Models, Dover. Wilson, Christopher, and Gerardo Silva, 2014. "Mexico's Latest Poverty Stats", Woodrow Wilson Center, Mexico Institute. News Articles "El gobierno de Pena 'abre la cartera' para 'reconstruir' Michoacan", 4 February 2014, CNN Mexico in http://mexico.cnn.com/nacional/2014/o2/o4/michoacanoperativo-pena-nieto-autodefensas-militares-alfredo-castillo. Montalvo, Tania , 8 December 2014. "Usan programas viejos para el nuevo Plan Guerrero", Animal Politico. Retrived from: http://www.animalpolitico.com/2014/12/plan-guerrero-repite-programas-y-objetivosanunciados-en-2013/. Sanchez, Enrique, 4 December 2014. "Desarrollo para tener paz: EPN; anuncia reactivaci6n econ6mica en Guerrero", Excelsior. Retrived from: http://www.excelsior.com.mx/nacional/2014/12/05/995887. 68 Appendix A. Human Development Index by Municipio, 2010 Human Development Index, 2010 [] 0.00- 0.58 E 0.58 - 0.63 m m 0.63 - 0.66 0.66 - 0.71 0.71 -0.92 .4 It.. 4 I- Source: United Nations Development Programme, Human Development Index, 69 2010. GDP per capita, Global Moran's I, 1940 Moran's I (map-pibepc4012_base95_queen): 1940 Moran's I: 0.111426 0 10 018 6O - r: V) 0 1 o -I -3.5 -2.1 0.7 -0.7 1940 70 2.1 3.5 GDP per capita, Global Moran's I, 198o -3 Moran's I (mappibepc4O12_-base95_queen): 1980 Moran's t. 0.0854763 - P 0 00 0 00 0 I, 0 0% 0 01 0 0 CI -3.2 -2 0.4 -0.8 1980 71 1.6 2.8 GDP per capita, Global Moran's I, 1990 Moran's I (mappibepc4O2_base95_queen): 1990 -C3 Moran's 1: 0.147043 0 10 9- C~4 - I 00 -4 -2.4 0.8s -0 8 1990 72 2.4 4 GDP per capita, Global Moran's I, 2010 Moran's I (map_pibepc4012_base95_queen): 2010 Moran's t 0.144356 0I1 4 110 0 0 % I 0 ...... ...... 0 0~ I', Ce 4 -4.1 -2.5 0.7 -0.9 2010 73 2.3 3.9 Appendix B. Density of Municipios in Southern Mexico (Inhabitants per Kma), 2010 [~] 1.9 - 21.0 EJ 21.0 - 41.4 41.4 - 75.9 m m 75.9 - 146.8 146.8 - 4665.6 $ ~K. -4 Source: National Institute of Statistic and Geography, Mexico 74 Appendix C. LISA Significance Map: chphgogromicoaxpueta, Iincomepc1 (... Map: chp LiA Significance Not Significant (838) U- p - 0.05 (199) p - 0.01 (206) p = 0.001 (0) p a 0.0001 (0) 75 - GDP per capita Cluster Significance Map and Global Moran's I, 2010 - Moran's I (chphgogromicoaxpueta): incomepcl I* - . Moran's 1: 0.538954 1 0 0 -8 -S -2 1 incomepcI 76 4 7 HDI per capita Cluster Significance Map and Global Moran's I, 2010 4 LISA Significance Map: chphgogromicoaxpuetia, Iidh (99 perm) USA Significance Map chp SNot Sgnificant (837) p * 0.05 (220) p - 0.01 (186) p- 0.001 (0) p* 0.0001(o) 77 I Moran's I (chphgogromicoaxpueta): idh Morans t 0.531881 10 0 0 000 * CD4 o I I 0008 oo o 0100 * 00 0 0 0 0 OD 0 0 o 0, 0 0 o 008 8 .4 -4 -2.4 0.8 -0.8 idh 78 2.4 4 Gini Index Cluster Significance Map and Global Moran's I, 2010 USA Significance Map: chphgogromicoaxpuetia, ILgini_10 (99 p... LISA Significance Map: chp SNot Significant (857) p , 0.05 (225) p - 0.01 (161) 0.001 (0) p - 0.0001 (0)4 p- " L1, 79 Moran's I (chphgogromicoaxpuetla): gini_10 Morai's t 0.368019 0 00 00 90 C8 0 0 0 0 o 0 0 OD 0 it 00 0 e4 (0 -4.6 -2.8 -1 0.8 gvin10 8o 2.6 44 SLISA Significance Map: chphgogromicoaxpuetla, Iimr (99 perm) LISA Significance Map: chp Not Significant (990) p 0.05 (172) P .01(81) 0.001 (0) P 0.0001 (0) P- I 81 - Infant Mortality Rate Cluster Significance Map and Global Moran's I Moran's I (chphgogromicoaxpuetla): imr (n - Moran's t 0 188214 0 - 0 o .......... 0 0 ............................................. ? - I V- a -7 -5 -3 1 -1 imr 82 3 S ' LISA Significance Map: chphgogromicoaxpuetia, Lporcyvivsn (9... USA Significance Map: chp LI U U U U Not Significant (865) pz 0.05 (195) p 0.01 (183) pa 0.001 (0) pa 0.0001 (0) 83 - Percentage of Households without Sewage Cluster Significance Map and Global Moran's I, 2010 Moran's I (chphgogromicoaxpuetla): porcvivsn Moran's t 0.464152 00 00 00 o 0 0 00 -28 -1 0.8 porcvvsn 84 o 0~0' 0O 00 -4.6 CD 900 0 0 % - 0 0 2.6 44 LISA Significance Map: chphgogromicoaxpuetla, Lporcbas_5 (... - Percentage of Population without Elementary Education Cluster Significance Map and Global Moran's I, 2010 LSA Significance Map: ctvp Not Signflcmnt (856) p - O.OS (157) P - 0.01 (230) P - 0.001 (0) P - 0.0001 (0) '4W1 85 Moran's I (chphgogromicoaxpueta): porc bas_5 Morans t 0598874 8 0 0 jo (0 -4O3 -2.6 08 -09 porc..bes_.5 86 2.5 42