Provision of Local infrastructure in Indonesia A Cultural

Document Sample
Provision of Local infrastructure in Indonesia A Cultural Powered By Docstoc
					NORMS, CULTURE AND LOCAL
INFRASTRUCTURE IN
INDONESIA
                         Sarmistha Pal
            CEDI Brunel University (UK)
                       & IZA (Germany)
INTRODUCTION
 Inrecent years there has been a renewed
 interest to analyse whether the varied
 economic paths of different societies over
 time can be traced to differences in
 culture, customs, social norms and
 religion (e.g., Guiso and Zingales, 2006).
      The present paper builds on this literature
       to assess the role of norms and culture on
       the     provision     of    local    public
       infrastructure,   which    remains    little
       understood.
AIMS AND OBJECTIVES
 Public goods are not allocated through markets.
 Collective action is key to the provision of public
  goods, but it can often give rise to the problem of
  conflicting community preferences.
 Norms and culture could however help reconcile the
  conflicting community preferences.
 We explore one possible way through which norms
  and culture can affect a community’s choice of local
  public infrastructure.
THE INDONESIAN              CASE
 The  analysis is based on information from 314 Indonesian
  communities drawn from 13 provinces over a period of
  1993-2007.
 Indonesia is an important case in point. Indonesians are
  grappling with the problems of living in a world of differing
  ideas of norms.
   There are long-standing efforts to shape lives in an Islamic way;
   There are even longer standing and diverse efforts to share them
    according to complex local norms and tradition called adat.
   Indonesians have been trying to reconcile these diverse ideas
    over time, as the concept of the nation state evolves under
    different forms of governments since the 1950s.
A CULTURAL PERSPECTIVE
TO THE PROVISION OF PUBLIC GOODS
 Social norms may influence social organisation in different ways:
    Individualist cultural beliefs led to the formation of efficient
     agency relations in middle ages (Grief 1994)
      Ethnically linked parochial groups could achieve high
       levels of cooperation in informal contracts while
       engaging in exclusionary practices (Bowles and Gintis, 2004).
 Religious norms could also play an important role in community
  development in an intriguing way.
    Religious people and institutions may be agents of
     advocacy, funding innovation, empowerment, social
     movement and service delivery.
    Religious people and institutions can also incite
     violence, oppose empowerment, deflect advocacy, absorb
     funding and cast aspersions on service delivery.
A POSSIBLE MECHANISM
 People   may find it easier to trust those who have
  faith in the same norms, which in turn may affect
  optimal investment in social capital.
 The latter could be facilitated by the perception of
  a common external threat, giving rise to feelings
  of loyalty and norms of solidarity to protect
  community identity/livelihood.
 Social and religious norms could thus help
  reconcile conflicting preferences within a
  community, thus resulting in preferences (or lack
  of it) for certain local public infrastructural goods
  as opposed to others.
ADAT NORMS IN INDONESIA
 Indonesia’s  pluralistic identity gives rise to a coexistence of
  traditional adat laws, Islamic Sharia laws and positive laws of
  the modern state.
 Literally ‘adat community’ translates to ‘autonomous’
  groups of indigenous people who are able to manage their
  lives without knowing western laws and established their
  own regulations and social control.
 Adat laws are a set of local and traditional laws concerning
  marriage, inheritance and dispute resolution systems.
 Adat livelihoods are often linked to land, water and natural
  resources, thus giving rise to a culture that is primarily rural
  in nature.
ISLAMIC NORMS IN INDONESIA
 In  addition, there has been a historical division
  between ‘traditional’ and ‘modern’ Islamic
  values/practices in Indonesia:
    One can distinguish ‘Muslim modernists’, who seek to
    reform Indonesia, from the traditional ones.
   While the traditional Islamic practices lean on Sharia
    laws, Muslim modernists have close ties with the positive
    laws of the state.
   Islam is the main religion in more than 80% of the sample
    communities while it is so in more than 86% of the
    communities which strongly adheres to adat. Thus adat
    communities are predominantly Islamic communities.
ISLAMIC LAW AND ADAT ENCOUNTER
   Indonesians have successfully harmonized the two legal
    traditions, namely adat and Islamic laws.
     It is envisaged that Adat and Islamic laws have existed side by side
      long before the intervention of the colonial powers in Indonesian
      legal affairs.
     The dialogue between the two sets of laws persists even in
      modern Indonesia which has been reflected in Indonesian laws on
      conditional repudiation, common property in marriage,
      obligatory bequest and also conflict resolution.
A HISTORICAL PERSPECTIVE
 Suharto’s New Order Period starting in the late 1970s
  witnessed efforts to undermine the ethnic identities of adat
  communities with a view to promote the integration of the
  nation. This has been implemented through policies and
  programmes, e.g., significant changes in property rights in land
  and other natural resources, which threatened the very basis of
  adat livelihoods.
 This period has witnessed efforts to boost investment in
  infrastructural development. Several major projects were
  undertaken, which have significantly improved the availability
  of community-level infrastructure and public services in the
  country.
 Onset of the economic crisis of the 1990s had however cast a
  major blow to the infrastructural investment/ development in
  the country, which was further undermined by the introduction
  of decentralisation at the turn of the century.
NORMS & LOCAL GOVERNANCE IN INDONESIA
 The   Dutch colonial rule recognized village governments as
  lawful entities and encouraged self-rule according to Adat
  laws, which were in place until Suharto took power in 1978.
 While adat laws were formally banned during Suharto’s
  regime, the formal ban did not result in the abandonment of
  these adat laws and the extensive decentralization process that
  followed the demise of Suharto only reinstated them in 2001.
 Given      this   long-standing     tradition    of   political
  decentralisation, community culture is likely to reflect the
  preferences of the community for the provision of local
  infrastructure through decentralised community decision
  making process.
STRUCTURE OF THE LOCAL GOVERNMENT
 Local government consists of a headman assisted by an appointed
  village assembly (LMD) throughout the country.
 Until 2001, the headman was generally elected every 8 years and was
  accountable to the district government. Development projects and
  assistance were managed by community resilience boards (LKMD)
  who allocated development grants (mainly from the central
  government) across households and projects.
 Law 22/99, enacted by January 2001 gave villages more autonomy in
  raising local revenues replacing central grants (i.e., fiscal
  decenralisation). Elections for both the headman and the council now
  take place every five years and the headman is directly accountable to
  the council.
COMMUNITY DECISION MAKING
 Although  the headman is generally elected, there could be
  one of the four ways of electing the headman:
   Voting
   Consensus building
   Local elite
   Local government institutions
 Similar methods are used to make decision making regarding
  provision of local public goods.
 While the IFLS data provides information on selection of
  headman as well decision making process, we do not include
  nature of governance in analysing community’s choice of
  local infrastructure, as it is likely to be endogeneous to the
  choice of public goods that we determine.
EMPIRICAL FRAMEWORK
 We  use 4 available rounds of Indonesian Family Life Survey
  (IFLS) data for the years 1993, 1997, 2000 and 2007 to
  study the provision of local infrastructure.
 This includes data from 314 rural and urban communities
  in 13 provinces including, Jakarta, Bali, Java (central, east
  and south), Sumatra (north, west and south), Lampung,
  Wntenara and south Kalimantan.
 The panel nature of the data allows us to control for the
  potential endogeniety bias arising from community-level
  unobserved heterogeneity. Time span of our sample also
  allows us to control for time fixed effects.
LIST OF PHYSICAL INFRASTRUCTURAL GOODS
 We  consider a number of basic infrastructural goods, both
 physical and social, that can directly impact on sustainable
 livelihoods and provide opportunities for all, especially for
 the poor.
     Since culture economic backwardness and poverty in the country
      have often been caused by remoteness and isolation, roads and
      different modes of motorized transport have a crucial role for
      economic development and poverty alleviation.
     Similarly, communication goods like post and telephone have a role
      in reducing the disadvantages related to location and distance.
     We       include     two      more       essential    infrastructural
      services, namely, banks and markets that could facilitate formal
      exchange, thus contributing to the process of economic
      development.
LIST OF SOCIAL INFRASTRUCTURAL GOODS
 We consider two types of social infrastructure, namely, share
 of government schools (elementary, junior and senior schools)
 and government health facilities (health centres and health
 posts) in the community.
     An important distinction between social and physical infrastructure like
      transport is worth highlighting here. Social infrastructure could boost
      exchange both within/outside the community, as it enhances skills and
      productivity of the community population. Transport and communication
      infrastructure, in contrast, brings different communities closer and thus has
      the innate feature of improving exchange with the external communities.
ACCESS TO LOCAL INFRASTRUCTURAL GOODS
 Table   1 summarises the sample communities’ access to
  various local public infrastructure during 1993-2007.
 There has been an improvement in the provision of a
  number of public goods over this period, especially in the
  pre-crisis period.
 The progress slowed down somewhat after the crisis for
  some infrastructural goods most notably banks. The latter
  could be linked to the fact that the 1997 crisis started in the
  financial and banking sector though it quickly spilled over to
  the real sector.
 Introduction of fiscal decentralisation in 2001 has a mixed
  effect. While access to PTO has gone up significantly, access
  to many other local infrastructure has declined in 2007.
ACCESS TO LOCAL INFRASTRUCTURE 1993-2007
                      % of total sample communities           Mean
                                                              proportion
                                                              (sd)
    Community’s        1993       1997         2000   2007     1993-2007
       access to
   Bus                 33.1       27.3         43.9   32.69   0.35 (0.48)

   Public transport    78.6       78.9         81.0   75.0    0.78 (0.41)

   Public telephone    42.4       50.8         64.6   75.32   0.60 (0.62)
   office (PTO)
   Post office (PO)    25.2       26.7          30    24.68   0.31 (0.66)

   Pucca       Road    71.3       79.7         83.3   99.36   0.84 (0.20)
   (PROAD)
   Electricity          91        96.2         98.1   99.4    0.97 (0.22)
   (ELEC)

   Market (MKT)        36.9       39.7         45.3   45.19   0.43 (0.63)

   Bank (BANK)         41.9       40.3         15.4   18.59   0.36 (0.74)
INTER-COMMUNITY HETEROGENEITY
 Despite   the overall improvement over the sample period,
  inter-community heterogeneity has been pronounced. One
  could classify a community to be strongly adat (those who
  strongly adhere to adat norms and laws) and others.
 Table 2 summarises the mean differences in the provision of a
  range of basic infrastructural goods communities.
 Clearly, access to motorized public transport, telephone office,
  post office, cemented road are significantly lower in the under-
  developed communities.
 The difference is however not statistically significant in case of
  share of health facilities while adat communities tend to have
  significantly higher share of public schools in our sample.
 STRONGLY ADAT & OTHER COMMUNITIES
                                Community classification
Variables                       Strongly     Others        T-stat
                                adat
Access to busstop               0.26          0.38         -3.913**
Access to any motorized         0.73          0.81         -2.888**
public transport
Access to market                0.35          0.47         -3.427**
Access to PTO                   0.47          0.66         -4.250**
Access to PO                    0.14          0.35         -6.912**
Access to bank                  0.23          0.35         -3.492**
Access to pucca Road            0.76          0.86         -4.023**
Access to electricity           0.95          0.98         -2.470**
Access to piped water           0.48          0.63         -4.958**
Share of government schools     0.67          0.60         4.204**
Share of government health      0.57          0.58         -0.919
facilities
Rural                           0.60          0.38         6.674**
Border with sea                 0.48          0.44         1.010
District HQ or Provincial       0.17          0.20         -1.531
capital
If Islam is the main religion   0.86          0.78         3.750**
Size of largest population      91            79           11.833**
group (%)
Ethnic heterogeneity            0.11          0.26         -7.387**
Community size                  4706          4368         0.170
Population (number)             7040          11005        -6.181**
IDENTIFICATION OF ADAT COMMUNITIES
   Adat communities have the following
    characteristics
       relatively larger in size but have significantly lower
        population so that population density is lower.
       Ethnic heterogeneity is significantly lower.
       Predominantly rural and have borders with sea
       Islam is the main religion
       A lower proportion of these communities have district
        head quarter or provincial capital.
INTER-PROVINCE VARIATION
 There is considerable variation in culture, geography and environment
  across the country. Table 3 shows the summary statistics for the
  selected community characteristics across the sample provinces.
 In general, provinces with higher average population per community,
  higher proportion of university educated population and lower
  proportion of adat communities tend to have better provision of all
  types of public infrastructural goods under consideration;
 Also more developed provinces tend to have relatively lower
  proportion of under-developed communities.
 A significantly higher proportion of strong adat communities in our
  sample is concentrated in Sulawesi, Lampung, south Kalimantan, Bali,
  Wntenara.
 INTER-PROVINCE VARIATION
                Community characteristics
                Mean (standard deviation)
                Population             University educated Strong           Islam is the main Under-developed
                                       population          adherence     to religion
                                                           adat laws
Jakarta         30023.25 (14228.6)     0.19 (0.26)         0.03 (0.17)      0.31 (0.46)       0.11 (0.32)
West Java       10056.3 (9693.6)       0.08 (0.12          0.20 (0.40)      0.98 (0.14)       0.14 (0.35)
East Java       7424.09 (5740.1)       0.69 (0.88)         0.38 (0.49)      0.96 (0.21)       0.24 (0.43)
Central Java    6513.25 (6275.7)       0.12 (0.32)         0.20 (0.40)      0.97 (0.17)       0.22 (0.42)
North Sumatra   5562.9 (5639.8)        0.14 (0.49)         0.15 (0.36)      0.46 (0.50)       0.19 (0.40)
South Sumatra   3869.6 (2499.6)        0.25 (1.10)         0.13 (0.34)      0.87 (0.34)       0.20 (0.40)
West Sumatra    2453.4 (1099.2)        0.11 (0.15          0.29 (0.46)      0.93 (0.26)       0.21 (0.42)
Bali            8624.3 (1599.6)        0.19 (0.22)         0.50 (0.51)      0 (0)             0.27 (0.45)
Wntenara        8206.4 (4621.3)        0.05 (0.71)         0.50 (0.63)      0.87 (0.33)       0.63 (0.49)
Ykarta          13411.00 (10081.7)     0.19 (0.25)         0.26 (0.44)      1.00 (0.00)       0.16 (0.37)
Lampung         5016.09 (2771.4)       0.03 (0.03)         0.45 (0.51)      0.81 (0.39)       0.27 (0.45)
Sulawesi        4897.0 (5218.15)       0.08 (0.13)         0.63 (0.49)      0.63 (0.49)       0.31 (0.47)
South           3850 (4040.6)          0.08 (0.12)         0.46 (0.51)      0.85 (0.37)       0.15 (0.37)
Kalimantan
MAP OF INDONESIA
DETERMINATION OF PUBLIC INFRASTRUCTURE
AND INDICES OF COMMUNITY DEVELOPMENT
 We     estimate the following equation
 Yit   = ’Xit + i +uit
   Yit is i-th community’s access to selected infrastructural goods in year
    t.
 Yit   refer to
       binary indictors of access to various public infrastructural goods
       a composite index of access to various physical infrastructural
        goods, using factor analysis.
       the likelihood of a community being underdeveloped. The latter is
        based on whether a community has been selected for the major
        anti-poverty programme IDT in the country during this period.
CENTRAL HYPOTHESIS
 We     argue that in a politically decentralised
 system, community norms would account for community
 preferences for different public infrastructure (physical and
 social).
     Initial descriptive statistics suggest that traditional communities
      tend to have more social rather than physical infrastructure
      considered in our analysis. It is possible that typically traditional
      communities would be averse to those physical (as opposed to
      social) infrastructural goods that removes barriers to distance and
      isolation in a bid to preserve their indigenous identity. In the
      absence of a prior we empirically explore this hypothesis.
MEASURING NORMS & CULTURE
   It is not easy to measure norms and culture perfectly.
     IFLS data provides information on whether a community
      strongly adheres to adat laws.
     However initial analysis suggests that the concept of adat
      has many dimensions.
           It is closely linked to the location (e.g., rural and access to sea)
            of the community
           Often it is closely linked to the dominant religion of the
            country, namely, Islam.
     In order to exploit this multi-dimensional aspect of norms
      and culture in our context, we use factor analysis to
      construct a composite index of strong adherence to adat
      laws, strong adherence to Islam, rural location and access
      to sea and label it as PCNORM.
     We compare the estimates using this composite index with
      that using the individual constituent factors of adat to test
      the robustness of our estimates.
OTHER CONTROL VARIABLES
   A competing hypothesis would be investments in physical
    infrastructure have been neglected in traditional communities because
    they were deemed less justified in economic terms.
       We thus control if the district head quarter and/or provincial capital is
        located in the community.
   In order to ascertain the nature of collective action, one needs to
    control for group size and sustainability as well.
       To this end, we include community population, proportion of total
        households with junior high school and also if the largest population group
        has 90% or more population.
   Finally, characteristics of the headman may importantly influence the
    community preferences for one public good or other.
      To this end, we include community leader’s characteristics , e.g. if
       the leader has senior high school education or more, if his/her
       tenure in the community has been >=10 years.
ECONOMETRIC ISSUES AND SPECIFICATIONS
 Often social and religious norms tend to be persistent and change only
  slowly, even when they are no longer efficient. We could thus take
  indices measuring norms to be exogenously given. Reverse causality is
  also unlikely.
 For each dependent variable of our choice, we try three specifications:
     We start with conventional covariates group size, its sustainability and
      community leader’s characteristics.
     We then include adat, Islam, access to sea and rural location individually
     We replace individual constituent factors of adat culture by the composite
      measure of culture, namely PCNORM.
           We argue that given the multidimensional nature of social norms, the composite index
            PCNORM is a better measure of norms in our context .
MODEL CHOICE
 For  the binary indicators of infrastructural goods and
  also community development we apply binary probit
  RE or logit FE models.
 Composite index of selected physical infrastructural
  goods (PCDEV) is continuous in nature – so in this
  case we consider standard fixed and random effects
  regression models.
 Similarly,   share of government schools and
  government health facilities are continuous in nature.
  Hence, in these cases too we use standard fixed and
  random effects regression models.
 For each case we also estimate the corresponding
  pooled model. For robustness checks, we compare
  pooled results with the panel estimates.
FIXED VS. RANDOM EFFECTS ESTIMATES
 We   use Hausman test to choose between fixed/random
  effects model for determining the continuous variable
  PCDEV. The null hypothesis is that i is uncorrelated; in
  other words, acceptance of the null implies an acceptance of
  random effects estimates.
 We consider the Lagrange Multiplier (LM) test when the
  dependent variable is binary in nature. The LM test boils
  down to a test of significance of ρ. Rejection of the null
  hypothesis that ρ=0 leads to an acceptance of the random
  effects estimates.
 Given that the parameter ρ is statistically significant, we
  focus our attention on the random effects estimates.
CENTRAL RESULTS
 An advantage of these random effects estimates is that we can directly
  control for the time-invariant factors (e.g., social norms).
 Even after controlling for all other factors, composite index of adat
  norms is statistically significant for the provision of pucca road, post
  office, telephone office. In other words, communities with strong
  adherence to adat norms tend to have less access to these local
  infrastructure.
 Contrasting effect of local norms for access to social infrastructural
  goods is noteworthy. Communities strongly adhering to adat norms
  seem to significantly greater share of public schools while the effect is
  not statistically significant for the share of government health facilities
  (the coefficient is still positive).
PANEL ME ESTIMATES OF INDIVIDUAL PHYSICAL
INFRASTRUCTURAL GOODS
                             (1)         (2)         (3)        (4)         (5)          (6)
       VARIABLES           pubtrans      pto         po        proad       market       bank

        Population          0.182*    1.084***    0.691***    0.683***    0.955***     0.479***
                            (0.107)     (0.152)     (0.167)     (0.162)     (0.225)      (0.119)
       Hhs with JHS        2.058***   3.491***    2.222***    4.216***       1.492     1.826***
                            (0.525)     (0.646)     (0.756)     (0.880)     (1.013)      (0.544)
   Largest pop group>90%    -0.0319      -0.370      0.718     -0.0961       0.148       0.844*
                            (0.454)     (0.541)     (0.601)     (0.706)     (0.700)      (0.488)
        Head>=shs            0.109       0.224       0.186       0.148       0.123       0.0629
                            (0.146)     (0.173)     (0.213)     (0.180)     (0.218)      (0.162)
        Tenure>=10          0.0821      0.0325     -0.0179       0.201       0.188       0.0367
                            (0.125)     (0.142)     (0.159)     (0.160)     (0.163)      (0.128)
      District HQ/PC       0.882***   0.580***    0.711***       0.219    0.629***     0.434***
                            (0.201)     (0.197)     (0.176)     (0.258)     (0.200)      (0.167)
          Pcnorm            -0.0631   -0.344***    -0.260*     -0.252*     -0.0452        -0.114
                           (0.0936)     (0.116)     (0.137)     (0.136)     (0.172)      (0.103)
         Constant            -1.119   -10.99***   -9.156***   -6.256***   -10.21***    -5.992***
                            (0.972)     (1.383)     (1.572)     (1.436)     (2.012)      (1.108)
         Province             Yes         Yes         Yes         Yes         Yes          Yes
         Year FE              Yes         Yes         Yes         Yes         Yes          Yes
         Lnsig2u            -0.423*    -0.4452*   0.629***     0.646**    1.570***    -0.7460***
                            (0.248)     (0.258)     (0.233)     (0.299)     (0.208)      (0.248)
      Observations            1183        1181       1181        1183        1181          1181
    Number of commid           310         310        310         310         310           310
OTHER RESULTS
 There   is confirmation that larger communities tend to have
  more access to different public infrastructure considered here.
 Community’s proximity to district head quarter or provincial
  capital is highly significant for its access to most infrastructural
  goods.
 Communities with higher proportion of JHS educated
  households are more likely to get most public goods
  (interesting exception being share of govt. schools).
 Leader’s characteristics are not so significant except for the
  likelihood of underdevelopment. Communities with shs
  educated leaders are less likely to be underdeveloped.
COMPOSITE PUBLIC GOODS, SOCIAL
INFRASTRUCTURE & UNDERDEVELOPMENT
                                  (1)        (2)         (3)           (4)
    VARIABLES                    pcdev   shgov_sch   shgov_hlth      undev
          Population         0.266***    -0.0255**   0.0248***     -0.229**
                              (0.0443)    (0.0126)   (0.00421)     (0.0958)
         Hhs with JHS        1.190***    -0.209***    0.0517**    -1.221***
                               (0.213)    (0.0604)    (0.0206)      (0.465)
     Largest pop group>90%      -0.107     0.0468      0.0179      0.987**
                               (0.186)    (0.0510)    (0.0187)      (0.420)
          Head>=shs           0.142**      0.0260     -0.00110     -0.301**
                              (0.0659)    (0.0170)   (0.00766)      (0.152)
          Tenure>=10           0.0779    0.0308**     -0.00255       -0.107
                              (0.0532)    (0.0137)   (0.00629)      (0.138)
             Rural             -0.131*    -0.00265    0.00435       -0.286*
                              (0.0776)    (0.0214)   (0.00778)      (0.166)
          Close to sea        0.00971     -0.00380   -0.000971    0.0991**
                              (0.0202)   (0.00517)   (0.00239)     (0.0406)
            dhq_pc           0.329***     -0.00512    -0.00145      -0.0819
                              (0.0676)    (0.0175)   (0.00780)      (0.203)
            pcnorm           -0.0807**    0.0191*     0.00371       0.101*
                              (0.0348)   (0.00977)   (0.00338)     (0.0575)
           Province               Yes        Yes         Yes           Yes
           Year FE                Yes        Yes         Yes           Yes
           Constant          -2.802***   0.963***     0.522***       0.493
                               (0.416)     (0.118)    (0.0402)      (0.908)
             SigU                                                 -1.033***
                                                                    (0.363)
         Observations          1181        1177        1160           1183
       Number of commid        310         310         310            310
ESTIMATES OF COMPOSITE PHYSICAL
INFRASTRUCTURAL GOODS

 In this case we consider a composite index of access to
  various public goods (PCDEV) namely, motorized
  transport, road, post office, public telephone office,
  bank and market, obtained by using factor analysis.
 Given the value of the Hausman statistic, we choose
  the random effects estimates.
 These estimates confirm that strong adat communities
  tend to have lower provision of composite public
  goods in our sample.
LIKELIHOOD OF COMMUNITY UNDER-
DEVELOPMENT
 Finally, we determine an index of under-development,
  which is a binary variable indicating whether a community
  has been selected for the on-going IDT programme in the
  relevant year.
 Given that ρ is significant, we choose the random effects
  estimates.
 Clearly these random effects estimates highlight that, ceteris
  paribus, traditional communities with strong adherence to
  adat are more likely to be underdeveloped.
CONCLUDING         COMMENTS        1
 Other things being equal, cultural considerations seem
  to exert a significant impact on the provision of local
  infrastructure in our sample.
 In particular, traditional rural communities (with
  strong adherence to adat and Islam) tend to have
  significantly   lower    access    to    key   physical
  infrastructure (e.g., road and communication
  goods), which may reduce barriers of distance and
  isolation.
 We argue that this is a reflection of these community’s
  lack of preference for these goods, especially in
  response to a common threat of modernisation. The
  latter may give rise to feelings of solidarity within
  traditional community members to protect their
  indigenous livelihoods, generally from land and
  natural resources.
    TABLE A1. POOLED PROBIT
 VARIABLES       pubtrans      pto          po          proad      bank       market

    Population   0.0306*    0.311***     0.113***     0.0383***   0.101***   0.164***
                 (0.0169)   (0.0325)     (0.0188)     (0.00988)   (0.0200)   (0.0241)
 Hhs with JHS    0.410***   1.020***     0.282***      0.244***   0.376***    0.254**
                 (0.0834)    (0.143)     (0.0846)      (0.0587)   (0.0907)    (0.110)
   Largest pop    -0.0271    -0.0427       0.131        0.00613    0.198**    -0.0574
   group>90%
                 (0.0786)      (0.129)    (0.0843)     (0.0336)   (0.0915)    (0.105)
   Head>=shs      0.0331       0.0365      0.0489       0.0246     0.0218     -0.0514
                 (0.0316)     (0.0509)    (0.0344)     (0.0151)   (0.0359)   (0.0447)
  Tenure>=10      0.0197       0.0435      0.0187       0.0110     0.0232    0.0806**
                 (0.0254)     (0.0428)    (0.0294)    (0.00974)   (0.0303)   (0.0360)
District HQ/PC   0.160***    0.223***     0.261***      0.0174    0.292***   0.269***
                 (0.0231)     (0.0516)    (0.0424)     (0.0135)   (0.0456)   (0.0436)
     pcnorm2      -0.0139   -0.0950***   -0.0543***   -0.0155**    -0.0204    0.00840
                 (0.0155)     (0.0268)    (0.0181)    (0.00721)   (0.0187)   (0.0218)
     Province       Yes          Yes         Yes         Yes         Yes        Yes
        Year        Yes          Yes         Yes         Yes         Yes        Yes

  Observations    1183        1052         1181         1183       1181       1181
TABLE A2: POOLED OLS/PROBIT
                       (1)         (2)         (3)          (4)

     VARIABLES       pcdev      shgov_sch   shgov_hlth    undev

     lvpop          0.253***    -0.0232**   0.0249***    -0.0527**
                     (0.0356)   (0.00948)   (0.00408)     (0.0240)
     pjhs           1.140***    -0.203***   0.0526***    -0.319***
                      (0.169)    (0.0447)    (0.0193)      (0.117)
     pop11             -0.114      0.0577      0.0192     0.311***
                      (0.163)    (0.0431)    (0.0186)      (0.114)
     headshs         0.165**      0.0307*    -0.00106     -0.0868*
                     (0.0666)    (0.0176)   (0.00765)     (0.0457)
     tenure10         0.105*       0.0195    -0.00254      -0.0354
                     (0.0548)    (0.0145)   (0.00628)     (0.0391)
     dhq_pc         0.452***     0.000538    -0.00145      -0.0424
                     (0.0686)    (0.0182)   (0.00780)     (0.0532)
     pcnorm2        -0.120***     0.0109*     0.00426     -0.00891
                     (0.0333)   (0.00664)   (0.00381)     (0.0225)
     Province           Yes         Yes         Yes          Yes
     Year               Yes         Yes         Yes          Yes
     Constant       -2.759***    0.933***    0.521***
                      (0.324)    (0.0866)    (0.0372)

     Observations     1181        1177        1160         1183
     R-squared        0.350       0.392       0.674
PANEL ESTIMATES USING CULTURE
COMPONENTS
                            (1)         (2)        (3)         (4)
        VARIABLES       shgov_hlth   shgov_sch    pcdev       undev

      Population size   0.0237***    -0.0259**   0.263***    -0.277***
                        (0.00429)     (0.0128)    (0.0450)    (0.0980)
         Hh edn>=jhs     0.0516**    -0.204***   1.189***    -1.256***
                         (0.0206)     (0.0604)     (0.213)     (0.464)
   Largest popn group      0.0132      0.0477       -0.121    0.866**
                         (0.0191)     (0.0516)     (0.189)     (0.423)
      Head edn>=shs      -0.00132      0.0265     0.142**     -0.310**
                        (0.00766)     (0.0171)    (0.0659)     (0.153)
         Tenure>=10      -0.00285     0.0318**     0.0794       -0.114
                        (0.00630)     (0.0137)    (0.0534)     (0.138)
              dhq_pc    -0.000914     -0.00524   0.330***      -0.0658
                        (0.00781)     (0.0175)    (0.0677)     (0.202)
                  sea    -0.00198      0.0152      0.0300      0.234*
                        (0.00725)     (0.0172)    (0.0654)     (0.142)
                rural     0.00404     -0.00241     -0.130*     -0.274*
                        (0.00779)     (0.0214)    (0.0777)     (0.166)
                adat1    0.000628      0.0291     -0.141**      -0.113
                        (0.00651)     (0.0197)    (0.0692)     (0.146)
                islam     0.0161*      0.0310       -0.109   0.671***
                        (0.00953)     (0.0253)    (0.0929)     (0.234)
            Province        Yes          Yes         Yes         Yes
               Year         Yes          Yes         Yes         Yes
            Constant     0.521***     0.929***   -2.643***      0.458
                         (0.0399)      (0.118)     (0.415)     (0.909)
             Lnsig2u                                         -1.075***
                                                               (0.372)