2 Summary of the Paper
2.1 Conceptual Framework
2.1.1 Historical Background
2.1.2 Theoretical Argument
2.2 Empirical Framework
2.2.5 Alternative Explanations
3 Critical Discussion
3.1 Pareto Estimator
3.2 RDiD Design
With (according to the 2010 Indonesian Census) 87 % of its 264 million inhabi- tants being Muslims, Indonesia is the largest Muslim country and the third largest democracy in the world. The modern history of this Southeast Asian country is shaped by an era of Dutch colonization followed by a period of authoritarian rule in the 20th century, until the current democratic system was established in 1999. This thesis discusses a paper about a natural historical experiment in Indonesia looking at the long-run eﬀects of Islamic institutions on political Islamism. The authors use the deﬁnition of Islamism as a political movement which relies on fundamentalist Islamic values that it tries to implement into modern social and political life (Euben and Zaman, 2009, p.4). The paper adds to the current research about the economics of religion which considers religion as an independent variable and looks at the in- ﬂuence of religion and its cultural implications on society (Iyer, 2016, pp.396-397), in this case the question how political outcomes are related to religion and religious institutions (Bazzi et al., 2018, p.1).
The foundation of the natural experiment is the Basic Agrarian Law (BAL) from 1960 during the era of president Sukarno. The goal of this reform was to reduce land inequality through expropriations of private landholdings. Nonetheless, religious properties were excluded, which had the eﬀect that more people endowed Islamic lands called awqaf (singular: waqf). These properties are inalienable and used as sites for mosques and Islamic schools. The reform failed and was later reversed, but the newly created awqaf and the institutions located on them remained due to their religious status. Bazzi et al. argue that this has had a long-run eﬀect on the support for political Islamism, since especially the Islamic schools inﬂuence local political views. (Bazzi et al., 2018, pp.2-3).
To test their hypotheses empirically, the authors apply an ”RDiD” design (Grembi et al., 2016, pp.9-13), which combines a Diﬀerence-in-Diﬀerences approach with a Regression Discontinuity design. For the DiD, Bazzi et al. look at whether the number of landholdings that exceeded a certain size (so that they were subject to expropriation according to the BAL) in a district is above or below the sample me- dian. For the RD, the authors use a population density cutoﬀ speciﬁed by the reform above which more landholdings of a district were targeted (Bazzi et al., 2018, p.2).
The results show that the number of lands under waqf and the Islamic institu- tions situated on them indeed increased as a reaction to the land reform. Apart from the positive long-run eﬀect on the support for political Islamism in districts which were targeted more intensely by the BAL, measured by looking at the voting results of elections in 1999 and 2009, the empirical analysis reveals a higher importance of candidates’ religion and religiosity as well as an increase in sharia (Islamic law) regulations in these regions (Bazzi et al., 2018, pp.2-4).
I divide my thesis in two parts. Firstly, I summarize the paper by Bazzi et al. and then I continue with a critical discussion. The summary consists of a section for the conceptual framework, in which I explain the historical background and the logical steps for the theoretical argument, and a section for the empirical framework, in which I present the data sets of the empirical analysis, the model, an illustration of important points, the results and possible alternative explanations for the observed eﬀects. For the discussion, I look at two mathematical methods used in the paper in more detail.
2 Summary of the Paper
2.1 Conceptual Framework
I start by explaining the theoretical argument that the authors put forward in their paper. In order to understand the argumentative steps of Bazzi et al., it is necessary to know some historical facts which I present in the following section.
2.1.1 Historical Background
The Sukarno regime (1945-1967) followed the political concept of Nasakom (an acronym of the Indonesian words for nationalism, religion and communism) (Cribb, 2001, p.228). Following a partially communist ideology, it was an important goal of the government to give more power to poor households and reduce inequality, speciﬁcally land inequality, which was quite pronounced in Indonesia (Bazzi et al., 2018, pp.5-6). There were some districts such as Java and Bali where 60 % of cultivators did not own any landholdings (Soemardjan, 1962, p.26).
The reform speciﬁed rules for minimum and maximum sizes of private land- holdings (varying depending on the population density of a region), distinguishing between ”dry” and ”wet” land (the latter referred to irrigated land mainly used for rice cultivation, while other properties were considered ”dry”). The government did not want to oppose the Islamic religious elites, especially on the basis of the historically established fear of seizing sacred land (Abbasi, 2012, p.128). Therefore, religiously-held awqaf were not subject to expropriation (Bazzi et al., 2018, pp.6-7).
A waqf is an area of land declared as holy and inalienable. It is meant to last eternally and serve the community (Kuran, 2001, p.842). In Indonesia, the ﬁrst awqaf were founded in the 1500s, but they were not fully legally accepted by the Dutch colonial administration (Bazzi et al., 2018, p.9). The institution became more popular in the 20th century, especially in the 1930s and the Sukarno era after 1945 (Djatnika, 1985, p.29). Most houses of worship (mosques, prayer rooms) and Islamic p esantren schools in Indonesia are institutions located on waqf land. In other Muslim countries, especially in the Middle East, these lands are used more often for economic purposes and cover a larger fraction of total land area (Bazzi et al., 2018, p.9).
Even though anyone can endow a waqf , it is an institution of the aﬄuent class, as the operating costs and the salary of the administrator have to be paid by the founder. Supplying a waqf is beneﬁcial for the reputation of the donor because it is seen as a pious act. He hopes to secure himself a better position in the afterlife and to better his relationship with God. Furthermore, public opinion can be inﬂuenced and therefore political power can be gained by having control over the contents taught at educational institutions located on waqf land. Another motivation to endow a waqf is to protect land from expropriation or taxation, as was already done earlier in history before the BAL (Kuran, 2001, pp.853-855).
The creation of these religious properties at the time of the BAL was fueled by complementary interests of the elites, who feared losing their properties and the Islamists, who were concerned about the Communists gaining more political power and the possibility that they might overthrow the government. The Islamists also needed the landowners since they were important for funding religious institutions (Utrecht, 1969, pp.83-85). The relationship between local elites and Islamists was strengthened in districts where more awaqf were endowed (Bazzi et al., 2018, p.9).
Implementing the land reform turned out to be diﬃcult. The local redistribution committees were in action from 1962, however, they included representatives of the local elite who supported the land owners and thus slowed down the expropriation process. As a response, vigilante groups associated with the peasant movement be- gan conﬁscating landholdings on their own, causing an increase in conﬂict (Utrecht, 1969, pp.80-82).
A mass violence against leftists and supporters of Sukarno broke out after a failed Communist coup in 1965 (Cribb, 2001, pp.231-233). Consequently, the land reform stopped and landowners retook their properties, except for the awqaf , since the legal status of these religiously-held lands could not be changed due to their inalienability (Bazzi et al., 2018, p.7). The land reform was mostly undone by the late 1960s (Utrecht, 1969, pp.86-87).
2.1.2 Theoretical Argument
The argumentation of Bazzi et al. consists of three elements: they explain why today more awqaf , more Islamic institutions and more support for political Islamism should be observable in districts where the land reform was likely to be binding, i.e. districts with high population density and many large landholdings.
The ﬁrst argumentative step is that if more awqaf were founded in these districts (which should be the case due to the strong incentives) and awqaf are inalienable, then more of these religious landholdings should exist today in targeted districts (Bazzi et al., 2018, p.10).
For the second step Bazzi et al. argue that those districts should feature more Islamic institutions that are commonly built on waqf land such as mosques and p esantren (Islamic schools). Other types of Islamic practice that do not require a waqf to function, e.g. zakat (Islamic alms-giving) should show no change in popularity (Bazzi et al., 2018, pp.11-12).
In order to argue for the third step, it is necessary to explain p esantren in more detail. These boarding schools play a role in inﬂuencing political Islamism in In- donesia. Their curricula are often based on the fundamentalist Islamic ideology of Salaﬁsm with a strong focus on Islamic studies, while not under governmental control. The schools often support Islamist parties and keep a close relationship to them, which gives the Islamists more local inﬂuence (Sukmajati, 2011, pp.85-87). For example, in Tasikmalaya (which is among the districts with the greatest density of p esantren in Indonesia) the Islamist United Development Party PPP gained a vote share of 41 % in 1999, the national average being 10 % (Bazzi et al., 2018, p.12). Islamist political leaders often come from p esantren schools (e.g. Hidayat Nur Wahid, former leader of the Prosperous Justice Party PKS) (Bazzi et al., 2018, p.10). Islamic groups of former p esantren students at universities were important for the Islamists as a means of organizing their activities during the repressive period under president Suharto from 1967 to 1998 (Machmudi, 2008, pp.38-39). Pesantren provide public goods and play a role in community-based activism (Hamayotsu, 2011, pp.983-985). Additionally, p esantren sometimes have own militias to support Islamist parties logistically and politically at the time of elections (Buehler, 2016, p.136).
Bazzi et al. construct the third argumentative step by concluding that given the inﬂuence of p esantren on politics, Islamist parties, policies and values should be more dominant today in the regions where more awqaf were founded as a reaction to the BAL (Bazzi et al., 2018, p.12).
2.2 Empirical Framework
The following sections provide information about the empirical approach of Bazzi et al., which uses a regression model combining the Diﬀerence-in-Diﬀerences and the Regression Discontinuity design.
The data sources for the empirical analysis are the following (Bazzi et al., 2018, pp.13-15) :
Demographic Data: To ﬁnd out the population densities of districts in 1960, the authors use the 1961 population census for the population numbers, while they calculate land area ﬁgures manually based on the boundaries of 188 historic districts that are left of 202 districts after taking into account the other available data sources.
Landholdings: The number of large landholdings per district is available in the 1963 Agricultural Census. However, the distribution of lands is only available for properties considered small in the following speciﬁcations of the model (< 5 ha). Thus, the authors estimate the distribution for large landholdings using a Pareto distribution (Saez and Zucman, 2016, pp.540-541) with the estimator λ.
Islamic Institutions: The amount and fraction of land under waqf and the num- ber of p esantren , mosques and zakat groups are taken from the 2003 Village Potential Statistics (Podes ) Survey, out of which 55 000 of 69 000 villages are left after link- ing with the 1963 Agricultural Census data. The Muslim population share in each village comes from the 2000 population census.
Political Islamism: The 2003 Podes includes the performance of Islamic and Is- lamist parties at village level for the 1999 election, but only the ﬁrst three ranks in each village, not the vote shares. Those are only available at district level in the na- tional legislative election data from 1955 to 2009. The authors categorize parties into Islamic and Islamist according to established political science literature (Baswedan, 2004, pp.679-680). The Indonesian Family Life Surveys from 2007 and 2014/15 in- clude preferences about the religion and religiosity of candidates. With a data set from Buehler (2016, pp.215-217) spanning over 443 regulations across 133 districts, Bazzi et al. measure sharia regulations implemented by district governments from 1998 to 2013.
Other Outcomes: For additional issues that I cover in 2.2.5, the authors use data from the 2003 Agricultural Census, the 2003 Podes , the Indonesian National Socio- Economic Survey ( Susenas ) from 1994 to 2011 and the 2000 Population Census.
The model combines a Regression Discontinuity design and a Diﬀerence-in-Diﬀerences approach. Since the reform used three cutoﬀs for the population density of a region (50, 250 and 400 people/km2) but the data set is not large enough for an RD design covering all thresholds, the authors focus on the 400 cutoﬀ. For grouping the data into two groups for the RD only the 50 and 400 thresholds are applicable, of which the 400 cutoﬀ makes more sense to use. It is the only relevant threshold on Java, the most populous island of Indonesia and the main targeted region of the BAL (Huizer, 1972, pp.32-33), where only 7 out of 81 districts had a population density below 250 people/km2 (Bazzi et al., 2018, pp.15-16). Therefore, the landholdings which are referred to as large in the model range from 5 ha to 7.5 ha (wet) and 6 ha to 9 ha (dry), which are the regulations for districts above the 400 cutoﬀ. Landholdings in this range were subject to expropriation, whereas larger landholdings were already aﬀected by the reform at lower population density thresholds. Bazzi et al. consider a district to have many large landholdings if the number is above the sample median (Bazzi et al., 2018, p.16).
The model is speciﬁed as follows (Bazzi et al., 2018, pp.16-18):
Abbildung in dieser Leseprobe nicht enthalten
This equation is used for all the main regressions in the paper. y ij takes the role of several variables, e.g. waqf land in 2003 or Islamist party support and the in- dexes are ”i” for villages and ”j” for districts. Abov e 400 j is a dummy for districts above the 400 threshold (the equivalent of the treatment dummy in a standard RD design). The dummy LH j represents districts above the sample median in the number of large landholdings at the onset of the land reform (the equivalent of the interacted variable in a DiD approach). g (D j) denotes a third-order polynomial term describing population density on both sides of the 400 threshold (term with η 1: below; term with η 2: above), interacted with LH j (D j is the equivalent of the running variable in a standard RD). X ij is a vector of control variables (total male and female populations, total amount of farms, irrigated land, agricultural land, geographic controls, vote share of Islamic parties in the 1955 and 1957 elections and six island ﬁxed-eﬀects). The controls are either measured prior to the reform or are time-invariant. The vector is interacted with Abov e 400 j and LH j (denoted by Ω). β is the most important coeﬃcient, it refers to districts with many large landholdings and a population density above the 400 cutoﬀ - the ”treatment group”.