«Abstract This paper estimates the causal impact of tertiary education on economic prosper- ity. To address the potential endogeneity of tertiary ...»
Tertiary Education and Prosperity: Catholic Missionaries to
Luminosity in India∗
Amparo Castell´-Climent,† Latika Chaudhary‡ and Abhiroop Mukhopadhyay‡
† ‡ ‡
University of Valencia Naval Postgraduate School Indian Statistical Institute
This paper estimates the causal impact of tertiary education on economic prosper-
ity. To address the potential endogeneity of tertiary education, we use the historical location of Catholic missionaries across Indian districts as an instrument. The choice of India is an important part of our identiﬁcation strategy because Christians con- stitute a small fraction of the population. Catholic missionaries, however, strongly impacted tertiary education but had a relatively modest inﬂuence on other institu- tions. Catholics were not at the forefront of tertiary education in colonial India. They established many high quality colleges following Indian independence. Controlling for a rich set of geographical and historical characteristics, we ﬁnd a positive causal eﬀect of tertiary education on development, as measured by light density at night. The ﬁndings are robust to diﬀerent measures of development, and they are not driven by alternative channels through which missionaries could impact current income.
JEL classiﬁcation: I25, N35,O15 Key words: Human Capital, Catholic Missionaries, Subregional Analysis ∗ We would like to thank participants at the Xth Annual Conference on Economic Growth and De- velopment, World Economic History Congress, Stanford Institute for Theoretical Economics Conference, ANAECO Workshop, and seminar participants at UNSW(Sydney), Monash University (Melbourne), Delhi School of Economics and CMI (Bergen) for their positive feedback. We are grateful to Federico Manto- vanelli for sharing the historical maps, and to Daron Acemoglu and Romain Wacziarg for comments. This work would not have been possible without the eﬀort put in by Athisii Kayina. We acknowledge the ﬁnancial support from the Spanish Ministry of Economy and Competitiveness through ECO2011-29283 project, and the Planning and Policy Research Unit at the Indian Statistical Institute (Delhi). The views expressed in this article are those of the authors and do not reﬂect the oﬃcial policy or position of the Department of Defense or the U.S. Government.
1 1 Introduction Identifying the fundamental determinants of development has a long pedigree in economics.
A large literature relying largelyon cross-country variation has emphasized the role of institutions (Acemoglu et al., 2001, 2002, 2014), geography (Sachs 2003), openness to trade (Frankel and Romer 1999), and human capital (Lucas 1998). Among these factors, the empirical evidence linking education to income has produced perhaps the weakest ﬁndings at the macro level (Benhabib and Spiegel 1994). The lack of a robust relationship between education and income is at odds with the vast labor literature, which ﬁnds strong causal eﬀects of each additional year of schooling on individual earnings (Card 2001). How do we reconcile the two sets of ﬁndings?
One explanation is perhaps that the macro literature has relied on incorrect measures of education. Most of the literature uses average years of schooling to capture education diﬀerences across countries (see Benhabib and Spiegel 1994; Cohen and Soto 2007; de la Fuente and Dom´nech 2007). The large number of people with no education skew average e years of schooling for poor countries. Moreover, primary education is often of poor quality in these countries (Chaudhury et al. 2006), further exacerbating measurement problems.
This may explain why years of education correlate poorly with economic outcomes at the macro level (Pritchett 2001). Another mutually nonexclusive explanation is the focus on cross-country analysis. The vast diﬀerences in culture, institutions and access to technology make it diﬃcult to identify the causal eﬀect of education on income. Problems of omitted variables and reverse causality plague many of the empirical studies (Acemoglu, Gallego and Robinson, 2014).
In this paper, we study the impact of tertiary education on development using data on Indian districts in 2006. The focus on a single country minimizes concerns of omitted variables because these sub-national units at least share common governance and national policies. Our focus on districts, an administrative unit below states, thus allows for tight comparisons because we exploit diﬀerences across districts within the same state using state ﬁxed eﬀects. This strategy allows us to address Acemoglu et al.’s (2014) concerns that empirical studies treating institutions and human capital as exogenous are misspeciﬁed.
Since institutions are mainly determined at the central and state level, the advantage of our approach is we can account for institutions without the need of an additional instrument.
However, district-level data pose one problem in the Indian context - current income levels are not well measured. We address this shortcoming by using night lights data 2 as a proxy for income, in line with the recent literature (see Henderson, Storeygard and Weil 2012; Micholapoulos and Papaioannou 2013, 2014; Alesina, Michalopoulous and Papaioannou 2015). We rely on information collected by the National Geophysical Data Center (NGDC) on the location of night lights between 8pm and 10pm, as captured by satellites of the United States Air Force Defense Meteorological Satellite Program (DMSP).
Observations are available for an area of one squared kilometer and can be aggregated to the district-level. To measure human capital, we focus on the share of the adult population with tertiary education as only higher levels of education appear to be correlated with economic growth in India (Castell´-Climent and Mukhopadhyay 2013).
o We ﬁnd a strong positive association between the share of the population with tertiary education and light density at night in an OLS model that controls for state ﬁxed eﬀects. This relationship is robust to a rich array of factors that may jointly inﬂuence tertiary education and luminosity such as current population, population shares of socially disadvantaged groups, geographical characteristics, and historical variables that control for initial conditions. The potential endogeneity of tertiary education, however, poses an empirical challenge because tertiary education and the evolution of income generally go hand in hand. We address this concern by using exogenous variation generated by the location of Catholic missionaries in the early 20th century to instrument for current tertiary education. Using the ﬁrst edition of the Atlas Hierarchicus, we extract the exact geographical location of Catholic missionaries in 1911 and overlay the historical maps on district borders as of 2001 using a geographic information system (GIS) program.
The location of Catholic missionaries has to satisfy two conditions to be a valid instrument. First, the location of Catholic missionaries has to be unrelated to any factor that may impact the subsequent development of districts other than through current tertiary education. If Catholic missionaries historically located in richer and more educated districts, then the instrument would be invalid. Our reading of the history allays some of these fears. Catholic missionaries arrived in large numbers in India with European traders beginning in the 16th century. The ﬁrst wave of Catholic missionaries accompanied Portuguese traders and they located in Portuguese settlements along the coast such as Goa, Daman and Diu. Over time the missionaries moved inland and set up missions in South India and beyond depending on their location preferences. Historical accounts suggest the individual preferences of missionaries were important. They also considered places where the potential for proselytization was high and the hospitality of the local people or their rulers more likely. In some cases this lead missionaries to focus on bringing Hindu upper 3 caste Brahmins into the fold; in other instances, it lead them to focus on areas with a large tribal population. Settling in richer or economically developed areas seemed to be of minor concern.
This account is supported by an empirical analysis of the historical data. Using a sample of districts from British India, we ﬁnd no positive correlation between the location of Catholic missionaries in 1911 and the historical provision of education in 1901. We also ﬁnd no signiﬁcant correlation between Catholic missionaries and measures of wealth such as income tax revenues. Thus, it does not appear Catholic missionaries located in richer districts with a higher potential to grow and develop. Consistent with the narrative above, we ﬁnd that missionaries were more likely to locate in districts along the coast, in districts connected to a railway line and in districts with a larger share of tribal groups.
In our instrumental variables speciﬁcation, we include many geographical and historical controls to address any potential selection problems regarding the location of Catholic missionaries. First, geographic variables such as latitude, longitude, average height, average river length, and a coastal district indicator ensure we control for any positive selection in the choice of location vis-`-vis geography. Second, as Catholic missionaries may have a chosen to locate in areas with more Brahmins (the traditional upper caste among Hindus) or more urbanized districts, we also include the historical composition of Brahmins and the urban population share of the district. Third, we include an indicator if a railway passed through the district in 1909 and the share of historically disadvantaged tribal groups, as they could have a direct inﬂuence on subsequent development. Finally, we include an indicator for districts that were historically part of Princely India and under the rule of hereditary kings as opposed to under direct British rule (British India) because Iyer (2010) shows that districts under Princely rule have better economic outcomes after independence. Including a rich set of controls derived from a detailed reading and analysis of the history ensures that the impact of historical Catholic missionaries is not confounded with location characteristics that may independently impact current development other than through current tertiary education.
To complete the story, however, it is important to discuss the link between Catholic missionaries and education. Indeed, in the colonial period, Catholics were less involved in the provision of education compared to the state, or even Protestants, who arrived later in the late 18th and 19th century. For example, there were only 9 Catholic colleges (5%) compared to 40 Protestant colleges (33%) as of 1911. The remaining colleges were either public or under private Indian management. Catholics were also largely absent from 4 primary education. Early Catholic eﬀorts were focused on training Indian priests and nuns, and on converting Indians to Christianity. The latter attempts had limited success given the small number of Indian Christians (less than 2% of the total population by 1911). This point is further supported by the historical data where we ﬁnd no signiﬁcant positive correlation between the location of Catholic missionaries in 1911 and the number of schools or colleges in 1901. While Catholic missionary location does not correlate with education in the colonial period, we ﬁnd a large, positive and statistically signiﬁcant impact of Catholic missionaries circa 1911 on the share of the 2001 adult population with tertiary education. One mechanism driving this correlation is the establishment of many Catholic colleges after Indian independence. The historical network of Catholic missionaries was a natural platform from where Catholic inﬂuence on education radiated out. This is borne out in regressions, which ﬁnd a strong positive correlation between missionary location in 1911 and the stock of Catholic colleges in the beginning of the 21st century. Qualitative evidence suggests stronger funding from the Vatican, better coordination among the diﬀerent Catholic groups and Indian independence in 1947 that guaranteed protection to minority institutions in the new constitution1 were collectively responsible for the increase.2 The strong correlation between the district tertiary education share in 2001 and the presence of Catholic missionaries in 1911 highlights that our instrument meets the second condition for being a valid instrument. The tertiary education share is 1.2 percentage points higher in districts with a Catholic mission. This is a substantial eﬀect on the order of 20% given mean tertiary education (5.8%). Using Catholic missionaries as an instrument, the second stage results ﬁnd a positive and statistically signiﬁcant eﬀect of tertiary education on current income as proxied by light density at night. A one standard deviation increase in the tertiary population increases log light density by 0.38, an economic eﬀect of 9% given mean log light density of 4.24.
Although we ﬁnd signiﬁcant and positive IV estimates on tertiary education, one may still be concerned about Catholic missionaries impacting current income via non-education 1 Article 30(1) of the Constitution of India gives linguistic and religious minorities a fundamental right to establish and administer educational institutions of their choice.
2 The end of British colonization also marked a water shed for the growth of Catholic educational institutions as the British government followed an active strategy of discouraging missionary inﬂuence on the provision of education in India. Moreover, an early emphasis on Indianizing the clergy generated a large pool of Indian priests and nuns giving the Catholic church a unique advantage in post-independence India (Frykenberg 2008).