Juni Singh


There is a large literature on the labor market and educational returns to military service. In most countries, the military is a unique employer, not just because of its size (the military is typically one of the largest employers), but also because the military tends to recruit disproportionally more from disadvantaged groups, it offers extensive training on the job, and it requires its employees to live away from their home and family. Hence, voluntary military service in professional armed forces could expose disadvantaged recruits to new ideas and skills, which in turn could improve the economic opportunities of theirs and their families. In the context of the US armed forces, evidence suggests that recruits from disadvantaged groups benefit mildly from voluntary service in terms of earnings (Angrist, 1998). However, very little evidence is available on how these findings translate to a developing country context. Nevertheless, the returns to military service in professional armies could be much higher in developing countries, as the average human capital in their populations tends to be lower. In a recent paper, Vanden Eynde (2016) shows how voluntary war time service in the colonial Indian Army raised literacy rates in heavily recruited communities. In spite of mild evidence of intergenerational transfers, this paper cannot assess long term impacts because of the dramatic population movements that took place during India’s Partition. Our project tries to estimate the long-term effects of military recruitment in a country that offers a particularly interesting environment to assess the long-term impacts of military service: Nepal.


      The British colonizers were particularly impressed by the fighting skills of the so-called “Gurkha’s” during the Gorkha war (1814-1816).[1] As a result, 200 years ago, the East India Company started to recruit soldiers from a large region that is now part of Nepal. The Gurkha soldiers provided many regiments to the colonial Indian Army, and they were heavily relied on during the two World Wars and more recently in the Falklands war or in UN peacekeeping operations. After India’s independence, both the British and the Indian army keep on recruiting soldiers from Nepal until the present days. Currently, 25,000 young Nepalese hill-men apply for not more than 250 positions each year. The transformational role that Gurkha soldiers played in the development of their villages is well documented by historians and sociologists (e.g., Kergoat, 2008, p.273), who often suggest that they are particularly instrumental in improving the educational facilities in their communities. Our paper aims to provide the first quantitative evaluation of the hypotheses that Gurkha recruitment contributed to the development of rural Nepal. Our empirical strategy, which we will develop in more detail below, relies on the fact that recruitment patterns were first established at the end of the 19th century but are likely to have remained persistent ever since, as the networks of existing soldiers were used to find new recruits.[2] A key constraint that gave rise to concentrated recruitment at the end of the 19th century could have been the reliance on a limited number of recruitment centers, each responsible for the recruitment of certain tribal groups. We propose to use the distance to these recruitment centers as an instrument for the intensity of Gurkha recruitment, in order to identify the impact of Gurkha recruitment on economic and educational outcomes.


      While our project naturally contributes to the literature on military service described above, we will also contribute to the literature on the long-term impact of migration on sending countries (Chauvet and Mercier, 2014). Moreover, our work will help to understand the persistent effects of military recruitment during the colonial period, adding to work about the long-term impact of colonial institutions (e.g. Huillery, 2009; Dell, 2010).


     Data and empirical approach: Our project will use data from the Nepalese census (1991, 2001, and 2011) and the Nepal Living Standard Survey (1995-1996, 2003-2004, and 2010-2011) to measure education, income, and public good provision at the level of individuals and villages (“VDCs”). I have already worked with this data for existing papers (Baland et al., 2015, Libois, 2015).


      Our suggested instrument will use distance to the 19th century recruitment centers for the relevant tribes. We have already located the 6 recruitment centers on GIS maps (see map in appendix).


      To test our first stage, we require information about the recruitment intensity in Nepalese villages. We suggest measuring the exposure of villages to military recruitment from different sources. First, we can use the distribution of World War casualties as a proxy for recruitment intensities, as in Vanden Eynde (2016) and Jha and Wilkinson (2015).[3] We cleaned the casualty data and have now extracted village names for almost all casualties. A second potential source of recruitment estimates are administrative sources from the British administration. We have identified historical sources that can be consulted in the British Library (London) that will probably contain lists of heavily recruited villages.


      Empirical approach included instrument, potentially other criteria imposed by the Brits as an IV (see VandenEynde 2016) and potentially exploit changes in the British recruitment rules, especially concerning the minimal educational background of applicants and their ethnic membership, to estimate the aspirational effect of recruitment based on difference-in-difference between cohorts.


     Preliminary results: As preliminary strategy, we use the location of the two recruitment centers and their outposts created by the British in 1885 (see map in the appendix) and correlate them with measures of education at the household level. The creation of these centers marks a turning point in Gurkha recruitment since Nepal and the British started to cooperate to ease the recruitment process (Rathaur, 2001). Only four ethnic groups of the Hills were recruited by the British: Magars, Gurungs, Rais and Limbus.


      Our hypothesis is that the recruitment in these specific communities has long term persistent effects on education and still has some direct influence because past recruitment hugely determines present recruitment, despite the displacement of the new recruitment centers (in Katmandu and Pokhara).


      Our preliminary results hold in three tables and are all organized in the same way. The explanatory variables of interest are (1) the interaction between a variable identifying household belonging to eligible ethnic groups (the “Gurkha” variable”) and the distance to the closest recruitment center of 1885 and (2) the share of eligible households in a village and the distance to the closest recruitment center. The first variable intend to capture benefit limited to the historically eligible groups, while the second intend to capture public good type effects. We control for Gurkha eligibility at the household or at the village level, distance to the closest recruitment center in 1885, distance to the Southern border of Nepal and distance to Kathmandu. In all regressions, we also include time and belt-zone (see figure 2) fixed effects to clean our estimations from South-North or East-West effects.

We find that the closest to the historical location of a recruitment center an eligible household lives, the highest is the average education of its teenagers. The first column of table 1 indicates that if an eligible household lives 100km further away from a recruitment center, one’s teenagers loose, on average, 0.4 years of education. The second column of table 1 is consistent with the first, but shows that the effect might be driven by the share of eligible households living in the village. This suspicion is confirmed by the third column. All households living in a village with a large proportion of eligible households have a higher level of education if they live closer to a recruitment center.


      The story is less sharp while looking at the average level of education of adults (column 4-6 in table 1), nevertheless average education of adult might suffer more from compositional effect. In table 2, we show that our results hold if we focus on the level of education achieved by the most educated member of the household. Consistently with the story of path dependence in the recruitment, columns 4-6 in table 3 show that eligible households do receive more income from pensions today if they live closer to an historical recruitment center, and less if they are far away. This is not the case for non-eligible households.

One consistent story with our preliminary findings is that former Gurkha soldiers have higher income and use part of it to finance public goods helping all inhabitants of their village to achieve higher level of education. The Gurkha recruitment would have positive spillovers on their immediate neighbors.


      Notice that our results are robust if we drop one of the historical recruitment centers. Point estimates are however slightly smaller. This is particularly reassuring in the case of Tribeni, the recruitment center based within Nepal because is it relatively close to Pokhara (75km).

[1] The Gorkha Kingdom comprised of a large area in the west of present-day Nepal.

[2] VandenEynde (2016) provides evidence of such persisitence for military recruitment in colonial Punjab.

[3] Both papers show that the distribution of casualties matches the distribution of recruitment very well, suggesting that mortality was not selective in terms of the region of origin.

Full Text:


The Himalayan Research Papers Archive is hosted by the
UNM University Libraries
MSC05 3020, Albuquerque, NM 87131, (505) 277-9100