WPS7965


Policy Research Working Paper                        7965




                             Push and Pull
         A Study of International Migration from Nepal

                                Maheshwor Shrestha




Social Protection and Labor Global Practice Group
February 2017
Policy Research Working Paper 7965


  Abstract
 This paper studies migration choices in the presence of                             that increases household income by US$ 100 increases
 liquidity constraints and varying costs of migration. This                          migration to India by 54 percent but has no effect on
 paper presents a simple theoretical framework that analyzes                         migration elsewhere. An increase in conflict, which reduces
 migration response to both push and pull factors in such                            consumption and amenity of the wealthier more, increases
 settings. This framework implies that a shock to the push                           migration abroad, especially from urban areas. An increase
 factors in the origin leads to differential observed response to                    in demand from the destination countries, especially the
 migration to various destinations, as the shocks affect differ-                     Gulf countries and Malaysia, has strong effects on migra-
 ent parts of the wealth distribution. The implications of this                      tion to those destinations. These findings are consistent
 framework are tested in the context of international migra-                         with the theoretical framework, and suggest the presence
 tion from Nepal, using a panel of 452 villages observed at                          of large liquidity constraints. An increase in income can
 three periods in the 2000s. The analysis uses rainfall shocks                       boost migration to India, whereas a reduction in the cost
 and deaths due to conflict as “push” shocks and growth in                           of migration might increase profitable migration elsewhere.
 manufacturing and construction in destination countries                             The responsiveness to “pull” shocks suggests that house-
 as “pull” shocks. The findings show that a rainfall shock                           holds are willing to take advantage of these opportunities.




  This paper is a product of the Social Protection and Labor Global Practice Group. It is part of a larger effort by the World
  Bank to provide open access to its research and make a contribution to development policy discussions around the world.
  Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted
  at mshrestha1@worldbank.org.




          The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
          issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
          names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
          of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
          its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                        Produced by the Research Support Team
   Push and pull: A study of international migration from Nepal
                                         Maheshwor Shrestha∗




JEL codes: F22, O15, O12, J61
Keywords: Migration, Conflict, Nepal
  ∗
    The World Bank. Email: mshrestha1@worldbank.org. I am extremely grateful to my PhD advisers Esther
Duflo, Abhijit Banerjee, and Benjamin Olken for their advice and guidance. I also thank the participants at the
MIT development lunch for helpful discussions and feedback. This paper was written as part of the dissertation
submitted to the Department of Economics at MIT. The views expressed in this paper do not necessarily reflect
the views of The World Bank or MIT.
1       Introduction
International labor migration has surged in recent years with an increasing number of workers
moving across international borders for work. The stock of international migrants increased from
154 million in 1990 to 244 million in 2015 (UNDESA, 2015). Over 70 percent of international
migrants are from developing countries. However, most of the migrants from developing countries
migrate to other developing countries, though the returns to migration may be higher in the
developed countries. Of the migrants that originate in the least developed countries, only 16
percent migrate to the global North. High costs of migration to and low demand for migrant
workers in the developed countries potentially prevent migrants from migrating there.
    In this paper, I investigate the migration response to the ‘push’ and ‘pull’ shocks, shocks that
affect the origin and destination respectively, when destinations have varying costs of migration
and individuals are liquidity constrained. For instance, a positive income shock in the origin
may increase migration by making it more affordable or it may reduce migration by making it
less desirable. The net effect on migration rates depends upon the distribution of households
around the affordability and desirability margins. I outline a simple theoretical framework that
captures these features and test its implications in the context of international migration from
Nepal.
    Several features make Nepal a good setting to study the effect of shocks on international
migration. First, migration of workers from Nepal is large and crucial to the national economy.
Though the historic migration rates are similar to the global average, by 2011 the migrant to
population ratio increased to 7.4 percent, more than double the global average (Table 1). Conse-
quently, foreign remittances became one of the largest sources of national income, contributing
over a fifth to the national GDP1 . Second, most work migrants from Nepal choose between two
distinct types of destinations that differ in terms of the costs and returns to migration. As seen
in Table 1, India has been one of the key destinations for Nepali workers for the past several
decades and continues to be the largest destination country. It serves as the low cost and low
return destination for Nepali workers. However, the surge in migration in the 2000s was driven
by migration to Malaysia and the Persian Gulf countries (especially Qatar, Saudi Arabia and
the United Arab Emirates). Most of the migrants, about 90 percent of those who migrate for
work, migrate to these destinations. These countries are the high-cost-high-return destinations
for Nepali workers. Third, migration to these destination countries is facilitated institutionally.
Nepal maintains an open border with India and workers can move across the border with extreme
ease. In addition, agreements with Malaysia and the Persian Gulf countries make it relatively
easy, though costly, for Nepali workers to migrate for (low-skilled) work in those destinations.
Almost 500,000 workers left Nepal in 2014 to work in these destination countries.
    However, the destinations chosen by Nepali workers are not unique to migrants from Nepal.
Nine Persian Gulf countries, which represent 1 percent of the global population, are destinations
to 13 percent of the international migrants in the world (UNDESA, 2015)2 . Along with India and
Malaysia, these destinations account for over 16 percent of international migrants (21 percent of
male migrants) in the world. These countries have become the workplace of low-skilled workers
from the Philippines and South Asian countries who have been migrating to these countries for
    1
     In 2015, this share was 32 percent (The World Bank) making Nepal the country with the largest share of
personal remittance to the GDP
   2
     The nine countries are Bahrain, Israel, Jordan, Kuwait, Lebanon, Oman, Qatar, Saudi Arabia, and the United
Arab Emirates. These are the largest destinations for international migrants in the Persian Gulf excluding the
Syrian Arab Republic.



                                                      2
decades.
    In the context of international migration from Nepal, I investigate the effect of three types
of shocks using a panel of 452 villages across the country and observed in years 2001, 2008 and
2010. The first is the shock to farm income in the origin. Deviation of rainfall from its historic
mean generates an exogenous variation in farm income both across the country for a given year
and between different time periods for the same village. The second shock is a measure of conflict
in origin which creates an income loss as well as utility loss, more so for the richer households.
The ten-year long Maoist insurgency, which killed over 13,000 people, generates a variation in
conflict intensity experienced by different parts of the country. In addition, the dramatic change
in the trajectory of the insurgency gives rise to inter-temporal variation in conflict intensity
experienced by a given village. In addition to the two ‘push’ shocks that affect the origin, I
investigate the effect of an increase in the demand for low-skilled workers from abroad. I use
the growth in the construction and manufacturing sectors in various destination countries to
proxy for these ‘pull’ shocks from the destination countries. Here as well, different destinations
countries that are plausibly in quite different macroeconomic cycles generates the variation that
identifies my estimates. Differential effects of the global recession of 2008, further generates
variation in growth rates across different destination countries.
    Consistent with the theoretical framework, I find that rainfall shocks increases migration
to India, but have no effect on migration to non-India destinations. Increase in farm income
of US$ 100 due to rainfall increases migration to India by 2.5 percentage points (54 percent
increase from its 2010 level) but has no effect on migration elsewhere. Rainfall shocks increases
the income of the poor, who rely on agriculture, and enable them to finance low-cost migration
to India. However, since rainfall shocks generate income gains that are small in magnitude
relative to the costs of migration to other destinations, they do not enable too many households
to migrate outside India. Similarly, an increase in conflict intensity by one death per 1000
population increases international migration from urban areas by 3.1 percentage points which
is equivalent to the effect of increasing household income by US$ 420 in absence of conflict
(after accounting for the average impact of conflict on urban consumption). Conflict reduces the
consumption and amenity of the wealthier more, and therefore, induces a larger effect from richer
(urban) areas. Lastly, I find that migration responds to increase in growth in the construction
and manufacturing sectors in the destination countries, particularly Malaysia and the Persian
Gulf countries. These findings suggest that increase in income can boost migration to India
whereas a reduction in cost of migration might increase profitable migration outside India as
well. Furthermore, the responsiveness to ‘pull’ shocks suggests that households are willing to
take advantage of the profitable opportunities abroad.
    This paper contributes to the literature on the effect of income at origin on migration. The
canonical model of Harris and Todaro (1970) suggests than an increase in income in the origin
reduces migration. Abramitzky, Boustan, and Eriksson (2013) finds evidence consistent with
this using historic migration from Norway. Studies in many other contexts, which use shocks
to weather as a source of exogenous variation in income, also find evidence consistent with
this3 . Examples of these studies include Munshi (2003) for migration from Mexico to the US,
Marchiori, Maystadt, and Schumacher (2012) for migration from Sub-Saharan Africa, and Gray
and Mueller (2012) for migration from rural Ethiopia. However, several other studies find the
   3
     The practice of using weather shocks as a source of exogenous variation in income is quite common throughout
the literature (see Jacoby and Skoufias, 1998; Kazianga and Udry, 2006; Miguel, 2005; Paxson, 1992; Wolpin, 1982,
among many others, for instance).



                                                       3
opposite relationship between income shocks and aggregate level of migration and argue that
liquidity constraints, particularly for poor households, prevent them from migrating. Examples
include Angelucci (2015) for migration from Mexico to the US, Bazzi (2014) for migration
from Indonesia, and Halliday (2006) for migration from El Salvador. In a cross-country study,
Clemens (2014) also finds the positive relationship between income and migration for very poor
countries. My results for migration to India are consistent with the latter studies that find a
positive relationship between income and migration. However, I also find that small income
shocks are unlikely to affect migration to destinations with very large costs of migration.
    This paper also contributes to the literature on the role of conflict on migration. Conflict
could affect migration directly as people want to avoid it, or it could affect migration by lowering
income at origin through its effect on the economy as a whole. Much of the literature on conflict-
associated migration examines the direct channel and argues that the threat of violence is a key
push factor behind such migration. Many studies have found empirical evidence for migration in
response to such threats of violence (see Davenport, Moore, and Poe, 2003; Moore and Shellman,
2004; Melander and Öberg, 2007, for example). Similarly, Naudé (2008) finds an increase in
international migration from Sub-Saharan Africa in response to armed conflict. In a context
similar to this study, Williams and Pradhan (2009) find that conflict related events like violence
and bomb blasts increased out-migration from the Chitwan valley of Nepal. However, these
studies only focus on internally displaced people and refugees and do not consider the effect on
the broader set of households who might choose to migrate even though they are not directly
forced to move.
    Finally, this paper contributes to the relatively scant literature on the role of migrant demand
from the destination on migration. A few examples include: Becker, Musabek, Seitenova, and
Urzhumova (2005) that find sharp migration response from Kazakhstan to macroeconomic shocks
in Russia, Yang (2008) that finds that migrants from the Philippines lengthen their stay when real
wages in the destination increases unexpectedly, and McKenzie, Theoharides, and Yang (2014)
that find more migrants leave the Philippines in response to positive shocks to the destination
country’s economy. Similar to these studies, I find that migration from Nepal responds positively
to growth in the foreign labor intensive sectors in the destination country. Another, related and
consistent evidence in the same context is Shrestha (2017a), where I find high wages due to
exchange rate shocks in the destinations increases migration outflow from Nepal4 .
    This paper stands out from the literature in two key ways. First, it looks at migration to
multiple destinations as a response to the same ‘push’ shock. Most of the literature investigates
migration response to all countries (Bazzi, 2014, for example) or to a specific destination country
(Abramitzky, Boustan, and Eriksson, 2013; Angelucci, 2015, for example). This shortcoming
exists in the literature because international migration is either dominated by migration to one
particular country (for instance, from Mexico to the United States) or because the destinations
are not too different in terms of the associated costs and returns to migration. The context of
Nepal makes it possible to investigate this because the two types of destinations are starkly dif-
ferent in terms of costs and return and the migration flows to these destinations are substantial.
Second, this paper studies the effect of ‘push’ as well as ‘pull’ factors in affecting the migra-
tion decision, an addition to the literature that focuses on either one of the shocks as reviewed
currently.
   4
    Besides the actual shocks at destinations, expectations about wages and conditions in the destinations can
also influence the migration decision. This aspect of influencing perceived migrant demand is investigated in
Shrestha (2017b).



                                                      4
    The remainder of the paper is organized as follows: Section 2 outlines a very simple theo-
retical framework, Section 3 describes the context and the data sources, Section 4 describes the
empirical strategy, Section 5 discusses the results, and Section 6 concludes.


2       Theoretical framework
In this section, I outline a very simple model of household migration decision to illustrate the
key forces that drive my empirical results.
    I assume that all households are farming households and their income depends upon the
realized rainfall shock during plantation season. Total wealth of the household represented
by w + r where w is a measure of permanent wealth, and r represents income from farming.
Households differ from each other only in terms of their permanent wealth which is distributed
F (·) in the population.
    In addition to total wealth, households value the peace p, of their place of residence. Conflict
in the district of residence lowers p and households want to avoid conflict. Consistent with the
nature of conflict (as discussed below), I assume that p is valued more by the richer households.
Formally, in absence of migration, households receive utility V (w + r, p) increasing in both
                                                                                     ∂2V
parameters and that marginal utility of wealth is increasing in peace. That is, ∂w∂p      > 0.
    Households can choose to send some members outside the country. This type of migration
serves two purposes: the migrating member can earn income while abroad, and if there is conflict
in the origin, migration protects the member as well as the rest of the family from the brunt of
conflict. This nature of avoiding conflict is slightly different from the usual conflict associated
migration where entire households move, but is consistent with the nature of conflict in this
context. By sending the productive members (usually the household ‘head’) abroad, households
can stop them from being forcefully recruited to fight, or from them being used as a means to
extort money from the households. If the key income earner migrates, then insurgents would
have a harder time harassing the rest of the family and extorting money from them.
    Migration is costly and the cost C of migration (and finding a job) depends upon the destina-
tion. As discussed in more detail later, there is a large variation in the cost of migration between
different destinations with India being the cheap destination compared to the non-India desti-
nations: Malaysia and the Persian Gulf countries. This cost will determine where the migrant
can and cannot go.
    Once abroad, I assume that migrants always find work. This assumption is different from
most studies in migration which focus on the risk associated with finding jobs in the destina-
tion (Bryan, Chowdhury, and Mobarak, 2014; Munshi, 2003, for example). This assumption is
motivated by the fact that for migration outside India, one needs to have a job lined up before
one can actually migrate abroad. For migration to India, I am assuming that, because of the
historic nature of this migration, migrants will have sizable networks in India, which will help
them find a job (more than 80 percent of migrants to India know someone in the destination
city)5 . Once the individuals migrate, their earning abroad, which depends upon the destination,
is high enough to recuperate the cost of migration. The net gain from a migration episode
increases with the cost of migration to that destination (as seen in Table 2). I assume a simple
parametric form mC for the net income gain from migration with m > 0. Once a member
migrates, the remainder of the household stay back but farm income suffers a loss. That is, the
    5
    Another interpretation is that the search cost of jobs abroad is borne before migration and is embeded in the
cost of migration.


                                                       5
household will have αr of farming income where α < 1. The case α = 1 would represent the case
where the migrant member was a surplus in farming. Formally, a household that has a migrant
abroad receives value U (mC + w + αr, p) which is increasing in both parameters. As discussed
above, I also assume that households are better off with a migrant at time of conflict. That is,
 ∂         ∂
∂p V (·) > ∂p U (·).
    Further, households are liquidity constrained and cannot borrow to finance migration. This
assumption seems extreme as most of the migrants report that they borrowed money, typically
from local moneylenders to finance their migration. However, as borrowers cannot credibly
commit to repay after migration, lenders do not want to lend without a collateral. Borrowers
usually post their house or land, worth more than the size of the loan, as collateral to get a
loan. Hence only richer households with total wealth (liquid or illiquid) greater than the cost of
migration can actually finance migration. That is, households can migrate only if w + r > C .
    Hence, households migrate if and only if

                               U (mC + w + αr, p) > V (w + r, p)

and
                                           w+r >C
Households will choose the destination that has the highest return if they can afford it.
    In this simple framework, increase in total wealth due to income shocks will have two opposing
effects. On one hand, it will enable households to migrate by alleviating liquidity constraint.
On the other hand, increased wealth makes migration less desirable by increasing the utility of
staying at home. Increasing conflict reduces the value of not migrating more than it reduces
the value of migrating. Therefore, households choose to migrate, with richer households more
likely to do so. Similarly, increasing the net benefit of migration mC increases migration. To
obtain specific predictions about the sizes of the effects, I will make simplifying assumptions on
the functional forms.

2.1   Linear functional form
To illustrate the key intuition and generate sharp testable predictions, I assume a linear func-
tional form for the utilities. Specifically, U (·) = mC + w + αr, and V (·) = w + r + wp. With
these functional forms, households migrate if and only if

                                               mC − (1 − α) r
                            A≡C −r <w <                       ≡D
                                                     p
The left end of the wealth distribution of migrants is bound by the affordability threshold. At
this margin are very poor households that cannot afford the cost of migration due to liquidity
constraints even though they may desire to migrate. The right end of the wealth distribution
of migrants is bound by the desirability threshold (for a given level of conflict). At this margin
are the wealthier households that choose not to migrate because the earnings gap between the
destination and the origin is too low and, at peaceful times, they get additional premium of
staying back. This simple framework predicts a stark inverted-U shape between household
wealth and migration rate in the cross-section, a common feature of many models of migration
(Bazzi, 2014; McKenzie and Rapoport, 2007, for instance).
    Furthermore, this simple framework suggests that poorer households migrate to cheaper
destinations (India) and richer households migrate to more expensive destinations. This pattern,

                                                6
in fact, is borne in the data. Figure 1 plots how the migration probability of households changes
with two measures of household wealth, w, separately for India and non-India destinations. The
measures used are the highest level of education in the household (bottom figure) and the value
of durables currently owned by the household that it had acquired it at least two years ago as
the measure of wealth (top figure)6 . Both plots in this figure shows that poor households are
more likely to migrate to India than to other destinations. This figure, particularly the bottom
plot, also shows the inverted U pattern as predicted by the framework. The extremely poor
cannot afford to migrate and the extremely rich do not desire to do so.
    Given the cutoffs above, the total observed migration rate in an economy, or a village,
denoted by M is
                                        M = F (D) − F (A)
for each of the destination types7 .
    The following comparative statics immediately follow:
                                    ∂M             D
                                        = − f (D )    <0
                                    ∂p             p
                                    ∂M             (1 − α)
                                        = − f (D )         + f ( A) ≶ 0
                                     ∂r               p
                                    ∂M C
                                        = f (D ) > 0
                                    ∂m    p
which suggests that the magnitude of the effects depend crucially on the distribution of wealth
and the thresholds.
    The first comparative static suggests that increase in conflict increases migration. This is
driven by the right (desirability) thresholds moving to the right. That is, richer households
migrate because they experience a big negative shock to the peace premium. If the migration
desirability threshold, D, is sufficiently high, then the effect is likely to be higher for the urban
areas than for the rural areas as the proportion of households with wealth close to D will be
higher.
    The second comparative static suggests that increase in farm income has two counteracting
effects. The first one is that the affordability threshold decreases. This effect is driven by
liquidity-constrained households who are now able to finance migration due to increased income.
Another effect is that the desirability threshold decreases as well. This effect is driven by
households who now choose not to migrate because of increased total wealth without migration.
The total effect depends, again, in the densities at the thresholds. If the distribution of wealth
and the cutoffs for migration to India and non-India destinations are as depicted in Figure 2,
then rainfall could have completely different effects on migration to India and migration to other
destinations. For India, the fraction of households affected by relaxed liquidity constraints is
much higher than the fraction of households for whom increased income makes migration less
desirable. The liquidity effect is more likely to dominate, thereby increasing migration to India
in response to a higher rainfall. On the other hand, for non-India migration, both thresholds
   6
     I choose these measure of wealth as the current consumption is endogenous to migration choice, as is the
value of durables recently acquired. Acquisition of durable assets (especially two years ago) and education can be
considered to be pre-determined. I get the same pattern with alternative measures of pre-migration wealth such
as the value of land owned by the household, and the imputed rental value of their houses.
   7
     M is zero if D ≤ A. I assume that D > A, which, in the model, translates to an upper bound on the peace
premium. That is, there is only so much utility you can get by staying back when it is peaceful.


                                                        7
have only smaller proportion of households around them and therefore the net effect is likely
to be very small and of ambiguous sign. See Appendix A for the details of how this figure is
calibrated.
    The third comparative static suggests that an increase in demand from abroad is likely
to increase the number of households migrating to those destinations. This is also driven by
increasing the desirability threshold: the households that chose not to migrate previously will
migrate now, as the returns to migration are higher. Following Figure 2 again, the right cutoffs
of migration to both destinations move further to the right, increasing migration in both places.
For the same unit change in migrant income, the effect for India migration is likely to be larger
than the effect for non-India migration. However, the same proportional change in migrant
income, may induce larger response to non-India migration due to higher levels of migrant
income.


3        Context and Data
3.1        International migration from Nepal
Historically, Nepal has seen reasonable rates of international migration, albeit mostly to India.
Between 1961 and 2001, the migrant to population ratio hovered around 3.4 percent, and was
almost entirely driven by migration to India (Table 1). This rate was slightly higher than
the global international migration rate of 2.9 percent (UNDESA, 2015). The high rate can be
attributed to the ethno-linguistic similarity between the two countries as well as the low cost of
migration to India. Nepal maintains an open border with India, where citizens from one country
are free to enter the other at any time without any restrictions, paperwork or clearances. This
allows workers of either country to take advantage of the economic opportunities in the other.
Historically, workers, mostly from Far-western and Mid-western regions of Nepal, have been
migrating to India to work as daily wage laborers or security guards or in restaurants in Indian
cities. Because of frequent migration to India over a long period, there are well-established
migration linkages between districts in Nepal and Indian cities that help newer migrants find
work in India (Seddon, Adhikari, and Gurung, 2002).
    Because of the open border and historical linkages, India serves as the low cost and low
return destination for Nepali migrants. A typical migrant worker to India pays Rs. 6,250 (USD
83) to migrate and find a job in India (Table 2). This amount is roughly over two months of
per-capita consumption in Nepal. They earn, on average, Rs. 6,400 (USD 85) a month, of which
they save almost two-thirds. The median migrant spends about 9 months in India per migration
episode. This is consistent with the view that migration to India is seasonal and workers stay
home during planting and harvest season and migrate at other times of the year. Almost a third
of these migrants finance migration through their own savings while 60 percent take out a loan,
mostly from village lenders and then from friends and relatives8 .
    On the contrary, migration outside India is a relatively recent phenomenon. Historically,
people migrated to non-India destinations mostly through recruitment in the British Army.
The numbers were small and directed to destinations like the UK, Hong-Kong, Singapore and
Brunei. It was only in the late 1980s that Nepalis have started to migrate to other destinations
for work. Foreign migration became easier after democracy was introduced in the country in
1990 and international travel was made easier and more systematic. Only in the mid-1990s, the
    8
        Author’s calculations from Nepal Migration Survey 2009 and The World Bank (2011).



                                                        8
Government of Nepal allowed private recruitment agencies to recruit workers to a selected set
of countries, mostly in the Persian Gulf and a few others like Malaysia, Japan and South Korea,
upon obtaining clearance from the Ministry of Labor (Seddon, Adhikari, and Gurung, 2002).
This has led way to the surge in migrant outflow from Nepal in the 2000s (see Table 1). This
outflow has been dominated by the migration of low-skilled, mostly male, workers to Malaysia,
and the Persian Gulf countries, especially, Qatar, Saudi Arabia, and the United Arab Emirates.
By 2011, 15 percent of the households had a current migrant in these destination countries.
     Most of the current migration to non-India destinations, especially to the Persian Gulf coun-
tries and Malaysia, happens through recruitment agents (The World Bank, 2011). Typically,
potential migrants contact, or are contacted by, independent local agents that are connected
to recruitment firms in Kathmandu. These recruitment firms receive demands for low-skilled
workers from firms or agencies abroad and are fully responsible to fill the demands and arrange
all necessary paperwork9 . Migrants to these destinations pay the intermediaries more than
Rs.100,000 (US$ 1333) for job-search, intermediation and other related costs (Table 2). This
amounts to three years of per-capita consumption in Nepal.
     Despite the higher costs, workers earn significantly more in Malaysia and the Persian Gulf
countries once they are abroad. Most of these workers (65 percent) work in the booming con-
struction and manufacturing industries. A typical worker in these destinations earns more than
Rs.14,000 (US$ 187) in a month. They tend to save almost three-quarters of this income (Table
2)10 . The median migrant to the Gulf countries and Malaysia spends 2.3 years abroad per migra-
tion episode. More than 85 percent of them finance migration by borrowing, mostly from local
loan merchants and from friends and relatives11 . Borrowing from formal financial institutions
is, however, quite infrequent with less than 5 percent of the migrants borrowing from banks. Of
the remaining 15 percent too, most finance the migration through grants and help from family
and friends.
     A few key points of international migration from Nepal are worth highlighting, which ties
the theoretical framework in the context. The cost of migrating to India and finding a job is
much lower than the costs of migrating elsewhere. Interestingly, these costs do not seem to
vary by location of origin community or by the education level of the migrants12 . The returns
to migration also seem to follow the same pattern so that migration is quite profitable on
average, given the typical duration of migration and the amount of savings while abroad. Jobs,
particularly to non-India destinations, are prearranged so that there is no risk of not finding
work conditional on migration. The vast majority of the migrants finance their migration by
borrowing, but borrowing from formal financial institutions is quite low. Most migrants borrow
from local moneylenders against a collateral.

Data on migration rates
The data on migration for this paper comes from three different surveys conducted by the Central
Bureau of Statistics (CBS) of Nepal over 2001-2010 period. All three surveys have information
on absentee members of the household and because of their unique and circumstantial survey
design, forms a panel of nationally representative villages observed thrice during the decade.
   9
     For more details, see Shrestha (2017b,a).
  10
     Though the earnings in Malaysia seem to be lower than the earnings in the Gulf countries, it is believed that
the wages are actually higher in Malaysia. Workers in the Gulf countries tend to work more hours.
  11
     Author’s calculations from Nepal Migration Survey 2009 and The World Bank (2011).
  12
     Author’s calculations from Nepal Migration Survey 2009 and The World Bank (2011).



                                                        9
    In June 2001, CBS conducted the tenth national population census, which counted individ-
uals and absentee members in the entire country. I use the sample census micro-data, which
covers 95 percent of the villages in the country. Within each village, one-eighth of the households
are interviewed.
    The census of 2001 was used to develop a sampling frame for the Nepal Labor Force Survey-II
(NLFS-II) of 2008. This survey selected 800 primary sampling units (PSUs), or village wards,
through a stratified random sampling method. In each of these PSUs, they randomly selected
and interviewed 20 households.
    For the Nepal Living Standards Survey-III (NLSS-III) of 2010, the list of PSUs selected
for NLFS-II was used as the sampling frame. Of the 800 PSUs, 500 were selected, again by
stratified random sampling to be part of NLSS-III PSUs. This was done to save time in the
cartography and listing exercise by the CBS so that they could complete the survey before they
began preparing the logistics for the census of 2011. In each of these PSUs, they randomly
selected and interviewed 12 households each.
    Though households cannot be tracked between these different surveys, 452 village wards are
observed in all three datasets. Furthermore, these datasets also have information on the absentee
members of the household: individuals that were considered member of the same household but
had been absent at the time of enumeration. I use this information at the individual level to
construct the migration rates at the village level. My measure of migration rate is the ratio of
migrant to resident population in each village.
    Additionally, I use district level migration rates from published census tables for years 1981
and 1991 to extend the panel back in time in some cases.
    Figure 3 shows the spatial distribution of these PSUs all over the country. There is at least
one panel village in 70 of the 75 districts. In all of the estimations, I use the NLSS-III sampling
weights of these PSUs to make my estimates nationally representative. The weighted migration
rates in the panel are: 9.57 for year 2010, 8.06 for 2008, and 3.47 for year 2001, very close to the
rates reported in Table 1. A fuller set of descriptive statistics is presented in Table 3. Figure
4 shows the migration rates to India and non-India destinations for the 3 recent periods. As
discussed above, migration to India tends to be more common from the mid and far-western
regions of the country whereas migration elsewhere is increasing all over the country.
    In the remainder of this section, I describe the context and relevance of the ‘push’ and ‘pull’
shocks in this context along with the source of the data.

3.2      Rainfall and income
Nepal is primarily an agricultural country with farm income heavily depending upon rainfall.
During the 2001/2010 period, the share of agriculture in total GDP hovered between 35-38
percent (MoAD, 2013). In comparison, the total share of industry and manufacturing in GDP
remained below 20 percent in the same period. In 2010, 80 percent of the households own
agricultural land and the same proportion of the labor force is involved in agriculture13 . Since
the fraction of land that is irrigated remains low, agriculture depends heavily on rainfall. Rice
and maize are the two major crops grown in the country, both of which are grown in the wet
monsoon season. Rice is the key staple food and contributes to a fifth of the agricultural GDP
(MoAC, 2008). In 2010, 74 percent of farming households grew rice and 61 percent of them grew
maize14 . Therefore, the amount of rainfall the country receives during the monsoon months is
 13
      Author’s calculations from NLSS-III.
 14
      Author’s calculations from NLSS-III.


                                                10
crucial for agricultural production. A one standard deviation increase in monsoon rainfall in a
district improves the yield of rice by 2.7 percent, improving total production by 4.5 percent and
total rice cultivated area by 1.8 percent. Similarly, maize production increases by 5.9 percent
and area of cultivation by 6.8 percent in response to one standard deviation higher rainfall in
monsoon15 .
    Consequently, increased rainfall in the monsoon months increases farming income and con-
sumption of farming households. I investigate the impact of rainfall directly on measures of
income and consumption using three waves of cross-sectional surveys conducted between 1995
and 2010. I find that one standard deviation of rainfall increases farming income by Rs. 2,400
(about 16 percent of total farm income in 2010) and increases consumption in the subsequent
year by almost the same amount (Rs. 2,300 or about 7 percent of total consumption in 2010)
for households who do not receive any remittance income. Details of this exercise are described
in Appendix B.

Data on rainfall
The rainfall data that I use for this paper are provided by the Department of Hydrology and
Meteorology (DHM) for years 1972 to 2010. The dataset contains average daily rainfall for each
of the months from 1972-2010 collected from over 300 rainfall stations throughout the country.
The distribution of these stations is shown in Figure 3. I interpolate these rainfall measures to
each of the 452 villages in my dataset using inverse of distance from the stations as weights. For
each of the villages in my sample, I compute normalized monsoon rainfall in standard deviation
units using historic (spanning 1972-2002) mean and standard deviation of monsoon rainfall in
the village. The measure that I use for my analysis is the normalized monsoon rainfall in the
year preceding the survey.
    There is considerable variation in rainfall between years as well as within a given year. Figure
5 plots the histogram of normalized rainfall for the three survey years. Within each year, the
range is well over 2 standard deviations and across the three years, the range is more than
4 standard deviations with considerable overlap between the years. From this graph alone, it
looks like rainfall is trending down in the 2000s, but this is driven by exceptionally high rainfall
around year 2000. The trend does not hold if rainfall distribution from earlier periods (years,
1981 and 1991) is added.

3.3    Conflict: The Maoist insurgency
When international migration from Nepal began to increase drastically in late 1990s and early
2000s, the ongoing Maoist insurgency took an unexpectedly sharp and violent turn. This led
experts to believe that the conflict, which took over 13,000 lives and displaced many more, also
pushed individuals outside the country (CBS, 2006; The World Bank, 2011, for instance). In
this subsection, I discuss the context of the conflict and my data source.
    The Maoist insurgency began, and gained momentum, as an anti-feudal and anti-elite move-
ment. The insurgency, or the “People’s War” as they call it, formally began on February 13,
1996 by a splinter faction of the leftist communist parties that was dissatisfied with the lack of
  15
    Author’s calculations from the historic rainfall data and historic production data for each of the districts
for years 1975-2000. The estimated regression includes district fixed effects as well as year fixed effects. The
identifying variation comes from geographic variation in rainfall measures within a given year and also inter-
temporal variation in rainfall within a given district.



                                                      11
progress since democracy was introduced in the country in 1990. As a sign of their movement,
they attacked three police outposts and a a few privately owned firms and banks located in re-
mote parts of the country to symbolize their attack against the feudal government, the capitalists
and the elites. Their movement spread gradually to other parts of the country with support
from the rural and poor masses. Their expansion and operation included few key features: they
seized land from the landlords and let poorer farmers cultivate under cooperative systems, they
extorted money from the wealthy, attacked private firms and banks, and encouraged the poor
and often marginalized people to join their movement by highlighting their pro-poor agenda
(Macours, 2011). Their attacks on government police posts and subsequent retaliation by the
police (which involved capturing and torturing people that they believed were militants) made
the conflict increasingly violent. By 2001, 2,000 people were killed, mostly in the mid and
far-western regions of the country.
    An unexpected incident changed the nature of the conflict and made it much more violent
after 2001. Before 2001, King Birendra considered the Maoist insurgency to be an internal
problem to be dealt by the government and the police force. He refrained from mobilizing the
army against the insurgency even upon requests from the government to mobilize the army. On
June 1, 2001, King Birendra and his entire family were massacred at a family dinner by the then
crown prince, who subsequently turned the gun to himself. In a matter of days, Gyanendra,
who was very far in the line of succession to the throne, and who survived as he was not present
for the dinner, became the new monarch. King Gyanendra was less reluctant to use the army
to suppress the insurgency. The ongoing peace talks with the government broke in late 2001
with the Maoists attacking an army barrack in one of the mid-Western districts. The Nepal
Army was deployed overnight and the conflict took a very violent turn. In 2002 alone, over
4,600 people were killed. By 2006, when the insurgency eventually ended, over 13,000 people
were killed and more than ten times as many displaced.
    Another equally unexpected coalition emerged during the conflict that ended the insurgency
in 2006. In a short duration, Gyanendra proved himself unpopular when he dismissed the elected
parliament and started ruling the country through his handpicked cabinet of ministers since 2005.
This presented an unusual opportunity for the Maoists to become a mainstream political party
by teaming up with the other political forces in a fight against the King. This movement led to
the demise of the monarchy and an election was held in 2008 to write the constitution for the
newly formed republic. The Maoists emerged as the largest party, confirming its transition to
mainstream political force.
    The geography and the trajectory of the insurgency produces spatial and inter-temporal
variation in conflict intensity, which I exploit in my empirical study. Conflict intensity was
lower before year 2001, when census data were collected. The conflict took an unexpectedly
violent turn after 2001, until its end in 2006. The threat of violent Maoist conflict ended in
2008 when the constituent assembly elections were held, the year when the NLFS-II data were
collected. The period between 2008 and 2010 represents a post-conflict peace period after the
Maoists emerged as the largest party in the elections. Figure 6 shows the stark increase in
conflict intensity before and after 2001. Conflict intensity increased from 0.1 deaths per 1,000
population on average to 0.6 deaths per 1,000 population in this period. The figure also shows
large geographic variation in conflict intensity in both periods. Further, post-election peace
ended violent conflict in all districts.




                                               12
Conflict data
My measure of conflict intensity is based on data provided in the annual Human Rights Year-
books published by the Informal Sector Service Center (INSEC), a Nepali non-governmental
organization that monitors human rights issues in the country. INSEC reports the number of
deaths due to Maoist conflict in each of the districts in the country. This dataset is identical
to the one used by Do and Iyer (2010) in their study of the determinants of the conflict. I use
conflict related deaths that occurred in the district between 1996 and 2001 as my measure of
conflict intensity for year 2001. For year 2008, I use conflict related deaths that occurred in the
district between 2001 and 2006. I set conflict intensity to zero for year 2010. I normalize the
conflict deaths by dividing by the population of the district in 1991.
    To convert conflict into monetary terms, I do an exercise similar to that with rainfall using
three cross-sections of consumption surveys conducted between 1995 and 2010. I find that
increase in conflict intensity of 1 death per 1,000 population is correlated with a consumption
loss of Rs. 3,490 (10.6 percent of consumption in 2010). This correlation is much stronger
for higher end of the wealth distribution, consistent with the nature of the conflict and the
assumption in the theoretical framework (Appendix Table B.1 and Appendix Figure B.1)16 .

3.4      Demand for migrant workers
Demand data
Another shock that drives migration in the theoretical framework is the increased demand for
workers in the destination countries, which increases migrant income. Since most migrant work-
ers from Nepal work in construction and manufacturing industries, I use a measure of demand
that reflects growth in construction and manufacturing industries in the key destination coun-
tries. Specifically, I use the growth rate in the levels of carbon dioxide emissions attributed to
construction and manufacturing industries in India, Malaysia and the Gulf countries (Qater,
Saudi Arabia, and United Arab Emirates). I get this measure from the World Development In-
dicators database (The World Bank) which reports carbon dioxide emissions calculated by U.S.
Department of Energy’s Carbon Dioxide Information Analysis Center (CDIAC) using data from
the United Nations Statistics Division’s (UNSTAT) World Energy Data Set and the U.S. Bureau
of Mine’s Cement Manufacturing data set. The UNSTAT collects this information from several
countries using questionnaires designed to collect information on emissions from various sectors.
Cement manufacturing data from the U.S. is used to estimate CO2 emission from construc-
tion. Further, CDIAC continually maintains and updates this database based on supplemental
information collected and upon availability of new information.
    The variation in this dataset comes from the fact that the destination countries are in different
cycles of construction and manufacturing. Further, the great recession of 2008 affected these
destination countries quite differently, creating a useful variation in the data on the growth of
migrant demand.
 16
      See Appendix B for details of this estimation.




                                                       13
4    Empirical Strategy
Push shocks
Ordinary Least Squares estimates of the effect of shocks on migration in a cross-sectional data
are fraught with problems. For instance, places with fertile land that is responsive to rainfall
might also have more entrepreneurial people who are more likely to migrate. Similarly, places
with extremely unequal land ownership might invite conflict as well as force people to migrate
due to lack of economic opportunities. These characteristics, which are not always observable,
will bias the OLS estimates obtained from cross-sectional data. Panel data solve this problem
if such characteristics are time invariant by comparing changes in migration rates with changes
in shocks within the same village over different periods.
    Given the structure of my migration data, I estimate the following model to study the impact
of ‘push’ shocks on migration:

                                    Mit = βXit + γt + µi + εit                                   (1)
where Mit is the migration rate in village i observed in time t, γt represents the survey year fixed
effects which absorbs any national trends common to all villages, µi is the village fixed effect
which absorbs all time invariant characteristics of the village, Xit is the shock that affects village
i at time t and εit represents the error term. I allow the error term to be correlated between
different villages within a district as well as over time. Xit are either conflict intensity in the
district as defined above, or normalized rainfall in very recent past (one or two years ago). The
identifying variation for this specification comes from variation within village over time as well
as variation across villages in a given year.
    Further, my measure of rainfall shock is in standard deviation units relative to the average
historic rainfall in the village and is likely to be uncorrelated with any village specific trends
in any of the related observed and unobserved variables. More importantly, individuals cannot
predict rainfall shocks and engage in anticipatory migration. Hence, the variation in rainfall
measure is likely to be uncorrelated with the error term.
    Conflict intensity could, however, be driven by village or region specific trends. This could
bias the estimates of the effect of conflict intensity. However, the overall trajectory of this
conflict, as discussed earlier, suggests that the two big factors that drove conflict intensity were
quite unanticipated. The deployment of Nepali Army, which escalated the death toll stemmed
from the massacre of the royal family and coronation of a new king who was not in the accession
line. Similarly, the end of the conflict was brought together by an unlikely alliance between the
mainstream political parties and the Maoists to overthrow the king.

Pull shocks
The regression model in Equation (1) cannot be used to estimate the effect of ‘pull’ shocks from
the destination countries, as they will be completely subsumed by the year fixed effects. To look
at whether migration responds to demand from the destination countries, I estimate a slightly
different model:

                                   Mijt = δDjt + γij + µit + εijt                                (2)
where Mijt is the migration rate from village i to country j in time t, γij is the destination-
village fixed effect, µit is the village year fixed effect and Djt is the recent growth rate in

                                                 14
manufacturing and construction sector in country j (the Gulf countries, Malaysia, and India).
As with equation (1), I allow the errors to be correlated across time and across different villages
in the same district.
    The identification of this specification comes from variation in demand from different desti-
nation countries within a village for a given year as well as from the variation in demand across
time for a village-destination pair. The fixed effects remove all other sources of variation that
might confound the estimation.
    In the next section, I present results of estimating equations (1) and (2) and discuss how the
results relate to the theoretical framework outlined in Section 2.


5         Results and discussion
5.1         A ‘push’: Rainfall shocks
As discussed in the previous section, income shocks due to rainfall could have an ambiguous
effect on migration. It increases wealth, making migration less desirable, but also relaxes liq-
uidity constraints, which makes migration affordable. As suggested by Figure 2, the fraction of
households that are induced by the rainfall shock in overcoming the liquidity constraints might
be very small relative to the fraction of households that now desire migration less when the
cost of migration is very high. The former proportion is likely to be higher when the cost of
migration is relatively low.
     Consistent with this interpretation of the framework, I find that rainfall shocks increase
migration to India whereas it has no effect on migration to non-India destinations. As Table 4
shows, a one standard deviation increase in rainfall increases migration to India by 0.8 percentage
points. This amounts to 31 percent increase in migration rate from its 2001 level, 19 percent of
2008 level, and 17 percent of 2010 level. On the other hand, rainfall has no impact on migration
to non-India destinations. The estimated effects are small and statistically insignificant.
     Since I have only three periods, arbitrary trends in normalized rainfall could be driving the
results. Indeed, Table 3 and Figure 5 shows that normalized rainfall in the past year has been
falling between 2001 and 2010, whereas migration rates have been increasing over this period.
To check that the trend in rainfall is not driving my results, I extend the panel further back
in time using district level migration rates from the census of 1981 and 1991 to proxy for the
migration rates in the village in 1981 and 1991. Table 5 shows the results for this extended panel.
The same specification produces almost identical results except for effects on India migration
from urban areas (columns 1, 3, and 5). The effect for the urban areas is about half the size
of the effect for rural areas. This could be because urban areas are less reliant on rainfall for
their income17 or, according to the theoretical framework, because the fraction of households
affected by the liquidity constraint is smaller in urban areas. Adding a linear time trend only
slightly reduces the point estimates. This indicates that time trends in rainfall are not driving
the results.
     In Table 6, I investigate whether rainfall shocks two years before the survey affects migration
rates. I do this to capture the effect on migrants who left more than a year ago for whom it
is not the last year’s rainfall but the rainfall two years ago, that matters. Baseline estimation
of Equation (1) shows that normalized rainfall two years ago matters almost as much as the
normalized rainfall last year for migration to India (columns 1, 3, and 5). But this relationship is
    17
         Only 45 percent of urban households owns agricultural land compared to 87 percent of rural households.



                                                          15
not robust to including region specific linear time trends (columns 2, 4, and 6). The magnitude
of the effect of rainfall two years ago becomes much lower with the linear time trends. For non-
India migration, the effects seem to be small and negative once I control for linear time trends.
The results suggest that rainfall facilitates migration to India but not to other destinations and
that, more recent rainfall matters more.
    To further ensure that the trends in rainfall are not driving the results, I use a falsification test
arising naturally in this setting. Though past rainfall may affect current migration rates, future
rainfall should not affect it, as households cannot anticipate future rainfall shocks. A failure
of this test would suggest that village specific trends, and not the increase in farm incomes
are driving the migration results. Table 7 shows the results of this check. As expected, the
coefficients are statistically insignificant with point estimates close to zero. This result is robust
to including region specific linear time trends (columns 2, 4, and 6).
    Since rainfall measure is essentially exogenous and affects migration only through farm in-
come, I can interpret the rainfall shock as an instrument that shifts household income. Since
I do not have income measures for the census data and the NLFS for years 2001 and 2008, I
cannot use an instrumental variable estimate directly. However, I do have income measures for
three cross-sections of NLSS rounds conducted in 1995/96, 2003/04 and 2010. As described in
detail in Appendix B, I find that one standard deviation increase in rainfall increases farming
income by Rs. 2,400 and, as seen from results in this section, increases migration to India by
0.008. Scaling the impact of rainfall on migration by the impact of rainfall on income, increase
in farm income of Rs. 7,500 (USD 100) increases migration to India by 2.5 percentage points,
a large 54 percent increase from its 2010 level. In terms of elasticities, the implied elasticity of
migration to India with respect to farm income is 1.1. Similarly, the implied elasticity of India
migration with respect to per-capita consumption is 2.5.

5.2    Another ‘push’: Conflict
As predicted by the theoretical framework, conflict increases international migration in general
(Table 8). The increase in conflict intensity of one death per thousand population increases
migration by 0.8 percentage points. This represents an increase of 11 percent in migration rates
if conflict increased from its pre-2001 level to the post-2001 level. Similar calculations suggest an
increase of 9 percent in migration to India and 18 percent in migration to non-India destinations.
Most of these results, though in the direction predicted by the framework, are estimated with
large standard errors and are not significant at conventional levels.
     The theoretical frameworks suggests that the migration response to conflict should be higher
for the richer households. Consistent with this, I find that, for each type of destination, the
migration response from the urban areas are about 6 times higher than the response in rural
areas. An effect of 3.1 percentage points means that an increase in conflict from its pre-2001
level to post-2001 level in urban areas increases international migration by a large and significant
40 percent. This is a third of the observed increase in migration between 2001 and 2008. The
corresponding increases are 62 percent for migration to India (two-thirds of the total increase)
and 24 percent (one-sixth of the total increase) for migration to non-India destinations. These
effects are large, and in case of urban migration to India, explain a large share of migration that
happened between 2001 and 2008. These results suggests that the urban households respond
more strongly to increased violence by migrating abroad.
     As discussed earlier, conflict can affect migration in two broad ways. First, conflict causes
a direct loss in income, which affects migration decision by affecting the desirability and afford-


                                                  16
ability of migration. Second, people do not like conflict either because of the threat it poses
to their safety or myriad of other reasons, which affects only the desirability of migration. The
monetizing exercise, detailed in Appendix B and tabulated in Table B.1 suggests that increase in
conflict intensity of 1 more death per 1,000 population lowers consumption by Rs. 3,500. Using
the results from rainfall shocks as the causal effect of income on migration, a fall in consumption
of Rs. 3,500 predicts a fall of migration to India from urban areas of 0.65 percentage points 18 .
But the observed increase in migration from urban areas to India in response to conflict is 1.98
percentage points (Table 8, column (3)). This suggests that the second channel must be more
dominant to explain the opposite effect that we observe. The migration to India from urban
areas in response to their dislike of conflict would then be 2.63 percentage points. This effect is
the size of migration response to India from a rainfall shock that increases household income by
Rs. 14,000 (USD 190).
    Though useful, this exercise assumes a constant effect of conflict on wealth for everyone.
This, however, is not true. As shown in Figure B.1, the effect of conflict on consumption is
much higher for the wealthier households. The estimate of US$ 190 is therefore a likely lower
bound of the average valuation of conflict by urban households. Repeating the same exercise
using the effect of conflict on consumption in urban areas (Rs 21,000 instead of Rs 3,500) suggests
that the equivalent increase in household income of Rs 31,000 (US$ 420). That is, the increase in
migration rates to India resulting from a unit increase in conflict would be offset by an increase
in household income by US$ 420. The offsetting effect, however, comes from different part of the
distribution: the liquidity constrained households being able to finance migration rather than
the richer households who are affected more severely by conflict.
    Furthermore, given the heterogeneous impact of conflict on consumption, the theoretical
framework suggests that an increase in migration could be possible following an increase in
conflict even through the consumption reduction channel. As Figure B.1 suggests, the effect
of conflict on consumption for the bottom 30 percentile is essentially zero. These are precisely
the households that are near the affordability margin for migration to India (see Figure 2). An
increase in conflict would not change the affordability status for liquidity constrained households,
whereas for the slightly richer households around the desirability threshold, it would make
migration more desirable and hence increase migration to India. Conflict has similar effects
for non-India migration as well. For the households in the desirability threshold for non-India
migration, conflict lowers income as well as amenity more than the poorer households, hence,
pushing them to migrate (particularly from the urban areas). However, because of the hight
affordability threshold for non-India migrants, a negative income shock from conflict reduces
their ability to finance migration, which creates a slightly more muted effect.
    Hence, the observed effects are consistent with conflict having a direct effect as a reduction
in amenity as well as an indirect and heterogeneous impact through income or consumption.

5.3    The ‘pull’: Migrant demand
According to the theoretical framework, increase in demand for migrant workers, reflected by an
increase in migrant income, increases migration by making it more desirable. Migrant demand
from a particular destination country affects all of Nepal, which makes it impossible to separate
the effect of migrant demand from within country trends in migration. Instead, what I can ask
is whether migration responds to differential growths in various destinations. That is, whether
  18
    The effect of rainfall on migration to India in urban areas is 0.005 (Table 5, column (6)), and the effect of
                                                                                                 .005
rainfall on income is Rs 2,400 (Table B.1, column (3)). Hence, the effect of Rs 3,500 is simply 0
                                                                                               2400
                                                                                                      × 3500.


                                                      17
migrants are more likely to go to Malaysia when migrant demand in Malaysia is growing faster
relative to the Gulf countries.
    The results show that migrant supply does respond to changes in migrant demand (Table 9).
In particular, the most recent growth rates have the strongest impact in migration outflows. The
coefficient of 0.015 in column 1 corresponds to an increase in migration of 3.7 percent (evaluated
at the mean) in response to an increase in growth rate of 5 percentage points. The corresponding
increase for non-India destinations implied by the coefficient of 0.033 is 14 percent.
    The results get stronger when India is removed from the list of destinations. As the last
three columns of Table 9 shows, the magnitude without India as a destination is almost twice
as when India is included19 . This suggests that migrant supply is less responsive to increased
construction and manufacturing in India. This could be because the same percentage growth
rate translates to a much bigger increase in migrant income in non-India destinations than in
India. Consequently, in the language of the framework, the share of households that is induced
to migrate to non-India destinations is much higher relative to India.
    Pre-existing migrant networks does not seem to be the primary channel by which potential
migrants learn about the rising migrant demand in the destination countries. To investigate
this, I interact the growth rate in destination with the size of migrant network from the village
to the destination in the previous period. I find, particularly in cases where India is excluded as
a potential destination, that the size of the pre-existing migrant network does not affect the level
of response to growth in destination countries. The estimated regression coefficients are negative
with p-values higher than 0.80 (results not shown). This suggests that potential migrants find
information about migrant demand in Malaysia and the Persian Gulf countries from channels
other than the existing local migrant network. The increased demand may be transmitted
though recruitment companies and the local agents who seek workers more aggressively when
migrant demand abroad is higher.


6    Conclusion
Workers want to move towards better economic opportunities and away from disamenities such
as conflict. The 2011 Gallup polls estimates that over 1 billion individuals desire to migrate
abroad, at least for temporary work (Esipova, Ray, and Publiese, 2011). But the fact that only
3 percent of the global population are international migrants suggests that many who want
to migrate do not, in fact, migrate. Lack of global mobility of workers has been suggested as
one of the biggest distortions in the global economy (Clemens, 2011). Workers typically cannot
migrate to places where the returns to their skills are highest for two key reasons. First, liquidity
constraints may prevent them from being able to afford the costs of migration. And, second,
the lack of demand for migrant workers, either economic or institutional, lowers the expected
benefit from migration.
    In this paper, I investigate how ‘push’ and ‘pull’ shocks affects the migration of Nepali
workers to destinations that vary in their costs of, and returns to, migration. First, I find
that income shocks, arising from positive rainfall shocks, increases migration to a low-cost low-
return destination (India), but these shocks do not affect the migration to high-cost high-return
  19
     These results are consistent with what I find in Shrestha (2017a) for migrant outflows from Nepal to Malaysia
and the Persian Gulf countries. In an exercise in that study, I find that when relative exchange rates between
destination changes, causing a shift in the wages that they earn, migrants respond by choosing the destination
with a favorable exchange rate. I cannot use the same variation as a measure of a shock to destination wages in
this study because Nepal has a fixed exchange rate with India.


                                                      18
destinations such as Malaysia and countries in the Persian Gulf. Second, I find that negative
shocks to amenity, measured by violent conflict, push people abroad, particularly from wealthier
urban areas. Third, I find that increase in migrant demand from the high-return destination,
measured by growth in migrant-employing sectors, increases migration to those destinations.
    These patterns are consistent with a simple theoretical framework of migration choices with
liquidity constraints and availability of destinations that differ in the costs of and returns to
migration. It suggests that different types of households are liquidity constrained to migrate to
India and to non-India destinations. Rainfall shocks, which increases the income of the poor
farmers more, only allow the liquidity constrained households to afford migrating to India. These
shocks, as they are small in size, are not able to push a sizable share of households above the
affordability threshold for the expensive but lucrative non-India destinations. Similarly, conflict,
which reduces the peace related amenity and the income of the richer more, makes migration
more desirable for the wealthier households. Conflict pushes the wealthy households more as they
are the ones taking a bigger hit in terms of income and amenity at the origin, and at the same
time can also afford to migrate away to avoid conflict. Finally, growth in destinations in migrant-
employing sectors translates to higher expected income from migration (either through higher
wages, or through higher probability of finding a job) which makes migration more desirable.
    My results highlight several aspects of international migration. First, conflict – at least in
the context of Nepal – selectively pushes wealthier households to migrate. Second, a large share
of households are liquidity constrained and are not able to finance migration to destinations
where the returns to their skills are the highest. Small increases in income may help them
finance migration, but only to the low-cost and low return destinations. Large transfers might
be needed to enable them to migrate to destinations that provide the highest return to their
(low) skills. Third, a substantively large share of households are also constrained by the lack
of demand from more lucrative destination countries. The large responsiveness to ‘pull’ shocks
suggests that these households are willing to take advantage of the opportunities abroad should
the demand increase. The rise in migrant labor demand from Malaysia and the Perisan Gulf
countries is perhaps the biggest factor behind the over six-fold increase in migration from Nepal
to these destination countries between 2001 and 2011.




                                               19
References
Abramitzky, R., L. P. Boustan, and K. Eriksson (2013): “Have the poor always been less
 likely to migrate? Evidence from inheritance practices during the Age of Mass Migration,”
 Journal of Development Economics, 102, 2–14.

Angelucci, M. (2015): “Migration and financial constraints: Evidence from Mexico,” Review
 of Economics and Statistics, 97(1), 224–228.

Bazzi, S. (2014): “Wealth heterogeneity and the income elasticity of migration,” Unpublished
 manuscript.

Becker, C. M., E. N. Musabek, A.-G. S. Seitenova, and D. S. Urzhumova (2005): “The
 migration response to economic shock: Lessons from Kazakhstan,” Journal of Comparative
 Economics, 33(1), 107–132.

Bryan, G., S. Chowdhury, and A. M. Mobarak (2014): “Underinvestment in a profitable
 technology: The case of seasonal migration in Bangladesh,” Econometrica, 82(5), 1671–1748.

CBS (2006): “Resilience amidst conflict: An assessment of poverty in Nepal, 1995-96 and
 2003-04,” Discussion paper, Central Bureau of Statistics, National Planning Commission Sec-
 retariat, Government of Nepal.

Clemens, M. A. (2011): “Economics and emigration: Trillion-dollar bills on the sidewalk?,”
 The Journal of Economic Perspectives, pp. 83–106.

        (2014): “Does Development Reduce Migration?,” CGD Working Paper 359, Center for
  Global Development.

Davenport, C., W. Moore, and S. Poe (2003): “Sometimes you just have to leave: Domestic
 threats and forced migration, 1964-1989,” International Interactions, 29(1), 27–55.

Do, Q.-T., and L. Iyer (2010): “Geography, poverty and conflict in Nepal,” Journal of Peace
 Research, 47(6), 735–748.

Esipova, N., J. Ray, and A. Publiese (2011): “Gallup World Poll: The many faces of global
  migration,” IOM Migration Research Series, 43.

Gray, C., and V. Mueller (2012): “Drought and population mobility in rural Ethiopia,”
 World Development, 40(1), 134–145.

Halliday, T. (2006): “Migration, risk, and liquidity constraints in El Salvador,” Economic
 Development and Cultural Change, 54(4), 893–925.

Harris, J. R., and M. P. Todaro (1970): “Migration, unemployment & development: A
 two-sector analysis,” The American Economic Review, 60(1), 126–142.

Jacoby, H. G., and E. Skoufias (1998): “Testing theories of consumption behavior using
  information on aggregate shocks: Income seasonality and rainfall in rural India,” American
  Journal of Agricultural Economics, 80(1), 1–14.

Kazianga, H., and C. Udry (2006): “Consumption smoothing? Livestock, insurance and
 drought in rural Burkina Faso,” Journal of Development Economics, 79(2), 413–446.

                                            20
Macours, K. (2011): “Increasing inequality and civil conflict in Nepal,” Oxford Economic
 Papers, 63(1), 1–26.
Marchiori, L., J.-F. Maystadt, and I. Schumacher (2012): “The impact of weather
 anomalies on migration in sub-Saharan Africa,” Journal of Environmental Economics and
 Management, 63(3), 355–374.
McKenzie, D., and H. Rapoport (2007): “Network effects and the dynamics of migration
 and inequality: Theory and evidence from Mexico,” Journal of development Economics, 84(1),
 1–24.
McKenzie, D., C. Theoharides, and D. Yang (2014): “Distortions in the international
 migrant labor market: Evidence from Filipino migration and wage responses to destination
 country economic shocks,” American Economic Journal: Applied Economics, 6(2), 49–75.
Melander, E., and M. Öberg (2007): “The threat of violence and forced migration: Geo-
 graphical scope trumps intensity of fighting,” Civil Wars, 9(2), 156–173.
Miguel, E. (2005): “Poverty and witch killing,” The Review of Economic Studies, 72(4), 1153–
 1172.
MoAC (2008): “Selected indicators of Nepalese agriculture and population,” Discussion paper,
 Ministry of Agriculture and Cooperatives, Government of Nepal.
MoAD (2013): “Statistical infomration on Nepalese agriculture 2012/2013,” Discussion paper,
 Ministry of Agricultural Development, Government of Nepal.
Moore, W. H., and S. M. Shellman (2004): “Fear of persecution: Forced migration, 1952-
 1995,” Journal of Conflict Resolution, 48(5), 723–745.
Munshi, K. (2003): “Networks in the modern economy: Mexican migrants in the US labor
 market,” The Quarterly Journal of Economics, 118(2), 549–599.
Naudé, W. (2008): Conflict, disasters and no jobs: Reasons for international migration from
 Sub-Saharan Africa. Research paper/UNU-WIDER.
Paxson, C. H. (1992): “Using weather variability too estimate the response of savings to
  transitory income in Thailand,” The American Economic Review, 82(1), 15–33.
Seddon, D., J. Adhikari, and G. Gurung (2002): “Foreign labor migration and the remit-
  tance economy of Nepal,” Critical Asian Studies, 34(1), 19–40.
Shrestha, M. (2017a): “Death scares: How potential work-migrants infer mortality rates from
  migrant deaths,” World Bank Policy Research Working Paper Series WPS7946, The World
  Bank.
         (2017b): “Get rich or die tryin’: Perceived earnings, perceived mortality rate and the
  value of a statistical life of potential work-migrants from Nepal,” World Bank Policy Research
  Working Paper Series WPS7945, The World Bank.
The World Bank, . (2011): “Large-scale migration and remittance in Nepal: Issues, challenges
 and opportunities,” Discussion Paper 55390-NP, The World Bank, Poverty Reduction and
 Economic Management Sector Unit, South Asia Region.

                                              21
         (2016): “World Development Indicators,” .

UNDESA (2015): “Trends in International Migrant Stock: The 2015 Revision,” Discussion
 paper, United Nations, Department of Economic and Social Affairs, Population Division.

Williams, N., and M. S. Pradhan (2009): “Political conflict and migration: How has violence
 and political instability affected migration patterns in Nepal?,” Discussion paper, University
 of Michighan, Population Studies Center.

Wolpin, K. I. (1982): “A new test of the permanent income hypothesis: the impact of weather
 on the income and consumption of farm households in India,” International Economic Review,
 pp. 583–594.

Yang, D. (2008): “International migration, remittances and household investment: Evidence
 from Philippine migrants’ exchange rate shocks,” The Economic Journal, 118(528), 591–630.




                                             22
7     Tables and Figures
7.1    Figures


                 Figure 1: Migration probability and measures of household wealth


                                        .6
                       Probability that a HH has a migrant
                                 .2     0       .4




                                                             2   4         6          8         10          12         14
                                                                                 Log(Lagged durables)

                                                                     migrant to India             migrant to Gulf-Malaysia
                                        .2
                       Probability that a HH has a migrant
                                 .1     .05    .15




                                                             0   2        4            6         8         10       12       14
                                                                              Highest adult education in HH

                                                                     migrant to India             migrant to Gulf-Malaysia


Source: Author’s computation from the NLSS-III, 2010.
Note: The top plot shows the relationship between the probability of a household having a migrant and the
value of durable assets acquired before the past two years. The bottom plot shows the relationship between the
probability of a household having a migrant and the highest adult education level in the household. Both plots
are estimated using locally linear regressions. In both plots, the solid (maroon) line shows the probability of a
household having a migrant in India, and the dashed (green) line shows the probability of a household having a
migrant in Malaysia and the Persian Gulf countries.




                                                                                        23
                                                                                     Figure 2: Model Illustration



              f(w)




                                                                                                                                      w

                                                                                 India migrants                                                    Non-India migrants


Note: This figures illustrates the plausible set of thresholds for migration to Indi and non-India destinations



                  Figure 3: Distribution of PSUs (village wards) and rainfall stations


                               (
                               !                         E
                           E
                           !
                           (                                                         E
                                                                  E
                       E      (E
                              !
                     (
                     !EE     (
                             !       ( E
                                     !                            EE
                                                                   !
                                                                   (
                     !
                     (                                !
                                                      (
                           (
                           !
                           E
                           (
                           !        (
                                    !
                                          E E (
                                              !           !
                                                          (
                        !
                        (E!(
                             ( E
                             !      E !
                                      (                  E    (
                                                              !       EE E
                                  (
                                  !       (
                                          !
                                            E(
                                             !
                                                     E        (
                                                             E!                                                                  E
                                                                                                                                 E
                                            E(
                                             !
                                             !
                                             (      E
                           E                                                                                                      E
              (E
              !                    !
                                   (!(           (
                                                 !       !
                                                         (                                                                   E
                                        E                                  E
               (
               !(!
                ! (                    E        E (!       !
                                                           (                                     E                       E
                   (
                   !      E                E                          !
                                                                      (                                              E
                        E
                        (
                        !( !
                         !                              !
                                                        (
                                                        E                                                                    E
                        E   (                E      E                                                                E                E
                                  E                (
                                                   !                                                                 E
                   E !  (!
                       (!(                                             E         !
                                                                                 (           (
                                                                                             !
                                 (
                                 !         E              !
                                                          (
                                                          (
                                                          !           (
                                                                      !E       ( E
                                                                               !                                 E                E       E
                     (!
                     !(    !
                           ( !(                  ( !
                                                 !  (E
                                                     !
                                                     (                                   E           E
                               E                                  (
                                                                  !                          E               E
                                     (!
                                     !(E!(              !
                                                        (                                                (
                                                                                                         !
                                                                                                         E
                                                                                                                                      E
                                      (
                                      !E      E       (!
                                                      !  (
                                                                        !
                                                                        ( ! (
                                                                                       E              EE E                                   E
                                           !
                                           ( E                !
                                                              (    E (                  E!               EE
                                                       E             ! ! ! (E !
                                                                                          ( !  (E E(E!
                                                                                                   !
                                                                                                   !
                                                                                                   (  ( E          E
                                                                                                                   (
                                                                                                                   !                      E
                                              !
                                              (( ( !
                                               !        E            E (
                                                                     (
                                                                     !          (                    !
                                                                                                     (E E  EE!(E           E              !
                                                                                                                                          (             E
                                             ( !
                                             !    ( (
                                                  !               E
                                                                      !
                                                                      (                           ( !
                                                                                                  !             (
                                                                                                                !
                                                                                                                !
                                                                                                                (
                                                                                                                E
                                                                                                                (
                                                                                                                !
                                                                                                                !
                                                                                                                (         (
                                                                                                                          !
                                             (
                                             !E !     !
                                                      (!
                                                       (                                    E          (E
                                                                                                        !
                                                                                                        ( E(
                                                                                                           ! (!
                                                                                                             ! (
                                                                                                               (
                                                                                                               !
                                                                                                               !
                                                                                                               (E
                                                                                                                E!(E      (
                                                                                                                          !       E               E ! (        E
                                                  ( (
                                                    !               (E
                                                                    !
                                                                      !
                                                                      (
                                                                      E                     (
                                                                                            !        EE     !
                                                                                                            ((
                                                                                                             !    !
                                                                                                                  (!
                                                                                                                   (      E !                        EE
                                                   EE E                           E!
                                                                                  !
                                                                                  ( (                        EE     E      ( (
                                                                                                                           !    ( !
                                                                                                                                !   ( ! (             E
                                                    !
                                                    (!
                                                     ( E E
                                                     (
                                                     !            E   ( !
                                                                      !  (!
                                                                          (
                                                                          E     (!
                                                                                ! ( ( !
                                                                                           (
                                                                                          (!
                                                                                          ! E
                                                                                               (
                                                                                               !    (
                                                                                                    !  E E! (!
                                                                                                             (     E     !
                                                                                                                         (        E
                                                                                                                                       E        E         !
                                                                                                                                                          ( E
                                                                                                                                  (
                                                                                                                                  !
                                                                         ( !
                                                                         !  ( !
                                                                              (       ( E !
                                                                                      !     ( E
                                                                                                      (
                                                                                                      !
                                                                                                          E           (!
                                                                                                                      ! (  (
                                                                                                                         E !E                    (
                                                                                                                                                 !
                                                                                                                                                 EE            EE
                                                          !
                                                          (                                                                 ( !
                                                                                                                            !  (    (
                                                                                                                                    !   (E
                                                                                                                                        !          (
                                                                                                                                                   !
                                                                          !
                                                                          (!(           (E
                                                                                        !          !
                                                                                                   (E !    EE!(     E                  ( !
                                                                                                                                       !  ( !(     (
                                                                                                                                                   !          EE      E
                      Legend:
                                                                                                        (!(                                         E
                                                                                                                                                    E   EEE     E E ! (                      EE
                                                                                      E                                            !
                                                                                                                                   (      (
                                                                                                                                          !         EEE
                                                                                                                                                      E(
                                                                                                                                                       !EE
                                                                                                                                                         E
                                                                                                                                                         E
                                                                                                                                                         EE
                                                                                                                                                          E(
                                                                                                                                                           !EE         (
                                                                                                                                                                       !
                                                                                                                                                                       E       !
                                                                                                                                                                               (                                 E
                                                                                                                      E                                (
                                                                                                                                                       !
                                                                                                                                                       E
                                                                                                                                                       (
                                                                                                                                                       ! !
                                                                                                                                                         (
                                                                                                                                                         E     !
                                                                                                                                                               ((
                                                                                                                                                                !                                                               E
                                                                           E    E        (
                                                                                         !       (
                                                                                                 !E
                                                                                                  (
                                                                                                  !!
                                                                                                   (              E!
                                                                                                                   (EE E    !
                                                                                                                            ((!
                                                                                                                             !
                                                                                                                             !
                                                                                                                             (
                                                                                                                             E (
                                                                                                                                                  EE(
                                                                                                                                                    !
                                                                                                                                                    E!
                                                                                                                                                     (
                                                                                                                                                     !
                                                                                                                                                     (
                                                                                                                                                     (
                                                                                                                                                     !
                                                                                                                                                     (
                                                                                                                                                     !
                                                                                                                                                      (
                                                                                                                                                      !
                                                                                                                                                      !
                                                                                                                                                      (
                                                                                                                                                      !
                                                                                                                                                      (
                                                                                                                                                      E
                                                                                                                                                      (
                                                                                                                                                      !
                                                                                                                                                      E
                                                                                                                                                      E(
                                                                                                                                                       !
                                                                                                                                                       (
                                                                                                                                                       !
                                                                                                                                                       E
                                                                                                                                                       !
                                                                                                                                                       (
                                                                                                                                                       !
                                                                                                                                                       (
                                                                                                                                                       !
                                                                                                                                                       (
                                                                                                                                                       E
                                                                                                                                                       (
                                                                                                                                                       !
                                                                                                                                                       E
                                                                                                                                                        (
                                                                                                                                                        !
                                                                                                                                                        (
                                                                                                                                                        !
                                                                                                                                                        E
                                                                                                                                                        (
                                                                                                                                                        !
                                                                                                                                                        (
                                                                                                                                                        !E
                                                                                                                                                         (
                                                                                                                                                         !
                                                                                                                                                         (
                                                                                                                                                         !
                                                                                                                                                         E
                                                                                                                                                          !
                                                                                                                                                          (
                                                                                                                                                          (
                                                                                                                                                          !
                                                                                                                                                          !
                                                                                                                                                          (  E !E(EE !
                                                                                                                                                                 !
                                                                                                                                                                 E     (
                                                                                                                                                                       ( !
                                                                                                                                                                       !   E
                                                                                                                                                                           (E!
                                                                                                                                                                            !
                                                                                                                                                                            ( (              E                            EEE
                                                                                                (!
                                                                                                ! (                            (!
                                                                                                                               ! ( !            !
                                                                                                                                                (     (
                                                                                                                                                      !E
                                                                                                                                                       (
                                                                                                                                                       !(
                                                                                                                                                        !(
                                                                                                                                                         !
                                                                                                                                                        EE E    (  !
                                                                                                                                                                   ( !
                                   Panel PSUs
                                                                                    !
                                                                                    (                              E!
                                                                                                                    (!(E E !   (      (        E E   E
                                                                                                                                                     EEE
                                                                                                                                                       E    E
                                                                                                                                                            (
                                                                                                                                                            !E
                                                                                                                                                             (
                                                                                                                                                             !( E
                                                                                                                                                              !
                                                                                                                                                              E         ( ! (E E   E
                           !
                           (                                                      (E E            !
                                                                                                  ((
                                                                                                   !         E !
                                                                                                               (  E(
                                                                                                                   !           E
                                                                                                                               !
                                                                                                                               (            (
                                                                                                                                            !     E  E   E  !
                                                                                                                                                            (E                  !
                                                                                                                                                                                (
                                                                                (
                                                                                !
                                                                                  !    !(
                                                                                       (!(!
                                                                                         !       E
                                                                                                 !
                                                                                                 (E !  (E
                                                                                                       !
                                                                                                      (E   (
                                                                                                           !
                                                                                                           E                           E         E E               E
                                                                                                                                                                           E        E ! (E                   E            E
                                                                                                                                         !
                                                                                                                                         (                                                       !
                                                                                                                                                                                                 (                        !
                                                                                                                                                                                                                          (
                                   Rainfall Stations                                   (!
                                                                                       !   (
                                                                                         ( E     E
                                                                                                 !
                                                                                                 ( ! !
                                                                                                     (
                                                                                                   ( !  (                                             (
                                                                                                                                                      !                    !
                                                                                                                                                                           (E     (
                                                                                                                                                                                  !  E    E                               E
                           E                                                                            (!
                                                                                                        ! (                                 E(
                                                                                                                                             !E
                                                                                                                                              (
                                                                                                                                              !
                                                                                                                                              E   E      !
                                                                                                                                                         (           E       !
                                                                                                                                                                             (           E                (
                                                                                                                                                                                                          !
                                                                                              !
                                                                                              (                                              !
                                                                                                                                             (!
                                                                                                                                              (                           (
                                                                                                                                                                          !                                (
                                                                                                                                                                                                           !
                                                                                              !
                                                                                              (                                                             E ! (                             E!         EE
                                                                                                                                                                                                         !
                                                                                                                                                                                                         (           EE
                                                                                                                                             E                       !
                                                                                                                                                                     ( E E !    ( !(( E
                                                                                                                                                                                    !          (       ( E E!
                                                                                                                                                                                                       !        (
                                                                                                                                                              !
                                                                                                                                                              (        (
                                                                                                                                                                       !                      !
                                                                                                                                                                                              (             E ! (       !
                                                                                                                                                                                                                        (
                                                                                                                                                                                                                        E
                                                                                                                                      (
                                                                                                                                      !                                !
                                                                                                                                                                       (                      !
                                                                                                                                                                                              ( EE ! (        (
                                                                                                                                                                                                              !      (
                                                                                                                                                                                                                     !         E
                                                                                                                                            E E              (
                                                                                                                                                             !                E
                                                                                                                                                                                      E
                                                                                                                                                                                           (
                                                                                                                                                                                           !           E
                                                                                                                                                                                                             E !         E
                                                                                                                                        (
                                                                                                                                        !
                                                                                                                                            (
                                                                                                                                            !           ( E !
                                                                                                                                                        !      ( E                                             (
                                                                                                                                                                                                               !  ( E       (
                                                                                                                                                                                                                            !
                                                                                                                                          (
                                                                                                                                          !
                                                                                                                                          ! E
                                                                                                                                          ( E!                                                        !
                                                                                                                                                                                                      (     EE  (!
                                                                                                                                                                                                                 (        !
                                                                                                                                                                                                                          (      E
                                                                                                                                      (E
                                                                                                                                     (!
                                                                                                                                     !   !
                                                                                                                                         (
                                                                                                                                         (
                                                                                                                                         !
                                                                                                                                         !
                                                                                                                                         (    (      ( E!
                                                                                                                                                     !     ( !
                                                                                                                                                           (
                                                                                                                                                           !   (       E       (
                                                                                                                                                                               !
                                                                                                                                                                                    (
                                                                                                                                                                                    !
                                                                                                                                                                                                 E            EE
                                                                                                                                                                                                               (
                                                                                                                                                                                                               !
                                                                                                                                                                                                                !      (
                                                                                                                                                                                                                       !      !
                                                                                                                                                                                                                              (
                                                                                                                                                                  (
                                                                                                                                                                  !                                       E E
                                                                                                                                             E E                   (
                                                                                                                                                                   ! !
                                                                                                                                                                     (
                                                                                                                                                                     (
                                                                                                                                                                     !        E         E !(   !
                                                                                                                                                                                               (      !
                                                                                                                                                                                                      ( E                     ( !
                                                                                                                                                                                                                              !
                                                                                                                                                                                                                              E    (!
                                                                                                                                                                                                                                    (
                                                                                                                                               !
                                                                                                                                               ( !  (!( E !
                                                                                                                                                      !
                                                                                                                                                      (
                                                                                                                                                      !
                                                                                                                                                      (
                                                                                                                                                             (E     E
                                                                                                                                                                         !
                                                                                                                                                                         (
                                                                                                                                                                            ( !
                                                                                                                                                                            !  ( !(                      !
                                                                                                                                                                                                         (E
                                                                                                                                                                                                          E  E            !
                                                                                                                                                                                                                          (
                                                                                                                                                                                                                              ( E
                                                                                                                                                                                                                              !    E
                                                                                                                                                                   !
                                                                                                                                                                   (( !
                                                                                                                                                                    !    E                (! (
                                                                                                                                                                                             E !  (          !
                                                                                                                                                                                                             (
                                                                                                                                                                                                             !
                                                                                                                                                                                                             (
                                                                                                                                                                                                             !
                                                                                                                                                                                                             (!
                                                                                                                                                                                                              (               E
                                                                                                                                                      E !
                                                                                                                                                      (
                                                                                                                                                      !  (              (
                                                                                                                                                                                  !
                                                                                                                                                                                  ((
                                                                                                                                                                                   !!
                                                                                                                                                                                    (     !         (
                                                                                                                                                                                                    !      !
                                                                                                                                                                                                             E
                                                                                                                                                                                                           ( ! ( E          (
                                                                                                                                                                                                                            !
                                                                                                                                                                   ( !
                                                                                                                                                                   !
                                                                                                                                                                   !
                                                                                                                                                                   (    (
                                                                                                                                                                        !
                                                                                                                                                                        (E(        ( E!
                                                                                                                                                                                   !   (!           E        E        !
                                                                                                                                                                                                                      (        !
                                                                                                                                                                                                                               (
                                                                                                                                                                    ( !
                                                                                                                                                                    E
                                                                                                                                                                    !    (!    !
                                                                                                                                                                               (E      ((
                                                                                                                                                                                       !(
                                                                                                                                                                                        !       !
                                                                                                                                                                                                (!
                                                                                                                                                                                                 (         (
                                                                                                                                                                                                           !
                                                                                                                                                                                                             !
                                                                                                                                                                                                             (
                                                                                                                                                                                                             !
                                                                                                                                                                                                             (
                                                                                                                                                                                                             !
                                                                                                                                                                                                             ((
                                                                                                                                                                                                              !(
                                                                                                                                                                                                               !
                                                                                                                                                                                                                   !
                                                                                                                                                                                                                   (    E!(
                                                                                                                                                                                                                                E
                                                                                                                                                                                                                                (
                                                                                                                                                                                                                                !
                                                                                                                                                                                                                                E
                                                                                                                                                                                                                                 !
                                                                                                                                                                                                                                 (
                                                                                                                                                                      (
                                                                                                                                                                      !                            E     !
                                                                                                                                                                                                         (      E      !
                                                                                                                                                                                                                       (!
                                                                                                                                                                                                                        (            !
                                                                                                                                                                                                                                     (
                                                                                                                                                                                                                      (
                                                                                                                                                                                                                      !      (
                                                                                                                                                                                                                             !E E
                                                                                                                                                                                          (!
                                                                                                                                                                                          ! ( E
                                                                                                                                                                                              !
                                                                                                                                                                                              (                                 (!
                                                                                                                                                                                                                                !  (
                                                                                                                                                                                                                                   !
                                                                                                                                                                                                                                   (
                                                                                                                                                                                            ( !
                                                                                                                                                                                            !    (           ( !
                                                                                                                                                                                                             E
                                                                                                                                                                                                             !
                                                                                                                                                                                                             E
                                                                                                                                                                                                             !
                                                                                                                                                                                                             (
                                                                                                                                                                                                             !
                                                                                                                                                                                                             (
                                                                                                                                                                                                             !
                                                                                                                                                                                                             (   (!(!
                                                                                                                                                                                                                    (
                                                                                                                                                                                                             !
                                                                                                                                                                                                             ( !
                                                                                                                                                                                                             !
                                                                                                                                                                                                             (
                                                                                                                                                                                                             (
                                                                                                                                                                                                             !   (
                                                                                                                                                                                                                 !
                                                                                                                                                                                                                 (              (E
                                                                                                                                                                                                                                !




Note: This map plots the distribution of the panel of village wards (PSUs) and the rainfall stations throughout
the country.




                                                                                                                                  24
                            Figure 4: International migration from Nepal




  India migration rate                                    Non-India migration rate
  2001                                                    2001
         0.00 - 0.01                                             0.00 - 0.01
         0.01 - 0.04                                             0.01 - 0.04
         0.04 - 0.07                                             0.04 - 0.07
         0.07 - 0.10                                             0.07 - 0.10
         >0.10                                                   >0.10




  India migration rate                                    Non-India migration rate
  2008                                                    2008
         0.00 - 0.01                                             0.00 - 0.01
         0.01 - 0.04                                             0.01 - 0.04
         0.04 - 0.07                                             0.04 - 0.07
         0.07 - 0.10                                             0.07 - 0.10
         >0.10                                                   >0.10




  India migration rate                                    Non-India migration rate
  2010                                                    2010
         0.00 - 0.01                                             0.00 - 0.01
         0.01 - 0.04                                             0.01 - 0.04
         0.04 - 0.07                                             0.04 - 0.07
         0.07 - 0.10                                             0.07 - 0.10
         >0.10                                                   >0.10




Source: Author’s calculation from the PSU level panel assembled from various surveys. See text for details.
Note: These maps show the migration rates to India (Left panel) and non-India destinations (Right panel). The
top row shows the rates for 2001, middle row for 2008, and bottom row for 2010.




                                                     25
                          Figure 5: Distribution of monsoon rainfall in 2001, 2008, 2010




                    1.5
                    1
              Density
                    .5
                    0




                           -4              -2                    0                2           4
                                                Normalized daily monsoon 1 yr ago

                                                 2001             2008                2010


Source: Author’s calculation from the PSU level panel assembled from various surveys and the rainfall data. See
text for details.
Note: This figures shows the distribution of past year’s monsoon rainfall for the survey years 2001, 2008, and
2010.




                                                            26
                         Figure 6: Deaths from conflict in 1996-2001 and 2001-2006




               Conflict Intensity
               (Deaths per 1000 population)
               1996-2001 (average 0.1)
                    0
                    0 - 0.5
                    0.5 - 1.0
                    1.0 - 2.0
                    2.0 - 4.0




               Conflict Intensity
               (Deaths per 1000 population)
               2002-2006 (average 0.6)
                    0
                    0 - 0.5
                    0.5 - 1.0
                    1.0 - 2.0
                    2.0 - 4.0




Source: Author’s calculations from the Conflict data from INSEC. See text for details.
Note: The top map shows the conflict intensity during 1996-2001. The bottom map shows the conflict intensity
during 2002-2006 period. Conflict intensity measures the number of conflict related deaths normalized per 1000
population in 1991.




                                                    27
7.2    Tables


                          Table 1: International migration and remittance
                                Migrant/Population       Remittance income
                        Year All India Non-India             % of GDP
                        1961 3.49
                        1981 2.68 2.48          0.19
                        1991 3.56 3.17          0.37            1.5*
                        2001 3.41 2.61          0.78             2.4
                        2011 7.43 2.80          4.63            22.4
Source: Migrant/Population shares from the Census reports for respective years; Remittance as a share of GDP
from the World Development Indicator database (The World Bank).
Note: * Figure for 1993 (earlier figure not available)



                             Table 2: Migration costs and incomes
 Destination         PC monthly cons Cost of migration Monthly income                   Monthly savings
                       (000 NPR)         (000 NPR)           (000 NPR)                    (000 NPR)
 Nepal                    2.90
 India                                       6.25                6.4                           3.86
 Gulf Countries                            102.92               16.21                          11.95
 Malaysia                                  133.67               13.51                          9.87
Source: Author’s calculations from the Nepal Migration Survey, 2009.
Note: Self reported numbers by household members in Nepal. (These numbers are very similar to self-reports by
returnees about their own income while abroad). Per-Capita (PC) consumption for Nepal from the Nepal Living
Standards Survey-III, 2010




                                                     28
                                      Table 3: Summary statistics
                                                          All     Rural            Urban
                                                          (1)      (2)              (3)
                 Year 2001
                 Migration rate abroad                        0.035     0.036      0.027
                                                             (0.004)   (0.004)    (0.003)
                 Migration rate to India                      0.026     0.030      0.011
                                                             (0.003)   (0.004)    (0.002)
                 Migration rate to non-India                  0.008     0.006      0.016
                                                             (0.001)   (0.001)    (0.002)
                 Conflict related deaths per 1000              0.110     0.124      0.049
                                                             (0.013)   (0.016)    (0.005)
                 Normalized daily monsoon 1 yr ago            1.441     1.472      1.301
                                                             (0.018)   (0.021)    (0.032)
                 Normalized daily monsoon 2 yr ago            1.381     1.374      1.409
                                                             (0.020)   (0.022)    (0.045)
                 Year 2008
                 Migration rate abroad                        0.081     0.085       0.063
                                                             (0.003)   (0.004)    (0.004)
                 Migration rate to India                      0.043     0.048      0.021
                                                             (0.003)   (0.003)    (0.003)
                 Migration rate to non-India                  0.037     0.036       0.042
                                                             (0.002)   (0.002)    (0.003)
                 Conflict related deaths per 1000               0.592    0.638       0.392
                                                             (0.028)   (0.034)    (0.029)
                 Normalized daily monsoon 1 yr ago            0.590      0.648      0.337
                                                             (0.028)   (0.030)    (0.072)
                 Normalized daily monsoon 2 yr ago            -1.410    -1.401     -1.446
                                                             (0.020)   (0.023)    (0.032)
                 Year 2010
                 Migration rate abroad                        0.096     0.098       0.086
                                                             (0.004)   (0.004)    (0.006)
                 Migration rate to India                      0.046     0.051      0.023
                                                             (0.003)   (0.004)    (0.003)
                 Migration rate to non-India                  0.050     0.047       0.063
                                                             (0.003)   (0.003)    (0.005)
                 Conflict related deaths per 1000               0.000    0.000       0.000
                                                                (.)       (.)        (.)
                 Normalized daily monsoon 1 yr ago            -0.639    -0.587     -0.865
                                                             (0.036)   (0.042)    (0.059)
                 Normalized daily monsoon 2 yr ago            0.559      0.627      0.257
                                                             (0.025)   (0.029)    (0.045)
Source: Author’s calculation from the village level panel assembled from Census 2001, NLFS 2008, and NLSS-III
2010.
Note: Means are weighted by the NLSS-III sampling weights.




                                                     29
             Table 4: Previous year’s rainfall and International          migration (2001-2010)
                                                          All              Rural     Urban
                                                          (1)               (2)       (3)
                All International Migration
                Normalized daily monsoon 1 yr ago 0.007*                     0.007      0.005
                                                        (0.004)             (0.005)    (0.005)
                Observations                             1356                 936        420
                Adj R-squared                            0.276               0.269      0.326
                Migration to India
                Normalized daily monsoon 1 yr ago 0.008**                  0.008**     0.008*
                                                        (0.003)            (0.004)     (0.004)
                Observations                             1356                936         420
                Adj R-squared                            0.052              0.050       0.081
                Migration to non-India
                Normalized daily monsoon 1 yr ago -0.001                     -0.001     -0.002
                                                        (0.002)             (0.003)    (0.004)
                Observations                             1356                  936       420
                Adj R-squared                            0.350                0.364     0.313
Source: Author’s calculation from the PSU level panel assembled from various surveys. See text for details.
Note: This table shows the impact of past year’s monsoon on migration rates using Equation (1). The first
column shows the effect for all the PSUs. The second and third columns split the sample into rural and urban
areas respectively. The top panel shows the effect on all international migration. The second panel shows the
effect on migration rates to India. The third panel shows the effect on migration rates to non-India destinations.
All regressions are weighted by NLSS-III sampling weights. The explanatory variable measures previous year’s
average daily rainfall in the village during the monsoon months normalized by the historic rainfall in the village.
Only recent years (2001, 2008, 2010) data used. Standard errors, reported in parenthesis, are clustered at the
district level. ∗ : p < 0.1;∗∗ : p < 0.05;∗∗∗ : p < 0.01.




                                                        30
             Table 5: Previous year’s rainfall and International migration (1981-2010)
                                            All                          Rural                     Urban
                                      (1)           (2)            (3)           (4)         (5)           (6)
  All International Migration
  Normalized daily monsoon 0.007**                 0.005         0.008*       0.006         0.003      0.002
  1 yr ago                    (0.003)             (0.003)        (0.004)     (0.004)       (0.004)    (0.004)
  Observations                  2260               2260           1560        1560           700        700
  Adj R-squared                 0.355              0.359          0.346       0.350         0.414      0.414
  Migration to India
  Normalized daily monsoon 0.009***            0.007***          0.009***   0.008***       0.006**    0.005**
  1 yr ago                    (0.003)           (0.002)           (0.003)    (0.003)       (0.002)    (0.002)
  Observations                  2260             2260              1560       1560           700        700
  Adj R-squared                 0.071            0.119             0.076      0.117         0.060      0.185
  Migration to non-India
  Normalized daily monsoon     -0.002              -0.002         -0.001      -0.001        -0.003     -0.002
  1 yr ago                    (0.002)             (0.002)        (0.002)     (0.003)       (0.003)    (0.003)
  Observations                  2260                2260           1560        1560           700       700
  Adj R-squared                 0.415               0.466         0.417        0.474         0.426     0.465
Source: Author’s calculation from the PSU level panel assembled from various surveys. See text for details.
Note: This table shows the impact of past year’s monsoon on migration rates using Equation (1). The first pair
of columns show the effect for all the PSUs. The second and third pairs of columns split the sample into rural
and urban areas respectively. The even numbered columns add linear time trends for each region. The top panel
shows the effect on all international migration. The second panel shows the effect on migration rates to India.
The third panel shows the effect on migration rates to non-India destinations. All regressions are weighted by
NLSS-III sampling weights. The explanatory variable measures previous year’s average daily rainfall in the village
during the monsoon months normalized by the historic rainfall in the village. Data for all years (1981, 1991, 2001,
2008, 2010) are used with district level rates for 1981 and 1991. Standard errors, reported in parenthesis, are
clustered at the district level. ∗ : p < 0.1;∗∗ : p < 0.05;∗∗∗ : p < 0.01.




                                                            31
                 Table 6: Previous rainfall and International migration (1981-2010)
                                           All                           Rural                      Urban
                                     (1)           (2)             (3)           (4)          (5)           (6)
 All International Migration
 Normalized daily monsoon    0.006*               0.005          0.007*        0.006        0.002        0.002
 1 yr ago                    (0.003)             (0.003)         (0.004)     (0.004)       (0.004)      (0.004)
 Normalized daily monsoon      0.002              0.000           0.002       -0.000        0.003        0.002
 2 yr ago                    (0.003)             (0.002)         (0.003)     (0.003)       (0.003)      (0.003)
 Observations                  2260               2260            1560         1560          700          700
 Adj R-squared                 0.355              0.359           0.346        0.350        0.414        0.413
 p-value of joint test         0.123              0.332           0.185        0.344        0.359        0.431
 Migration to India
 Normalized daily monsoon 0.007***               0.007***       0.007**      0.007**       0.004*        0.003*
 1 yr ago                    (0.003)              (0.002)        (0.003)     (0.003)       (0.002)      (0.002)
 Normalized daily monsoon 0.008***                 0.002        0.008***      0.001       0.007***      0.004**
 2 yr ago                    (0.002)              (0.002)        (0.003)     (0.002)       (0.002)      (0.002)
 Observations                  2260                2260           1560        1560           700           700
 Adj R-squared                 0.079               0.119          0.082       0.117         0.084         0.192
 p-value of joint test         0.001               0.008          0.005       0.035         0.001        0.000
 Migration to non-India
 Normalized daily monsoon     -0.001              -0.002          0.000       -0.001         -0.002      -0.001
 1 yr ago                    (0.002)             (0.002)         (0.003)     (0.003)        (0.003)     (0.003)
 Normalized daily monsoon -0.006***               -0.002        -0.006***     -0.001       -0.005**      -0.002
 2 yr ago                    (0.001)             (0.001)         (0.002)     (0.001)        (0.002)     (0.002)
 Observations                  2260                2260            1560        1560            700         700
 Adj R-squared                 0.422               0.466          0.424        0.474          0.429       0.465
 p-value of joint test         0.000               0.146          0.001        0.472          0.018       0.253
Source: Author’s calculation from the PSU level panel assembled from various surveys. See text for details.
Note: This table shows the impact of past two years’ monsoon on migration rates using Equation (1). The first
pair of columns show the effect for all the PSUs. The second and third pairs of columns split the sample into rural
and urban areas respectively. The even numbered columns add linear time trends for each region. The top panel
shows the effect on all international migration. The second panel shows the effect on migration rates to India.
The third panel shows the effect on migration rates to non-India destinations. All regressions are weighted by
NLSS-III sampling weights. The explanatory variable measures previous year’s average daily rainfall in the village
during the monsoon months normalized by the historic rainfall in the village. Data for all years (1981, 1991, 2001,
2008, 2010) are used with district level rates for 1981 and 1991. Standard errors, reported in parenthesis, are
clustered at the district level. ∗ : p < 0.1;∗∗ : p < 0.05;∗∗∗ : p < 0.01.




                                                           32
                Table 7: Future rainfall and International migration (1981-2008)
                                        All                   Rural                Urban
                                  (1)         (2)         (3)       (4)        (5)       (6)
   All International Migration
   Normalized daily monsoon 0.001            0.000      -0.000    -0.001      0.002    0.003
   1 yr later                   (0.002) (0.002)        (0.003) (0.003)      (0.002) (0.002)
   Observations                  1808        1808        1248      1248        560      560
   Adj R-squared                 0.296       0.300      0.286      0.289      0.407    0.414
   Migration to India
   Normalized daily monsoon 0.002            0.002      0.001      0.001      0.003    0.004
   1 yr later                   (0.003) (0.002)        (0.003) (0.003)      (0.002) (0.002)
   Observations                  1808        1808        1248      1248        560      560
   Adj R-squared                 0.058       0.093      0.062      0.089      0.068    0.231
   Migration to non-India
   Normalized daily monsoon -0.001          -0.002      -0.002 -0.002*       -0.001    -0.001
   1 yr later                   (0.001) (0.001)        (0.001) (0.001)      (0.001) (0.001)
   Observations                  1808        1808        1248      1248        560      560
   Adj R-squared                 0.422       0.474      0.425      0.481      0.432    0.472
Source: Author’s calculation from the PSU level panel assembled from various surveys. See text for details.
Note: This table shows the impact of the following year’s monsoon on migration rates using Equation (1). The
first pair of columns show the effect for all the PSUs. The second and third pairs of columns split the sample
into rural and urban areas respectively. The even numbered columns add linear time trends for each region.
The top panel shows the effect on all international migration. The second panel shows the effect on migration
rates to India. The third panel shows the effect on migration rates to non-India destinations. All regressions
are weighted by NLSS-III sampling weights. The explanatory variable measures previous year’s average daily
rainfall in the village during the monsoon months normalized by the historic rainfall in the village. Data for
years 1981, 1991, 2001, and 2008 are used with district level rates for 1981 and 1991. Year 2010 excluded because
rainfall data not available for 2011. Standard errors, reported in parenthesis, are clustered at the district level.
∗
  : p < 0.1;∗∗ : p < 0.05;∗∗∗ : p < 0.01.




                                                        33
                                   Table 8: Conflict and migration
                                                        All    Rural                     Urban
                                                        (1)     (2)                       (3)
                    All International Migration
                    Conflict related deaths per 1000 0.008      0.005                    0.031**
                                                      (0.006) (0.006)                   (0.015)
                    Observations                       1356     936                       420
                    Adj R-squared                      0.275   0.267                     0.335
                    Migration to India
                    Conflict related deaths per 1000 0.005      0.003                     0.020
                                                      (0.005) (0.006)                   (0.012)
                    Observations                       1356     936                       420
                    Adj R-squared                      0.046   0.046                     0.077
                    Migration to non-India
                    Conflict related deaths per 1000 0.003      0.002                     0.011
                                                      (0.003) (0.003)                   (0.008)
                    Observations                       1356     936                       420
                    Adj R-squared                      0.351   0.365                     0.314
Source: Author’s calculation from the PSU level panel assembled from various surveys. See text for details.
Note: This table shows the impact of conflict intensity in district on migration rates using Equation (1). The first
column shows the effect for all the PSUs. The second and third columns split the sample into rural and urban
areas respectively. The top panel shows the effect on all international migration. The second panel shows the
effect on migration rates to India. The third panel shows the effect on migration rates to non-India destinations.
All regressions are weighted by NLSS-III sampling weights. Data for years 2001, 2008, and 2010 are used in this
estimation. The conflict variable measures the conflict related deaths in the district during 1996-2001 for 2001,
during 2002-2006 for 2008, and is set to 0 for 2010. Standard errors, reported in parenthesis, are clustered at the
district level. ∗ : p < 0.1;∗∗ : p < 0.05;∗∗∗ : p < 0.01.



                      Table 9: Growth in destination and migration
                                     All destinations             Non-India destinations
                                 (1)        (2)       (3)        (4)       (5)         (6)
 1 yr growth in manu           0.015**                       0.033***
 and cons CO2 emission         (0.007)                        (0.009)
 2 yr growth in manu                      0.012*                        0.026***
 and cons CO2 emission                   (0.006)                         (0.007)
 3 yr growth in manu                               0.011**                         0.011***
 and cons CO2 emission                             (0.005)                          (0.004)
 Observations                   4068       4068      4068       2712      2712        2712
 Adj R-squared                  0.179      0.178    0.178      0.466      0.458      0.442
Source: Author’s calculation from the PSU level panel assembled from various surveys. See text for details.
Note: This table shows the impact of recent growth in construction and manufacturing sectors in the destination
categories on migration rates using Equation (2). The first three columns include India, Malaysia, and the Persian
Gulf countries (Qatar, Saudi Arabia, and the UAE) in the destination categories. The last three columns exclude
India. All regressions are weighted by NLSS-III sampling weights. Data for years 2001, 2008, and 2010 are
used in this estimation. The explanatory variable measures the growth in CO2 emission by manufacturing and
construction in each of the destinations over the previous 1, 2 and 3 year period preceding the survey years.
Standard errors, reported in parenthesis, are clustered at the district level. ∗ : p < 0.1;∗∗ : p < 0.05;∗∗∗ : p < 0.01.


                                                          34
A      Calibrating migration thresholds
The interpretation of the empirical results in this paper depends upon the wealth distribution
and the position of the affordability and desirability thresholds. In this subsection, I describe
how the cut-offs for Figure 2 are calibrated. For this exercise, I set p = 1, so that the rich get
the full peace premium.
    Wealth, w, used in the theoretical framework is a measure of pre-migration permanent wealth
and is separate from consumption. I use the monetary value of total amount of land owned by
the household as a measure of w. I estimate the distribution of this measure in the Nepal Living
Standards Survey - I (NLSS-I) of 1995/96 amongst the landed population using a log-normal
approximation. Then I scale this distribution to its 2010 level by adjusting for inflation as well as
allowing real wealth to grow between these two periods. I used the growth rate of real per-capita
consumption amongst landed households without migrants between these periods. Similarly, I
get distributions of per-capita consumption and farm income from NLSS-I after appropriate
scaling to their 2010 level.
    Then I simulate a dataset of 100,000 households with wealth, consumption and farm income
drawn randomly from this distribution. I set r equal to the difference between farming income
net of consumption. The costs and earnings from migration are as reported in Table 2 scaled
by the CPI to year 2010. I assume a migration episode of 2 years. That is, households choose
between 2 episodes of migration to India, each lasting 9 months or one episode of migration
to the Gulf countries or Malaysia, lasting 2 years. I use NLSS-III to estimate income loss in
farming resulting from migration of a member20 .
    Using these calibrations and the simple cut-off rule implied by the theoretical framework,
I classify each household as a non-migrant household, or a household with India migrant or a
household with non-India migrant. I find that about 27 percent of the households choose to
have a migrant in India, and 19 percent choose to have a migrant outside India. These numbers
are only slightly bigger than the household migration rates observed in NLSS-III in 2010. In
the NLSS-III data, about 15 percent of the households have a migrant in India and about 16
percent of the households have a migrant in non-India destinations.
    In this simple simulation, I find that 19 percent of the households are below the affordability
threshold for India and 23 percent are above the desirability threshold for the Gulf and Malaysia
migration. The wealth distribution used for simulation and the resulting average thresholds of
migration to India and non-India destinations is plotted in Figure 2.
    In this simulated data (with a sample size of 5,000, the expected size of NLSS-III survey), I
find that a random drop of Rs 7,500 (USD 100) increases India migration rate significantly by
3.7 percentage points (p-value of 0.004) whereas the cash transfer has an insignificant impact
on migration to non-India destinations (p-value of 0.850). This is also consistent with what I
find empirically21 .
   20
      The estimating equation is log (Yi ) = βMi + εi where Mi indicates whether a household has a foreign migrant,
and Yi is total farm income for the household. I estimate this specification on a sample of household with
agricultural land. The OLS estimate of β is −0.088 with standard error of (0.041).
   21
      However, the outcome variables are not identical in the simulated data and the one used in the paper, so
I cannot compare coefficients directly. In the simulated data, outcome variable is an indicator of whether the
household has a migrant or not. In the paper, the outcome variable is a ratio of migrant to population in the
village.




                                                        35
B     Monetizing rainfall and conflict
Data and method
In this exercise, I try to monetize the conflict and rainfall shocks using three different cross-
sectional surveys with consumption and income measures. I use three different waves of Nepal
Living Standards Surveys (NLSS) conducted in 1995/96, 2003/04 and 2010 by the Central
Bureau of Statistics using comparable instruments.
    In the pooled cross-sectional data, I estimate

                                    yit = βXit + γt + µj + εit

where yit is the consumption per-capita or farm income per-capita for PSU i observed in year
t, Xit is the conflict or rainfall measure for the PSU for year t; γt captures the survey-year fixed
effects. I employ different fixed effects to essentially create pseudo-panel data at the j × t level.
I use Development Region × Ecological belt which results in 15 groups, district fixed effects
which results in 74 groups and also use synthetic clusters of 97 clusters. I allow standard errors
to be correlated across observations within the same group (j ) and across survey periods.
    As these surveys were cross-sections, the probability that the same PSU is observed more
than once is quite low. However, several PSUs that are close to each other are observed across
different periods. I therefore create a pseudo-panel at a level lower than the administrative
districts by creating synthetic clusters of PSUs that are close to each other. A group of three (or
more) PSUs within 10km of each other are in the same synthetic cluster. Any PSU within 10km
of at least two of the PSUs in a cluster also belongs to the same cluster. In cases where this
algorithm maps one PSU to more than one cluster, I assign it to the cluster with fewer PSUs.
This method gives me a total of 97 synthetic clusters and covers 70 percent of the PSUs in the
dataset. I dropped the unassigned PSUs from this specification. Since the 70 percent of the
PSUs which are mapped to synthetic clusters are different from those that were not-mapped,
this specification would produce biased estimates of β . As a simple check, I test whether income
measures differ by whether a PSU is included in the synthetic cluster sample or not. I find
that farm income per-capita is not different (p-value of 0.31) whereas consumption per-capita
is different (p-value of 0.04) by their inclusion status in the synthetic cluster sample. When I
assign higher level clusters to unassigned PSUs, I get qualitatively similar results as in other
specifications with fixed effects (and clustering) at higher levels.
    The measure of rainfall is identical to the measure used in the rest of the text adapted to
different survey years. Conflict measures are set to 0 for NLSS-I and NLSS-III (years 1995/96
and 2010) whereas it measures conflict related deaths in a given district between 1996 and 2002
normalized by district population in 1991 for NLSS-II (year 2003/04). Since the conflict variable
is measured at the district level, I only use specification with fixed effects at the district or a
higher aggregation. All of my consumption and income measures are converted to 2010 prices.
    Table B.1 shows the result of the estimation restricted to households that do not receive any
remittance income. I use column (3) as my preferred specification for the income equivalent of
rainfall shock and column (5) as my preferred specification for the income equivalent of conflict
shock.
    The theoretical framework postulates that the wealthy suffer more from conflict. However,
Table B.1 only looks at the average effect of conflict on consumption. I use quantile regressions to
investigate whether conflict affects consumption differently in different parts of the consumption
distribution. Figure B.1 shows the estimates of the same equation at different consumption

                                                36
quantiles. Indeed, conflict does not affect the consumption of poorer households but reduces
consumption of the richer households. At the topmost decile, the effect is larger than Rs 5,000,
almost five times the effect at the median.

Tables and figures


                         Table B.1: Income effects of conflict and rainfall
                                         Farm income per-capita       Consumption per-capita
                                          (1)      (2)      (3)          (4)    (5)     (6)
    Normalized daily monsoon 1 yr ago 2051** 1921 2444**                2133   2323     484
                                         (880) (1481) (1034)          (1297) (1955) (858)
    Observations                         8081     8081     5207        8081    8081    5207
    Adj R-squared                        0.110    0.133    0.101       0.063  0.140    0.149
    Conflict related deaths per 1000                                    -4719 -3488**
                                                                      (2839) (1477)
    Observations                                                       8081    8081
    Adj R-squared                                                      0.065  0.141
    Year FE                                Y        Y        Y            Y      Y       Y
    Region x Belt FE                       Y                              Y
    District FE                                     Y                            Y
    Synthetic cluster FE                                     Y                           Y
∗
  : p < 0.1;∗∗ : p < 0.05;∗∗∗ : p < 0.01. Standard errors reported in parenthesis clustered at the level of the
group fixed effects. Regression run in pooled Nepal Living Standards Survey of 1995/96, 2003/04 and 2010 and
are weighted by their respective sampling weights. Sample further restricted to households that do not receive
any remittance income. Conflict variable set to 0 for 1995/1996 and 2010 and counts the deaths from 1996-2002
for year 2003/2004. All consumption and income deflated to 2010 prices. All regressions have survey year fixed
effects.
Columns (1) and (4) have Development region × Ecological belt fixed effects (15 groups). Columns (2) and (5)
have district fixed effects (74 districts). Columns (3) and (6) have synthetic cluster fixed effects (97 clusters).
Synthetic clusters group PSUs into one cluster if they are within 10km of each other.




                                                      37
                            Figure B.1: Effect of conflict on different quantiles of consumption




                                0
                    Effect of conflict on consumption
             -10000 -8000 -6000 -4000 -2000




                                                        .1   .3               .5          .7   .9
                                                                  Quantile of consumption

                                                                     Effect           95% CI


Note: Bootstrap standard errors used to estimate the confidence interval. Regression estimates are unweighted.




                                                                        38