Social change, Out-migration and Exit from Agriculture Dirgha J. - - PDF document

1

Social change, Out-migration and Exit from Agriculture

Dirgha J. Ghimire

William G. Axinn Prem Bhandari Population Studies Center Institute for Social Research

University of Michigan * Direct correspondence to Dirgha J. Ghimire at Institute for Social Research, the University of

Michigan, 426 Thompson St., Ann Arbor, MI 48106-1248, email: nepdjg@umich.edu. The authors thank all the staff of the Institute for Social and Environmental Research in Nepal. Finally, the authors also thank the Economic and Social Research Council under Award No. ES/L012065/1, DFID-ESRC Growth Research Programme for their generous support for this research. All errors and omissions in this manuscript are the sole responsibility of the authors.

2

Abstract Though international labor migration has become widespread, there is relatively little scientific evidence regarding the consequences for economic change in sending countries in general, and in the densely populated rural agricultural areas of Asia in particular. We examine the influence of outmigration and migrant remittances on agriculture change in Nepal. The literature offers two opposing views of the consequences of outmigration for agriculture change: (1) loss of farm labor reduces engagement in agriculture, vs. (2) loosening credit constraints from migrant remittances increases engagement in agriculture. Of course it is entirely possible that both of these mechanisms could operate at the same time. Using multilevel dynamic models we estimate the effect of outmigration and remittance by members of agricultural households on subsequent exit of those households from agriculture. The results show that outmigration increases the hazard of exit from agriculture. The amount of remittance, on the other hand, substantially decreases the hazard of exit from agriculture. These results are robust against several key variations in model specification.

3

Social change, Out-migration and exit from Agriculture International labor migration has become widespread, with important social and economic consequences for both sending and receiving populations. In 2015, 244 million people,

• r 3.3 percent of the world's population, lived outside their country of origin (UN 2016).

Unfortunately there is relatively little scientific evidence regarding the social and economic consequences of this migration for the sending populations. Massive labor out-migrations from the densely populated rural agricultural areas of Asia have the potential to create watershed changes in engagement in agriculture, the social organization of productive activities and the food supply resulting from those activities. But long term longitudinal studies of these sending populations are rare so these potential consequences remain undocumented. Here we use long term, multilevel panel data from rural Nepal to provide new information regarding the consequences of international labor migration for the sending populations. In many rural agrarian societies—which are home to a majority of the world’s population—both agriculture and labor outmigration are key livelihood strategies. More importantly, consequences of the massive international migration on poor agricultural setting are a primary concern in the policy arena14. A large body of literature has now documented important multidimensional influence of migration on subsistence agriculture. We focus on one specific, but crucial consequence—the impact of labor outmigration on exit from farming. Even though exit of small farm holders from farming in highly mechanized agricultural settings may lead to consolidation of farm land and potentially lead to increase in agricultural productivity. In much of subsistence agricultural settings with no or low farm mechanization, exit from farming generally results into conversion of agricultural land into non-agricultural land such as fallow, regeneration of forest or into built in environment (road, public infrastructure, residential area). This loss of agriculture land has potential for lower agricultural production, primarily food,

4

which is connected with one of the world’s epidemic problems: food security. The UN World Food Programme (WFP) reports that 110 out of 210 countries—primarily poor countries with subsistence agriculture—are facing food security problems and this number is expected to continue growing15. Knowing the influence of outmigration on exit from farming in subsistence agriculture settings will provide new insights into growing global food security problem. The consequences of out migration for rural, agricultural sending areas are hypothesized in two potentially opposing ways. First, the lost labor hypothesis predicts that out migration causes shortages of farm labor which will be associated with declines in agricultural production, decreased participation, or complete exit from agriculture (de Brauw 2007; Jokisch 2002; Adhikari 2001). Second, the financial credit hypothesis predicts that the inflow of remittances from migrants will encourage investments in agricultural improvement, including agricultural technology use, stabilizing or even increasing engagement in agriculture (Mendola, 2008; Quinn 2009; Jokisch 2002; Sharma and Gurung 2009; Pant 2008; Rivera, Jose 2005; Seddon 2004; Stark and Bloom 1985). We argue that both processes could occur simultaneously. We also argue that because migrant remittances may encourage farm mechanization that creates surplus labor that may encourage additional out migration (Massey et al 1988) scientific investigation of these relationships must use carefully constructed longitudinal measures to account for the potential reciprocal associations between migrant remittance and agriculture change over time. The empirical demands for adjudicating among these associations are high, limiting the ability of previous research to investigate these relationships. Jasso et al. (2000:127) write, “[i]n perhaps no other area of demographic and social science research has there been such a persistently large gap between information needs and existing data. Consequently, many fundamental questions remain unanswered.” The lack of sufficient data for the study of migration

5

has been lamented for several decades (Bilsborrow et al. 1997; Durand and Massey 2006; Fawcett and Arnold 1987; Jasso et al. 2000; Levine et al. 1985; Massey 1987; Massey and Capoferro 2004). Comprehensive panel measures of sending communities and households with temporally ordered measures of migration, remittance and agricultural change are necessary for assessing these relationships (Adams, 2011). The study we report here is a rare exception, with the long-term household and community measures to investigate the consequence of outmigration and remittance for agricultural change. This study makes several important advances in our understanding of the relationship between international migration and agricultural change in rural agrarian settings. First, uniquely detailed neighborhood history calendar measures provide information to account for effects of preceding community context on both subsequent labor migration/and exit from agriculture. Second, household registry measures of migration with monthly precision provide the time order

• f moves across a decade of high international labor migration. Third, an innovative household

agriculture and remittance calendar provides the temporal ordering of agricultural change and reception of remittances across the same ten years. Together these unique measures from rural Nepal provide an unparalleled window into the consequences of international labor migration for agriculture in the densely populated rural regions of Asia. Theoretical framework Careful investigation of associations between international labor migration and sending population outcomes in general, or sending population agricultural practices specifically, requires theoretical consideration of two complex topics. The first, of course, is theoretical consideration of the likely mechanisms linking international migration of household members to subsequent household decisions among those who remain behind. In the case of sending

6

population agricultural practices, this means links between international migration of household members and subsequent agricultural decisions by those who remain in the household. The second is theoretical consideration of the likely factors producing the international migration itself which may also shape subsequent household decisions among those who remain behind. These factors produce an observed association between international migration and subsequent

• utcomes in the sending population that is spurious and can create a misleading understanding of

the true consequences of migration. In the case of sending population agricultural practices, this means links among factors changing prior to the international migration of household members that may influence both the migration decisions and subsequent agricultural decisions by those who remain in the household. Here we give careful theoretical to both of these complex topics. Guided by a life course perspective, we begin with the second topic. Those factors shaping both migration and subsequent sending household decisions take place earlier in the life course of individuals, even when they take place at household or communities levels (Axinn and Yabiku 2001; Elder 1998; 1985; 1995). Therefore we consider factors likely to shape international labor migration decisions that are also likely to affect subsequent household decision before we consider the additional mechanisms likely to link international migration of one household member to the subsequent decisions of other household members. Social and Economic Changes, Social Organization, and Decisions to Migrate Since the earliest days of social research, scientists have focused on the fundamental links between evolving social and economic conditions and the social organization of the people living through those changes (Dukheim; Marx). Just as in rural agrarian Asia today, the early work on changes in Europe focused on improved transportation and communication, monetization of the

7

economy, and population growth and how these change increase the division of labor in society. Durkheim argued that these factors increase the numbers of people who interact with one another, or the “moral density” of society (1984, p .257). Durkheim mentioned three specific mechanisms that predicate the division of labor: population concentration, the formation and development of cities, and improved communication and transportation (1984, p. 257-258). The framework recognizes that changes in transportation, communication, and monetization can reorganize not only production but a wide variety of other social activities as well, which greatly impacts the social organization of families and households. As activities of daily living increasingly take place outside the home and away from family, the structure of social interactions changes and alters social relationships with both household members and others

• utside the household. This reorganization of family life is the key link between macro-level

social changes and micro-level household outcomes including both migration decisions and other subsequent household decisions. Research on population mobility and migration is consistent with the conclusion that the local context of household and versus nonfamily organization shapes both migration processes (Massey and Espinosa 1997; Massey et al. 2010). Access to social organizations outside of the household in the childhood community, such as schools, employment, or support groups, is hypothesized to determine an individual’s capital endowments (from one’s parents) and achievements (attained in the respondent’s lifetime), yielding two indirect pathways to

• migration. Specifically, the level of access to non-household experiences in childhood affects an

individual’s human and social capital endowments which, in turn, affect the person’s likelihood

• f out-migration in adulthood (Massey et al. 2010). Key theories of migration argue that human

capital provides skills, knowledge, and financial resources that increase individuals’ expectations

8

• f benefits of out-migration and reduce expectations of risks (Harris and Todaro 1970).

Research also shows that human capital accumulation, such as education, labor market experience, and past migration experience, increase the likelihood of out-migration (Donato 1993, Massey and Espinosa 1997, Stark and Bloom 1985, Stark and Taylor 1991). Human and social capital endowments from one’s parents may also affect out-migration by offering an individual greater access to knowledge and financial resources (Massey et al. 2010). What does this mean in a specific international migration sending population? The prediction depends upon the conditions in that population during the period being studied (Axinn and Yabiku 2001). For the present study we investigate a setting of long-term subsistence agriculture and relatively recent changes to new transportation and communication infrastructure, monetization of the economy, and population growth. These changes are associated with community and household changes away from social organization almost entirely within households toward social organization in a wide range of nonfamily

• rganizations, including schools, employers, buses and planes, government services, nonfamily

recreation, and nonfamily social groups (Axinn and Yabiku 2001; others). In this setting, parental and childhood access to the specific nonfamily organizations that promote education, labor market experience, and past migration experience, will increase the likelihood of out- migration (Massey and Espinosa 1997; Massey et al. 2010). As explained in Massey et al. 2010, in contrast to the long-term effects of access to social

• rganizations outside the household, the short-term influences of economic and social

infrastructure on migration are generally negative: the greater the local infrastructure experienced in adulthood the less likely people are to leave the community. Community infrastructure, or services outside the household, directly decreases the likelihood of out-

9

migration by providing a broader array of local opportunities and resources—more jobs, higher wages, greater consumption opportunities. It also influences out-migration indirectly by increasing the likelihood of local employment, making people less likely to migrate in search of jobs (Massey et. al. 2010). Social and Economic Changes, Social Organization, and the Reorganization of Agriculture The key issue that makes it difficult to assess the effect of migration on reorganization of agriculture, including exits of households from agriculture, is the fact that the local context of social organization that influences migration also reshapes agriculture. Beginning the second half

• f the 20th century, along with the dramatic changes in socioeconomic and social organization,

the world agriculture has gone through dramatic reorganization including rapid mechanization and commercialization of agriculture systems with many farmers existing from farming leading to consolidation of agriculture land (Mamdani 1972; Vosti et al. 1994; Majumdar et al. 2001; Bahndari 2008). The proponents of agricultural modernization have greatly emphasized the positive aspects of this transition including potential for increase in food production and productivity, decline in food prices and overall socio-economic development (for example, Hazel and Ramaswamy 1991; Lipton 1989; Mellor 1976; Sen 1975; Vosti, Witcover, and Lipton 1994). However, a large body of literature also points towards several negative consequences of this transition such as price inflation of agricultural commodities, detrimental health effects, and negative environmental outcomes such as air and water pollution potentially leading to global warming and climate change (Biswas 1994; Cleaver 1972; Griffin 1974; Jacoby 1972; Pimentel and Pimentel 1991; Hill et al. 2009). Indeed, in many rural areas the dramatic reorganization of social and economic context greatly shaped the changes in agriculture, including exit from farming, leading to conversion of agriculture land into built in environment (Axinn and Ghimire

10

2011). The result can be decline in food production. Studies of these topics have also shown that local-level contextual characteristics can be particularly important determinants of agricultural change (Axinn and Ghimire 2011). Moreover, some of the same specific dimensions of local context we hypothesize will affect migration are also known to affect local production processes including agriculture. Thus, our understanding of the effects of migration on agricultural change may be misleading if local context is ignored. The relationships between migration and land use are embedded within a common set of contextual factors. As a result, precise specification of the relationship between migration and land use (agriculture) at the micro-level requires a clear understanding of the influence of contextual characteristics on both migration and agriculture change. Labor migration and exits from agriculture The literature offers two opposing views of the consequences of outmigration for agriculture change: (1) loss of farm labor reduces engagement in agriculture, vs. (2) loosening credit constraints from migrant remittances increases engagement in agriculture. Of course it is entirely possible that both of these mechanisms could operate at the same time. The labor loss argument suggests that outmigration negatively influences migrant-sending communities in two important ways. First, in labor intensive subsistence agriculture, the loss of family labor negatively affects household agricultural production (Adhikari 2001; de Brauw 2007; Jokisch 2002; Taylor, Rozelle, and de Brauw 2003). Second, even when migrants send much of their income back home, there is some agreement among researchers that a large proportion of this money is used for consumption rather than on productive investments, leading to economic dependency and stunted development in migrant-sending areas (Ecer, and Tompkins 2010; Koc and Onan 2004; Lipton 1980; Oberai and Singh 1980; Reichert 1981; Rempel and

11

Lobdell 1978; Taylor 1999). Thus, lost farm labor due to outmigration is likely to negatively affect agricultural productivity and ultimately result into complete exit from agriculture. In contrast, the New Economics of Labor Migration (NELM) theory argues that there is a positive impact of outmigration, and particularly of remittances, on migrant-sending areas (Stark and Bloom. 1985) 21. The NELM hypothesizes migration to be a decision made to overcome market failures that constrain local investment and production, implying that migration can raise productive investments in the origin. Adams (Adams 1998) 22 demonstrates that in Pakistan, remittances have a significant positive effect on the accumulation of irrigated and rain-fed land and productive investment assets but not on non-farm or consumption assets. Others (Ecer, and Tompkins 2010; Oberai and Singh 1980) also find that some amount of remittances is spent on productive investments such as the purchase of land and labor saving farm technology. In this paper, we investigate the possibility that both of these mechanisms could operate at the same time. We hypothesize that (1) lost farm labor due to outmigration is likely to encourage exit from farming; (2) remittances that households receive, on the other hand, are likely to be used on productive investments, including investment in farm mechanization that is likely to lower the demand for farm labor and provide buffer from exiting agriculture. Setting, Data and Methods Setting Our study setting, the western Chitwan Valley, lies in south central Nepal. Nepal boasts a long history of migration through trade routes between the Himalayan regions and the Indian

• plains. Because the country was kept in relative isolation from the rest of the world until the

1950’s, international labor migration was primarily limited to service in the British Army and to non-regulated migration across the border in India. Even though the country opened up to the rest

12

• f the world in the 1950’s, it was not until the Nepalese government promulgated the Foreign

Employment Act of 1989 that international labor migration to destinations other than India became a viable option. This Act licensed non-governmental institutions to export Nepalese workers abroad and legitimized certain labor contracting organizations. This ignited large streams of international migration to many countries besides India (Kollmair et al. 2006; Thieme and Wyss 2005). Recent data for 2013/14 from the Department of Foreign Employment shows that Nepalis migrated to 131 countries globally for work. The number of permits issued for international labor migration increased steadily and dramatically from 220 thousand in 2008/09 to 522 thousand in 2013/14, a more than 130 percent increase within a six year period (Government

• f Nepal 2014). It is reported that over 1500 Nepalis move outside Nepal every day (Pattison

2014). Although India, Japan, and Hong Kong were the major migration destinations until the late 1990’s, in recent decades the booming economies and construction projects in the Persian Gulf region and East Asian countries have made these areas increasingly popular destinations for Nepali migrants (Graner and Gurung 2003). Malaysia, Qatar, and Saudi Arabia were the most frequent destinations, with 207 thousand , 104 thousand, and 75 thousand migrant workers respectively in 2013/14 ( Government of Nepal 2014). However, because of free borders, low transportation costs, and shared culture, India still continues to receive a large number of Nepali migrants (Graner and Gurung 2003; Seddon et al. 2002). Nepal is one of the world’s top 10 remittance-receiving countries with a total amount of US$ 3.5 billion remittances, or 22% of Nepal’s GDP in 2011(MOF 2012). At the household level, 56% of families in Nepal receive remittances, which on average make up 31% of their household income (CBS 2011). This high

13

remittance dependency induces changes in household structures and reorganizes the agricultural system (CBS 2011; MOF 2012). Nepal is predominately an agricultural country, with an overwhelming majority of the population relying on labor intensive, subsistence-based farming. However, change in this area has been rapid. First, the proportion of the population employed in agriculture has declined from 76 percent in 1998 to 67 percent in 2008 (CBS 199; 2009). Second, many farm households are transitioning away from labor intensive farming to a more commercialized farming system with increasing use of farm technology Pariyar, Shrestha, and Dhakal, 2001. MOAC 2003; Agriculture Perspective Plan 1995. For example, tractors are gradually replacing labor traditionally involved in land preparation. Moreover, farmers are increasingly using chemical fertilizers, pesticides and

• ther farm implements (Boserup 1965; Rauniyar and Goode 1992). The fact that the use of

technologies—particularly those designed to perform labor intensive jobs—has the potential to replace labor (Boserup 1965; Rauniyar and Goode 1992) may have important demographic implications, particularly on migration. Data and Measures This study used the individual-, household- and community-level data from multiple surveys collected by the Chitwan Valley Family Study (CVFS). Data to test our hypotheses came from a longitudinal study of 151 neighborhoods scattered throughout the Western Chitwan Valley in Nepal. A neighborhood was originally defined as a geographic cluster of five to fifteen

• households. These neighborhoods were selected as an equal probability, systematic sample of

neighborhoods in Western Chitwan (Barber et al. 1997). This project features analysis of recently available CVFS data covering household dynamics in a prospective design from a 2006 baseline through 2015. The CVFS features a controlled comparison design that limits macro-

14

level variation in climate, commodity pricing, and public policy by focusing on a single geographic area. Within that area the study uses a systematic, stratified sample of neighborhoods that encompass variation in household livelihood strategies, migration, farm technology use, exit from farming, and household migration from the area. Given that the factors we aim to investigate cannot be randomly assigned to people or households, this controlled comparison, longitudinal design mimics experimental design conditions and the high level of measurement substantially reduces the possibility that unobserved factors bias the observed results (Axinn and Pearce 2006). This research uses multiple data sets from the CVFS. CVFS data sets include: a retrospective history of community change, collected in 1995, 2005, and 2015 using neighborhood history calendar (NHC); a survey of household consumption and agriculture practices, conducted in 1996, 2001, and 2006; individual interviews with Life History Calendar

• f all household members aged 15-59, collected in 1996 and 2008; and a monthly prospective

household registry system from 1997 to date. Particularly, relevant to this study is a new data set, collected in 2015, using an innovative household agriculture and remittance history calendar method. This survey collected information on farming and farming practices, farm technology use, and remittances received on an annual basis since 2006 to 2015. In order to enhance respondent’s recall the HARHC was pre- edited with important community and household events that were collected in other CVFS

• surveys. Similarly, to improve the accuracy of the amount of remittances received by the

household, detailed information on the migration history (including dates and places of migrations) of each household member was pre-edited in the calendar. This data covers 1436 households that were interviewed in 2006. It includes all households, including ones that moved

15

• ut of the study area (which we tracked to their destination and interviewed there), new

households formed from splits of the original households, and households that moved into the sample neighborhood after 2006. The HARC data was collected using a face-to-face interview technique with a 99 percent response rate. Exits from agriculture. Measures of exit from agriculture come from the 2015 household agriculture and remittance calendar data. The household survey collected information on whether

• r not the household was farming in a specific year. We operationalize exit from agriculture as

time-varying dichotomous measures of agriculture. We create household-year data files from this information by coding the dependent variable 0 in all years the household involved in agriculture and 1 in the year the household stopped agriculture. Once stopped agriculture, the household is censored from the analysis. Households that did not stop agriculture during the observation period are censored at the end of this period. Community level Measures of Social Organization. Prior research in this setting carefully documents the community-level spread of new non-family organizations and services that reorganized daily social life increasingly outside of the family and away from the family farm over time (Axinn and Yabiku 2001). Our measures of the community-level spread of these changes come from the neighborhood history calendar (NHC) data collection first launched in 1995 and repeated in 2005/06 (Axinn, Barber and Ghimire 1997). NHC data provides a measure

• f distance in walking time from the respondents’ current neighborhoods to the nearest

employment center, market, bank, health service, and bus stop. This data collection employed a mixed-method multi-mode data collection approach including semi-structured group interviews, key informant interviews, and verification of archival records, and it was employed to collect the retrospective histories of the sample neighborhoods. The specific techniques involved in this

16

method are now widely used in social science research. The calendar techniques improve respondents’ recall and produce reliable retrospective measures (Beli 1998; Caspi et al.1996; Freedman et al. 1998). These data provide dynamic measures of how far away each service was from the neighborhood for each year from 1953 to 2006. These walking times vary from 0 minutes (when the service is located within the neighborhood) to hundreds of minutes (more than a day’s walk from the neighborhood). We create dichotmous variables indicating whether or not the nearest service was within 15 minutes of walking distance from the respondent’s neighborhood in a specific year. We then sum up these measures to calculate the total number of years a certain service organization was within a 15-minute walking distance. Measure of distance to urban center. Measures of distance to an urban center also come from neighborhood history data. However, unlike the distance to non-family services, the unit of distance here is miles, not minutes. During the neighborhood history data collection the exact latitude and longitude location of each neighborhood was also calculated from 1:25,000 maps based on aerial photographs of the valley. These locations were entered into a Geographic Information System (GIS), which calculated the distance in miles between each neighborhood and Narayanghat, the valley’s only urban center. Household Ethnicity. As elsewhere, household ethnic background has important influences on both decision to farming and migration. Although the diverse ethnic mosaic of Nepalese society presents a unique complexity, scholars have often categorized ethnicity into five major groups for analytical purposes: Brahmin\Chhetri, Dalit, Newar, Hill Janjati, and Terai Janjati (Blaikie et al. 1980; Axinn and Yabiku 2001). We have adopted these categories for this

• analysis. For more information about these ethnic groups see Bista (1972), Gurung (1980), and

17

Macfarlane (1976). We coded households “1” if the households are members of a specific category and “0” if not, using Terai Janjati as a reference group for comparison. Household Size. Household size as source of farm labor has an important influence on decision to continue farming. Our measures of household size comes from the registry data that tracked household demographic events such as birth, death, marriage, divorce, in and out

• migration. This record provides information on the total number of people living in a household.

Baseline household engagement in agriculture. Our measures of baseline household engagement in agriculture come from 2006 household interview that included a series of questions about different level of engagement, including the total land area the household farmed, the number of farm animals owned, and the number of poultry owned. Because the scale

• f responses to each of the questions varies, we standardized the values in each of these domains

into Z-scores (mean of 0 and standard deviation of 1) and summed them to construct a standardized index of household engagement in agriculture. Household assets and income. Our measures of household assets and income also come from 2006 baseline household interviews which included measures of ownership of assets, including total land owned, ownership of house plot, ownership of radios, televisions, and

• motorcycles. The baseline also measures housing quality and annual household income. The

measures of ownership of assets and income come from the responses to the survey itself, the measure of housing quality comes from interviewers’ observations. Again because the scale of responses to each of the questions varies, we standardized the values in each of these domains into Z-scores (mean of 0 and standard deviation of 1) and summed them to construct a standardized index of household assets and income.

18

Number of Migrants. Our measures of number of migrants comes from an ongoing monthly prospective household registry system (HHRS). Though a household interview HHRS collects information on household demographic events such as birth, marriage, divorce, death, and living arrangements including both out-migration and in-migration with biweekly precision. Using the living arrangement information from HHRS, we constructed time varying measure of number of migrants by counting the number of household members who were living away from home in a specific year.

• Remittance. Our measures of remittance received by a household come from the 2015

household agriculture and remittance calendar data. This calendar collected information on remittances received on an annual basis since 2006 to 2015. As mentioned above, in order to enhance the accuracy of the amount of remittances received by the household, detailed information on the migration history (including dates and places of migrations) of each household member was pre-edited in the calendar. Using the information on amount of remittance received in Nepali rupee from each of the migrants member of the household, we created a categorical measures of remittance received coded as 0 = no remittance received, 1=1- 99,999, 2=100,000-199,999, 3=200,000-299,999, 4=300,000-399,999 and 5= More than 400,000 Nepali rupees (NRS) (approx. US$ 1= RS100). Analytical strategy We adopted a nested model approach to estimate effects of migration and remittance on the annual odds of existing from farming. First, we estimate the effects of community context - access to the community services and distance to urban center. Second, we modeled the effect of household background and characteristics – ethnicity, household size, engagement in agriculture and assets and income, first individually, than combined together, controlling for the community

19

• context. Finally, we modeled the effect of migration and remittance, first individually, than

combined together, controlling for the community context and household measures. We use event history methods to model the risk of exiting from farming. Because the data are precise to the year, we use discrete-time methods to estimate these models (Allison 1982, 1984; Petersen 1991). Household-year of exposure is the unit of analysis and we consider household to be at risk if farming in year 2006 (beginning of observation period).1 Moreover, because the objective of our analysis is the estimation of effects of migration and remittance on the yearly odds of existing from farming we employ a special extension of discrete-time event history methods—the multilevel, random effects, discrete-time hazard model (see Barber et al. 2000 for a detailed explanation of this estimation strategy). This specific modeling strategy has been used successfully before with similar measures and clustering (Axinn and Barber 2001; Yabiku 2004, 2005).2 We estimated all of our models using the GLIMMIX macro for SAS following the Barber et al. 2000 strategy. We discuss the results as odds ratios. These odds ratios can be interpreted as the amount by which the odds are multiplied for each unit change in the respective explanatory factor. That means that if the odds ratio is greater than 1, the effect is positive and every unit change in the explanatory factor increases the odds of the household exiting from farming. If the odds ratio is less than 1, every unit change in the explanatory factor decreases the odds of a household exiting from farming. Results

1 Although it may appear that the discrete-time method of creating multiple person-years for each

household inflates the sample size resulting in artificially deflated standard errors, this is not the case (Allison 1982, 1984; Petersen 1986, 1991). In fact, the estimated standard errors are consistent estimators of the true standard errors (Allison 1982, pg. 82).

2 Multilevel estimation is also imperative because in these data household are clustered within

• neighborhoods. Our multi-level models are two-level models with households being level 1

factors and community characteristics being level 2 factors.

20

Descriptive statistics The first Panel of Table 1 presents the mean, standard deviation, and minimum and maximum values for measures used in the analyses. Table 1 displays descriptive statistics of the

• utcome – exit farming. Note that during the observation period (2007-2015) almost one fifth

(19 percent) percent of the household existed from farming. The next sets are the measures of community context that include distance to urban center and access to community services. The distance to an urban center from current neighborhood ranges from 0.02 to 17.70 miles, with a mean of 9.05 miles. This suggests that the whole study area is fairly small. Recall that the measures of access to community services are the cumulative number of years each of the services existed within the 15-minute walk from the respondent’s current neighborhood since Chitwan was opened for settlement (53 years). Note that, except for bank, most of these services were within a 15-minute walk from the respondent’s current neighborhood for more than 15

• years. For example, on average respondents’ neighborhoods had an employer within a 15-minute

walk for 18 years. The sum of average years of the five community services ranges from 11 to 244 years with a mean of 90.25 years. This suggests that most of the respondents had these community services within a 15-minute walk from their neighborhood for some time.3

3 One of the common concerns in the studies of community influence on household behavior is

that the measures of community are likely to be correlated. As a result, instead of being distinct dimensions as we theorize, these measures could be multiple measures of a single the theoretical

• construct. In order to examine the issue of multicolliniarity we calculated the Pearsons

correlation across our measures of community context. We found the magnitude of the correlations between access to community services ranging from - 0.004 to 0.19; between community services and distance to urban center ranging from -0.02 to 0.11. Although the magnitude of correlation coefficients between access to community services and distance to urban center are modest, none of these coefficients are larger than -0.43. Because these correlation coefficients are modest we treat these factors as independent dimensions of community context.

21

(Table 1 About Here) The middle panel of Table 1 displays descriptive statistics for household background and characteristics: ethnicity, household size, household engagement in agriculture and household assets and income. In terms of household ethnic background, almost half (48 %) are Brahmin/Chhetri, 5% are Newars, 11% are Dalit, 15% are Hill Janjati, and 21% are Terai Janjati. The size of household range from a single person household to 11 with a mean of 6.31. Recall that our measure of household engagement in agriculture is a standardized index of household involvement in agriculture, poultry raising and livestock raising. This standardized index of household engagement in agriculture ranges from -2.93 to 10.36 with a mean of 0.04 and standard deviation of 2.11. Likewise, measure of household assets and income is also standardized index of total land owned, ownership of house plot, ownership of household assets such as radio, television, and motorcycle, housing quality and annual household income. This standardized index of household assets and income range from -8.93 to 12.73 with a mean of 0.13 and standard deviation of 3.09. Finally, the bottom panel of Table 1 displays descriptive statistics for our explanatory measures - migration and remittance. The number of migrants per household range from 0 to 5 with a mean of 1.39 and standard deviation of 1.38, suggesting that out migration is quite common in this setting. As mentioned above, using the amount of remittance received by a household we created a categorical measure of remittance that range from 0-5 with a mean of 1.28 and standard deviation of 1.73. Influence of community context. Table 2 displays the first set of estimates of our event history models. Guided by nested modeling strategy we begin with a simple model with one community context measure at a time

22

with a basic set of controls. In Model 1 of Table 2, we estimate of the effect of distance to urban center from the respondent’s neighborhood on the odds of exist from farming. The distance to urban center from the respondents’ neighborhood has a strong influence on the annual odds of exist from farming. The odd ratios of 0.94 for distance to urban center suggests that one mile away the respondent’s neighborhoods from urban center decreases the odds of exist from faming by 6%. This is quite consistent with previous findings from Nepal and other settings around the

• globe. Next, controlling for distance to urban center we estimated the effect of community

services, both ways, first individually (Models 2 through 6) and then summed together (Model 7). In Model 2 of Table 2, we estimate of the effect of the presence of employer within a 15- minute walk from the respondent’s neighborhood on the odds of exist from farming. The odds multiplier of 1.01 indicates that each additional year of having an employer within a 15-minute walk from the respondent’s neighborhood increases the odds of exist from farming by 1%. This means households in neighborhoods that have had an employer within a 15-minute walk for ten years had annual odds of exiting from farming (1.0110 =1.1046; 1-1.1046= 0.1046) 10% higher than household in neighborhoods with no employer within 15-minute walk. Model 3 of Table 2 estimates the impact of the presence of market within a 15-minute walk from the respondent’s neighborhood on the odds of exist from farming. Again the odds multiplier of 1.01 indicates that each additional year of having a market within a 15-minute walk increases the rate of exist from farming by 1%. Model 4 of Table 3 estimates the impact of having a bank within 15-minute walk on the odds of exit from farming. The odds multiplier of 1.03 for bank indicates that each additional year of having a bank within a 15-minute walk increases the odds of exist from farming first birth by 3%, much stronger effect than an employer and market. Model 5 of Table

23

3 estimates the impact of having a health service within a 15-minutewalk on the odds of exist from farming. The odds multiplier of 1.02 indicates that each additional year of having a health service within a 15-minute walk increases the odds of exist from farming by 2%. Model 5 of Table 3 estimates the impact of having a bus stop within a 15-minutewalk on the odds of exist from farming. Similar to health service, the odds multiplier of 1.02 indicates that each additional year of having a bus stop service within a 15-minute walk increases the odds of exist from farming by 2%. (Table 2 About Here) Model 7 of Table 3 estimates the effect of total number of years of having all five community services within 15-minute walk from the respondent’s current neighborhood. As shown in Model 7, this has a strong positive, statistically significant effect on the odds of exist from farming. The odds multiplier of 1.01 indicates that each additional year of having these community services within a 15-minute walk increases the odds of a first birth by 1%. Although at the face value 1 % increase seems small given the number of years of having all five community services within 15-minute walk ranges from 11 to 244 with a mean of 90 years this effect is very strong. Overall, the effects of community context – distance urban center and access to community (having community services within 15-minute walk from the respondent’s neighborhood) are in the expected direction. More interestingly, these effects are quite strong and consistent with the idea that the immediate context is important predictors of household livelihood strategy – whether or not to continue farming. Influence of household background and characteristics. Table 3 displays the effect of household background and characteristics – ethnicity, household size, assets and income and

24

engagement in agriculture, controlling for community context. As shown in Model 1, household ethnic background has strong statistically significant effect on exist from farming, as expected. Compared to the Terai Janjati, (group indigenous to the study site) all other ethnic groups are more likely to exist from farming. The odds multiplier of 1.60 for Brahimin\Chhetri indicates that compared to Terai Janjati Brahimin\Chhetri households are 60 percent more likely to exist from farming. Likewise compared to the Terai janjati the odds of existing from farming is 57 % higher for Dalit, 102 % higher for Newar, and 96% higher for Hill Janjati. Model 2 of Table 3 estimates the effect of household size in 2006 on exist from farming. We find that household size has strong, statistically significant negative effect on exist from farming. The odds multiplier of 0.88 indicates that each additional member in the household decreases the odds of existing from farming by 12 %. (Table 3 About Here) Model 3 of Table 3 estimates the effect of household assets and income in 2006. We find a strong statistically significant negative effects of household assets and income on exist from farming. The odds multiplier of 0.93 indicates that one unit change in household assets and income index (standardized scale of household assets and income) decreases the odds of existing from farming by 7 %. Model 4 of Table 3 estimates the effect of household engagement in agriculture in 2006

• n exist from farming. We find a strong negative statistically significant effects of engagement
• f household in agriculture on exist from farming. The odds multiplier of 0.80 indicates that

each additional unit change in the household engagement in agriculture index decreases the odds

• f existing from farming by 20 %.

25

Overall, the effect of household background and characteristics – positive effect of household ethnicity, and negative effect of household size, and engagement in agriculture are consistent with the idea that the Terai Janjati- the indigenous group, household with large family size, more assets and income and more engaged in agriculture are likely to continue farming. Influence of migration and remittance. Finally, Table 4 displays the estimates of the effect of migration and remittance on the odds of exist from farming, controlling for community context and household background and characteristics. We find strong statistically significant effects of both- migration and remittance, in opposite direction as we predicated. Model 1 of Table 4 displays the estimates of the effect of migration on the odds of exist from farming. The odds multiplier of 1.11 for number of migrants indicates that each additional migrant increases the

• dds of exist from farming by 11%. This effect is consistent with our hypotheses that

subsistence agricultural the loss farm labor through migration encourages complete exist from

• farming. Thus finding supports the loss labor hypothesis, which argues that in labor intensive

subsistence agriculture, the loss of family labor negatively affects household agricultural production (Adhikari 2001; de Brauw 2007; Jokisch 2002; Taylor, Rozelle, and de Brauw 2003). (Table 4 About Here) Next, we added the measures of remittance in Model 1. Recall that remittance is a six point categorical variable, with 0 = no remittance received, 1=1-99,999, 2=100,000-199,999, 3=200,000-299,999, 4=300,000-399,999 and 5= More than 400,000 Nepali rupees (NRS) (approx. US$ 1= RS100). Model 2 of Table 4 displays the estimates of the effect of remittance

• n the odds of exist from farming. We find strong negative statistically significant effects of

remittance, as we predicated. The odds multiplier of 0.92 for remittance received indicates that each additional unit increase in the remittance scale reduces the odds of exist from farming by

26

8%. This effect is consistent with the New Economics of Labor Migration (NELM) hypothesis that argues that there is a positive impact of outmigration, and particularly of remittances, on migrant-sending areas (Stark and Bloom. 1985). The NELM hypothesizes migration to be a decision made to overcome market failures that constrain local investment and production, implying that migration can raise productive investments in the origin (Adams 1998; Ecer, and Tompkins 2010; Oberai and Singh 1980). Independent effects of migration and remittance. Model 2 in Table 4 displays the independent effects of number of migrants and remittance on exit from farming. As shown in Model 2, we find that not only the amount of remittance received and number of migrants, each has strong, statistically significant effects on exist from farming but are independent of each other. Finally, note that we find substantial and statistically significant effects of migration and remittance. This finding is consistent with our argument that the two opposing mechanisms – loss of farm labor and loosen credit constraint – could operate simultaneously. Discussion This study examines the influence of labor out migration on the odds of exiting from farming – a fundamental societal transformation. We explore the relationship between labor out- migration and agricultural change in rural Nepal, a society characterized by subsistence agriculture until the recent past, but now experiencing dramatic social and economic reorganization and massive international out-migration. These changes have had a substantial influence on the reorganization of local peoples’ social lives. This setting provides a unique

• pportunity to examine the influence of new nonfamily community organizations and

international out-migration on the transition from a predominantly agricultural society to a non- agricultural society.

27

The results of this study show, overall, that the shift in community access to nonfamily social and economic organizations has important total consequences for the odds of exiting from

• farming. Living in a community with nonfamily services nearby for many years substantially

increases the odds of exiting from farming. This finding provides new evidence that community context – at least in terms of access to nonfamily organizations – influences subsequent household decisions about whether or not to continue farming. The consequences of community we document here are consistent with microeconomic models that argue immediate context may influence household decisions through local

• pportunity structures that facilitate and constrain economic opportunities. Our findings are

both consistent with this argument and also indicate that models of the transition from farming to non-farming must account for the contextual effects in order to correctly estimate the association

• f other factors with the decision to exit farming. As explained in the framework guiding our

study, the consequences of international labor migration for exits from farming are exactly this type of topic. Building on the above we investigate the influence of out-migration and remittance on the transition from farming to non-farming. The current literature offers two opposing views of the consequences of outmigration on agriculture change through (a) loss of farm labor and (b) loosening credit constraint. We investigated both of these mechanisms. Indeed, the results show that out-migration affects agriculture change through both mechanisms. The number of migrants from a household in each year substantially and significantly increases the odds of exiting from farming in the following year. This strong and important result is consistent with the hypothesis that loss of farm labor from international labor migration will increase exits from farming among the household members who remain behind. This is important new evidence that international

28

labor migration has significant consequences for the organization of food production in sending area settings. On the other hand, the amount of remittance received from household migrants in a year substantially and significantly reduces the odds of exiting from farming in the following year. This strong and important result is consistent with the hypothesis that income from international labor migration produces new sources of investments that actually reduce exits from farming among the household members who remain behind. This is important new evidence that international labor migration has significant consequences for the organization of food production in sending area settings that are actually in the opposite direction from those predicted by the lost labor hypothesis. Though lost labor from migration reduces the spread of farming across households, the remittances from those migrants prevents household from abandoning agriculture. A key contribution of the study we report here is a simultaneous test of both of these

• pposing hypotheses. The findings show that the number of migrants and the amount of

remittance received have independent influences on subsequent exits from farming. That is, though the two hypotheses yield opposite predictions about the reorganization of agriculture, empirical evidence is consistent with the conclusion that both processes occur simultaneously. Also important, these findings are net of community context and household characteristic

• measures. Rarely has any research on agricultural change been able to disentangle the influence
• f outmigration from the influence of community context in a predominantly agricultural area

that is experiencing dramatic reorganization in community access to nonfamily organizations and massive out-migration. This study demonstrates that the web of influences on agricultural reorganization is just as complex as theory predicts – community context, household baseline,

29

lost labor and increased credit all play significant roles on how these processes unfold in a predominantly poor, rural agricultural society. Social scientists and policy makers alike increasingly emphasize the important consequences

• f out-migration in migrant sending countries. This study primarily focused on one specific, but

crucial consequence—the impact of labor out-migration on agricultural organization that is closely connected with food production and local area food security. Local area food security is a worldwide concern that is predicted to grow in the coming years (FAO and IFAD 2012). The evidence provided here has important implications for policies aimed at food security, particularly in poor agrarian societies experiencing dramatic change in agriculture production and frequent food shortages. The findings of this study indicate that the loss of farm labor through out-migration encourages exits from farming thus may reduce local agriculture production and aggravate food shortages. However, remittances from migrants reduce exits from farming. The remittances received from migrants can be used by those remaining in the household to substitute for loss of farm labor and continue to farm. This study demonstrates that because the both processes occur simultaneously the labor out migration consequence for food production are not nearly as severe as predicted by that hypothesis alone – international labor migration may not actually harm food production or affect food security. Moreover, this study also shows that the came community level changes that increase international labor migration (Massey and Espinoza 1997; Massey et al. 2010) also increase exits from farming. This means that some portion of the change in food production attributed to labor migration in studies that do not consider local community context is actually a product of those community change, not the labor migration itself.

30

References

• 1. United Nations, Department of Economic and Social Affairs, Population Division (2016).

International Migration Report 2015 (ST/ESA/SER.A/384).

• 2. Townsend, R. F., I. Ceccacci, S. Cooke, M. Constantine, and G. Moses. 2013.

"Implementing Agriculture for Development: World Bank Group Agriculture Action Plan (2013-2015)." Retrieved: June 8, 2016 (documents.worldbank.org/curated/en/2013/01/17747135/implementing-agriculture- development-world-bank-group-agriculture-action-plan-2013-2015).

• 3. FAO, IFAD, and WFP. 2015. "The State of Food Insecurity in the World 2015." Meeting the

2015 International Hunger Targets: Taking Stock of Uneven Progress. Rome: FAO.

• 4. Mendola, Mariapia. 2008. “Migration and Technological Change in Rural Households:

Complements or Substitutes?” Journal of Development Economics, 85: 150-175.

• 5. Quinn, Michael A. 2009. “Estimating the Impact of Migration and Remittances on

Agricultural Technology.” The Journal of Developing Areas, 43(1), 199-216.

• 6. de Brauw, Alan. 2007. “Seasonal Migration and Agriculture in Vietnam.” ESA

Working Paper No. 07-04. Agricultural Development Economics Division, The Food and Agricultural Organization of the United Nations, www.fao.org/es/esa.

7.

FAO, WFP, and IFAD. 2012. The State of Food Insecurity in the World 2012: Economic Growth is Necessary but not Sufficient to Accelerate Reduction of Hunger and Malnutrition. Rome, FAO.

• 8. Jokisch, Brad D. 2002. “Migration and Agricultural Change: The Case of

Smallholder Agriculture in Highland Ecuador.” Human Ecology, 30(4), 523-550.

• 9. Adhikari, Jagannath. 2001. “Mobility and Agrarian Change in Central Nepal.”

Contributions to Nepalese Studies. The Free Library, http://www.thefreelibrary.com/Mobility and agrarian change in central Nepal (1).- a092840368.

• 10. Sharma, Jan and Ganesh Gurung. 2009. “Impact of Global Economic Slowdown on

Remittance Inflows and Poverty Reduction in Nepal.” Institute for Integrated Development Studies (IIDS), Mandikhatar, Kathmandu, Nepal.

• 11. Pant, Bhubanesh. 2008. “Mobilizing Remittances for Productive Use: A Policy-oriented

Approach.” NRB Working Paper. Nepal Rastra Bank, Research Department, Serial Number: NRB/WP/4. December 2008.

• 12. Rivera, Jose Jorge Mora. 2005. “The Impact of Migration and Remittances on

Distribution and Sources Income: The Mexican Rural Case.” United Nations Expert Group Meeting on International Migration and Development. UN/POP/MIG/2005/06. Population Division, Department of Economic and Social Affairs, United Nations Secretariat, New York.

• 13. Seddon, David. 2004. “South Asian Remittances: Implications for Development.”

Contemporary South Asia, 13(4), 403–420.

• 14. Stark, Oded and D.E. Bloom. 1985. “The New Economics of Labor Migration.” American

Economic Review,75(2), 173–178.

• 15. Adams, Richard H., Jr. 2011. “Evaluating the Economic Impact of International

31

Remittances on Developing Countries Using Household Surveys: A Literature Review.” Journal of Development Studies, 47(6), 809- 828.

• 16. Government of Nepal. 2010. Nepal Agriculture and Food Security: Country Invest

Plan.http://www.gafspfund.org/sites/gafspfund.org/files/Documents/Nepal%204%20of%20 9%20Country%20Investment%20Plan.pdf, Retrieved on March 8, 2013.

• 17. Rudel, T., O. Coomes,m E. Moran, F. Achard, A. Angelsen, J. Xu et al. 2005. “Forest

Transitions: Towards a Global Understanding of Land Use Change.” Global Environmental Change Part A., 15:23-31.

• 18. de Brauw, Alan, J. Edward Taylor, and Scott Rozelle. 1999. “The Impact of Migration and

Remittances on Rural Incomes in China.” Paper presented at the American Agricultural Economics Association Annual Meetings, Nashville, August 8-11, 1999.

• 19. Taylor, J. Edward, Scott Rozelle, and Alan de Brauw. 2003. “Migration and Incomes in

Source Communities: A New Economics of Migration Perspective from China.” Economic Development and Cultural Change, 52, 75-101.

• 20. Ecer, Sencer and Andrea Tompkins. 2010. “An Econometric Analysis of the Remittance

Determinants Among Ghanaians and Nigerians in The United States, United Kingdom, and Germany.” International Migration, 1-24.

• 21. Koc, Ismet and Isil Onan. 2004. “International Migrants’ Remittances and Welfare Status of

theLeft-behind Families in Turkey.” International Migration Review, 38(1), 78–112.

• 22. Lipton, Michael. 1980. “Migration from Rural Areas of Poor Countries: the Impact on Rural

Productivity and Income Distribution.” World Development, 8(1), 1-24.

• 23. Oberai, A.S. and H.K.M. Singh. 1980. “Migration, Remittances and Rural Development:

Findings of a Case Study in the Indian Punjab.” International Labor Review, 119, 229-241.

• 24. Reichert, Joshua S. 1981. “The Migrant Syndrome: Seasonal U.S. Labor Migration and Rural

Development in Central Mexico.” Human Organization, 40(1), 56–66.

• 25. Rempel, Henry, and Richard A. Lobdell. 1978. “The Role of Urban-to-Rural Remittances in

Rural Development.” Journal of Development Studies, 14(3), 324–341.

• 26. Rozelle, Scott, J. Edward Taylor, and Alan deBrauw. 1999. “Migration, Remittances, and

Agricultural Productivity in China.” The American Economic Review, 89 (2): 287-291.

• 27. Taylor, J. Edward. 1999. “The New Economics of Labour Migration and the Role of

Remittances in the Migration Process.” International Migration, 37(1), 63–88.

• 28. Stark, Oded and D.E. Bloom. 1985. “The New Economics of Labor Migration.” American

Economic Review, 75(2):173–178.

• 29. Adams, Richard H., Jr. 1998. “Remittances, Investment, and Rural Asset Accumulation in

Pakistan.” Economic Development and Cultural Change, 47(1), 155-173.

• 30. Bajracharya, Bijaya. 1994. “Gender issues in Nepali agriculture: a review.” HMG

Ministry of Agriculture/Winroch International Policy Analysis in Agriculture and Related Resource Management.

• 31. Boserup, Ester. 1990. Economic and Demographic Relationships in Development:

Essays Selected and Introduced by T. Paul Schultz. Baltimore, London: The Johns Hopkins University Press.

• 32. Prasad, C. and R. P. Singh. 1992. “Farm Women: A Precious Resource.” Pp. 3-16 in

Women in Agriculture: Education, Training and Development, Vol. 2, edited by R. K.

• Punia. New Delhi: Northern Book Center.
• 33. Rani, Seema and A. Malaviya. 1992. “Farm Mechanization and Women in Paddy

Cultivation.” Pp. 261-67 in Women in Agriculture: Their Status and Role, Vol. 1, edited

32

by R. K. Punia. New Delhi: Northern Book Center.

• 34. Singh, Ram Iqwal, Daulat Singh, and R. K. Singh. 1991. “Green Revolution and Rural

Women.” Pp. 277-84 in Women in Agriculture: Their Status and Role, Vol. 1, edited by R. K. Punia. New Delhi: Northern Book Center. [R03-xx]

• 35. Central Bureau of Statistics (CBS). 2011. National Population and Housing Census 2011

(National Report). National Planning Commission Secretariat, Government of Nepal, Kathmandu, Nepal.

• 36. Ministry of Finance(MOF) Nepal. 2012 Economic Survey: Fiscal Year 2011/2012.

Ministry of Finance: 535 Singhadurbar, Nepal, 2012; pp. 105-107.

• 37. Central Bureau of Statistics (CBS), 1999. Nepal Labor Force Survey 1998/99

Statistical Report. National Planning Commission Secretariat; Kathmandu, Nepal.

• 38. Central Bureau of Statistics, (CBS) 2009. Report on the Nepal Labor Force Survey
• 2008. National Planning Commission Secretariat, Central Bureau of Statistics,

Kathmandu, Nepal.

• 39. Pariyar, Madan, Khadga Bahadur Shrestha, and Narahari Dhakal. 2001. “Baseline Study
• n Agricultural Mechanization Needs in Nepal.” Facilitation Unit, Rice-Wheat

Consortium for the Indo-Gangetic Plains, National Agricultural Science Centre (NASC) Complex, DPS Marg, Pusa Campus, New Delhi 110 012, India.

• 40. Ministry of Agriculture and Cooperatives (MOAC). 2003. Nepal Fertilizer Use Study.

Kathmandu, Nepal: Agrifood Consulting International.

• 41. Agriculture Perspective Plan. 1995. Nepal Agriculture Perspective Plan (APP).

Agricultural Projects Services Center, Kathmandu and John Mellor Associates, Inc., Washington, D.C.

• 42. Boserup, Ester. 1965. The Conditions of Agricultural Growth: The Economics of

Agrarian Change under Population Pressure. Chicago and George Allen and Unwin Ltd: Aldine Publishing Company.

• 43. Rauniyar, Ganesh P. and Frank M. Goode. 1992. “Technology Adoption on Small Farms.”

World Development, 20(2): 275-282.

• 44. Pariyar, Shrestha, and Dhakal, 2001. MOAC 2003; Agriculture Perspective Plan 19.

33

• Table1. Descriptive Statistics of measures used in the analyses (N=1436 Households)

Measures Mean

• St. Dev.

Min Max Outcome Exit from agriculture (yes, no) 0.19 0.39 1 Predictors Community measures Distance to urban center from respondent’s (R) neighborhood (miles) 9.05 3.72 0.02 17.70 Access to community services Number of year employer within 15-minute walk from R’s neighborhood in 2006 18.07 14.81 54 Number of year market within 15-minute walk from R’s neighborhood in 2006 29.35 13.58 54 Number of year bank within 15 minute walk from R’s neighborhood in 2006 1.72 5.83 47 Number of year health service within 15 minute walk from R’s neighborhood in 2006 16.40 14.71 48 Number of year bus stop within 15 minute walk from R’s neighborhood in 2006 24.70 13.31 51 Sum of number of years of all five services within 15 minute walk in 2006 90.25 44.85 11. 244 Household measures Ethnicity Brahmin-Chhetri (yes, no) 0.48 0.50 1 Newar (yes, no) 0.05 0.23 1 Dalit (yes, no) 0.11 0.32 1 Hill Janjajati (yes, no) 0.15 0.35 1 Terai Janajati (yes, no) 0.21 0.40 1 Household size (number of household members) 6.31 2.39 1 11 Household engagement in agriculture (standardized index) 0.04 2.11

• 2.93

10.36 Household assets and income (standardized index) 0.13 3.09

• 8.93

12.73 Migration measures Number of migrant 1.39 1.38 5 Remittance measures Amount of remittance in Nepali Rupee received in category (0=none, 1=1- 100K, 2=100-200K, 3=300-400K, 4= 300-400K, 5= more than 400K) 1.28 1.73 5

34

Table 2. Multi-level Logistic Regression Estimates of the Effect of Community Context on Annual Hazard of Exit from Agriculture from 2007 -2015 (N=1436 Households) Measures Exit from agriculture Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Community access to nonfamily organizations Employer (# year within 15-minute walk from R’s neighborhood in 2006) 1.01* (1.73) Market (# year within 15-minute walk from R’s neighborhood in 2006) 1.01* (1.88) Bank (# year within 15-minute walk from R’s neighborhood in 2006) 1.03* (3.06) Health service (# year within 15-minute walk from R’s neighborhood in 2006) 1.02** (2.88) Bus stop ((# year within 15-minute walk from R’s neighborhood in 2006) 1.02** (2.57) Sum of all five services (# year within 15-minute walk from R’s neighborhood in 2006) 1.01** (3.25) Distance to urban center (miles) 0.94** 0.96* 0.94** 0.94** 0.94** 0.95* 0.96* (-2.74) (-1.77) (-2.63) (-2.76) (-2.56) (-1.95) (-1.90) Time (Year) 0.52** 0.52** 0.52** 0.52** 0.52** 0.52** 0.52** (-7.28) (-7.24) (-7.24) (-7.20) (-7.23) (-7.25) (-7.20) Time squared 1.06** 1.06** 1.06** 1.06** 1.06** 1.06** 1.06** (6.53) (6.49) (6.49) (6.46) (6.49) (6.51) (6.46) household-year 11591 11591 11591 11591 11591 11591 11591 Deviance 2359.99 2361.69 2360.83 2359.95 2361.24 2361.95 2363.76

• 2 Res log Like

76944.9 77053.5 77086.6 77143.5 77195.1 77120.4 77242.0 AIC 76948.9 77057.5 77090.6 77147.5 77199.1 77124.4 77246.0 BIC 76954.8 77063.4 77096.5 77153.5 77205.0 77120.3 77252.0 Note: *** p<.001; ** p<.01; * p<.05; +p<.10

35

Table 3. Multilevel Logistic Regression Estimate of Effect of Household Characteristics on Annual Hazard of Exit from Agriculture from 2007-2015 (N=1436 Households) Measures Exit from agriculture Model 1 Model 2 Model 3 Model 4 Household characteristics Household engagement in agriculture 0.80** (-5.23) Household assets and income 0.93** 0.96+ (-3.39) (-1.53) Household size 0.88** 0.90** 0.93** (-4.68) (-3.89) (-3.38) Ethnicity (Terai Janajati reference group Brahmin-Chhetri 1.60* 1.57* 1.93** 1.57* (2.21) (2.15) (3.00) (2.05) Dalit 1.57* 1.53+ 1.57* 1.26 (1.68) (1.58) (1.68) (0.85) Newar 2.02** 2.04** 2.46** 2.13** (2.33) (2.38) (3.97) (2.47) Hill Janajati 1.96** 2.02** 2.21** 1.88** (2.72) (2.86) (3.22) (2.54) Community access to nonfamily organizations in 2006 Sum of five community services 1.01** 1.01** 1.01** 1.01** (2.86) (2.94) (3.20) (2.82) Distance to urban center (miles) 0.96+ 0.96* 0.96* 0.98 (-1.59) (-1.66) (-1.73) (-1.02) Time (Year) 0.52** 0.53** 0.53** 0.53** (-7.14) (-6.82) (-6.80) (-6.78) Time squared 1.06** 1.06** 1.06** 1.06** (6.42) (6.20) (6.19) (6.18) Household-year 11581 11581 11581 11581 Deviance 2354.63 2341.04 2333.54 2306.46

• 2 Res log Like

77285.4 77609.9 77832.8 78708.4 AIC 77289.4 77613.9 77836.8 78712.4 BIC 77295.3 77619.8 77842.7 77718.3 *** p<.001; ** p<.01; * p<.05; +p<.10

36

Table 4. Multilevel Logistic Regression Model Estimate of the Effect of Migration and Remittance on Annual Hazard of Exit from Agriculture from 2007-2015 (N=1436 Households) Measures Exit from agriculture Model 1 Model 2 Remittance Amount of remittance received 0.92* (-1.71) Migration Number of migrants 1.11** 1.14* (1.71) (2.15) Household characteristics Ethnicity (Terai Janjati reference group) Brahmin-Chhetri 1.53* 1.54* (1.94) (1.95) Dalit 1.23 1.24 (0.75) (0.79) Newar 2.11** 2.09** (2.43) (2.41) Hill Janjati 1.82** 1.83* (2.39) (2.41) Household size 0.91** 0.91** (-2.83) (-2.91) Household engagement in agriculture 0.80** 0.80** (-5.13) (-5.18) Household assets and income 0.96+ 0.97+ (-1.61) (-1.46) Community access to nonfamily community services Mean of five community services 1.01** 1.01** (2.83) (2.88) Distance to urban center (miles) 0.98 0.98 (-1.04) (-1.02) Time (Year) 0.51** 0.52** (-6.96) (-6.72) Time squared (Year squared) 1.06** 1.06** (6.40) (6.30) Household-year 11581 11581 Deviance 2303.64 2302.07

• 2 Res log Like

78767.5 78850.6 AIC 78771.5 78854.6 BIC 78777.4 78860.6 *** p<.001; ** p<.01; * p<.05; +p<.10