Mechanize or exit farming? Multiple‐treatment‐effects model and external validity of adoption impacts of mechanization among Nepalese smallholders

• DOI: http://doi.org/10.1111/rode.12407
• Author: Hiroyuki Takeshima
• Date: 01 November 2018
• Published date: 01 November 2018

REGULAR ARTICLE

Mechanize or exit farming? Multiple-treatment-

effects model and external validity of adoption

impacts of mechanization among Nepalese

smallholders

Hiroyuki Takeshima

International Food Policy Research

Institute, Washington, DC, USA

Correspondence

Hiroyuki Takeshima, International Food

Policy Research Institute, Washington,

DC, USA.

Email: H.takeshima@cgiar.org

Abstract

The future of smallholders in developing countries is be com-

ing increasingly uncertain in the face of rising farm wages.

The custom-hiring of tractors, in which tractor owners provide

non-owner farmers with land preparation and transport ser-

vices for fees, has spread among smallholders in Asia, includ-

ing Nepal. However, estimating the adoption impacts of

agricultural mechanization by smallholders is complex as we

must also take into account smallholders’options to exit farm-

ing. We investigate this issue by applying multinomial logit

inverse-probability weighting and sample selection panel data

methods to data on smallholders in lowland Nepal. Our results

are generally consistent with the hypothesis that smallholders

whoarelikelytobenefitmorefrom adopting tractors are also

more likely to exit farming. Where smallholders are less likely

to exit farming, the use of tractors through custom-hiring may

help smallholders on average to earn greater total and agricul-

tural incomes. However, where they are more likely to exit

farming, the ability of custom-hired tractors to sustain small-

holder farming systems may become weaker. The results also

offer insights into how the external validity of technology

adoption impact evaluation may be affected in some settings.

DOI: 10.1111/rode.12407

1620

|

©2018 John Wiley & Sons Ltd wileyonlinelibrary.com/journal/rode Rev Dev Econ. 2018;22:1620–1641.

1

|

BACKGROUND

The future of smallholders in developing countries, particularly Asian countries such as Nepal, is

increasingly being investigated in the literature (Masters et al., 2013; Rigg, Salamanca, & Thomp-

son, 2016). Smallholders in Asia may lose comparative advantages in the near future du e to rising

rural wages, with potentially grave consequences for food security (Otsuka, 2013). In this context,

the literature has paid more and more attention to the role of mechanization, including through

tractor custom-hiring

1

services. In some cases, mechanization has helped smallholders to continue

farming despite rising rural wages (Diao, Cossar, Houssou, & Kolavalli, 2014; Zhang, Yang, &

Reardon, 2017), raising returns to scale in both land and non-land inputs (Takeshima, 2017). In

other cases, mechanization is found to benefit relatively larger farms more than smaller farms (Liu,

Violette, & Barrett, 2016; Wang, Yamauchi, Otsuka, & Huang, 2016). All in all, a knowledge gap

still exists regarding whether and how smallholders benefit from mechanization.

Estimating the effects of mechanization on smallholders is further complicated by the option to

exit farming.

2

Most of the aforementioned studies focus on those who continue farming. However,

some smallholders may choose to exit farming, by sharecropping or renting out all of their farm

plots, instead of continuing farming by adopting mechanization. If the factors affecting the decision

to continue or exit farming are correlated with those affecting the decision to adopt mechanization

upon deciding to continue farming, the failure to incorporate the self-selection into or out of farm-

ing can bias the estimated effects of adopting mechanization.

We fill this knowledge gap by focusing on smallholders in the Terai (lowland) zone of Nepal. Nepal

Terai has experienced fivefold increases in the proportion of tractor-using smallholders over a period of

15 years (Takeshima, 2017). Simultaneously, some smallholders have effectively exited farming by share-

cropping/renting out their entire plots or abandoning cultivation, and specializing in off-farm income-

earning activities. The coexistence of three types of smallholders, non-mechanized farmers, mechanized

farmers, and non-farmers, within the relatively homogenous production environment of Nepal Terai offers

an ideal setting to test the hypothesis that estimated effects of tractor adoption on key outcomes of small-

holders, such as their incomes,

3

are affected by their self-selection to continue or exit farming.

We empirically test this hypothesis by applying the multinomial logit inverse-probability weight-

ing (MIPW) method (Cattaneo, 2010; Cattaneo, Drukker, & Holland, 2013) to the repeated cross-

sectional sample of the Nepal Living Standard Survey (NLSS), and the sample-selection panel (SSP)

data method (Dustmann & Rochina-Barrachina, 2007) to small panel samples of the NLSS, respec-

tively. The MIPW method extends conventional IPW methods commonly used for impact evalua-

tions (Wooldridge, 2007) to a multiple-treatments framework. The SSP method incorporates

selections into or away from farming samples as attritions or accretions, while the endogeneity of

tractor adoption decisions is addressed through the generalized method of moments (GMM).

We show that ignoring the self-selection to continue or exit farming can bias the estimated

effects of tractor adoption for smallholders. The true effects of tractor adoption may be larger than

the estimates based only on those who continue farming. Results are consistent with the hypothesis

that, among smallholders, those who tend to benefit more from tractor adoption may also be more

likely to exit farming. Consequently, for smallholders, mechanization may substitute exit from

farming when such exit faces constraints.

The conventional impact evaluation literature addresses the self-selection into treatment or control

status. Few studies, however, also consider the effect of sample selection in the whole population from

which treatment and control groups are extracted, which can have important implications for the exter-

nal validity of the estimated impact. Methodologically, our paper fills this important gap in the litera-

ture.

TAKESHIMA

|

1621