1. Annual World Bank Conference on Land and Poverty, Washington, DC
Agricultural Intensification and Market Participation under Learning Externality: Impact Evaluation on Small-scale Agriculture
Tigist Mekonnen Melesse (PhD)
UNU-MERIT and Maastricht University
Responsible Land Governance: Towards an Evidence Based Approach
Annual World Bank Conference on Land and Poverty, Washington, DC
March 20-24, 2017

2. Outline Introduction Research aim Methodology Empirical analysis
Conclusion and policy implications

3. Introduction
There is a widely agreed belief that improved agricultural technologies are crucial for pro-poor development.
In developing countries, especially in SSA where agriculture is the predominant sector that underpins the livelihood of the poor, adoption of improved agricultural inputs are central to transformation of farming systems and a path out of poverty (Kelsey, 2011).
Improving productivity, profitability and sustainability of smallholders is the main pathway out of poverty (World Bank, 2008).

4. ...Contd
Ensuring food security and reducing poverty in SSA can be realized by enhancing the productivity in agri. that can be achieved through adoption of agri. technologies such as HYVs and chemical fertilizer or expansion of arable land.
Expansion of arable land is less viable for spatial or environmental reasons.

5. ...Contd
Hossain (1989) reported since area expansion and irrigation have already become a minimal source of growth at a world scale, agricultural growth will more depend on yield-increasing technological change.
Jayne et al. (2003) found declining average per capita arable land over the last 40 years in SSA.
Degefa and Nega (2000) reported declining average land holding in Ethiopia.
A viable option that remains is adoption of yield enhancing agricultural technologies.

6. ...Contd
Thus, the application of improved agricultural inputs enable resource poor smallholder farmers to produce marketable surplus crops.
While, the continuous use of low yielding local crop varieties stagnate the sector into subsistence farming.
Braun et al. (1994) the need to promote smallholders market participation is an important effort necessary to bring agricultural transformation.
Since producer's production technology choice might affect productivity, the increased productivity gain is not an end by itself and it should be linked to the market for driving profits.

7. ...Contd
Thus, commercialization is the result of a simultaneous decision marking behavior of farm households in production and marketing.
Recently, developing countries have started to promote the diversification of agricultural production and exports away from traditional commodities.
This accelerate economic growth, create employment opportunities, and reduce poverty.

8. For instance:
The export of flower farm industry in Ethiopia experienced a boom growing from 0.3 million US dollar in 2001 to 113 million US dollars in 2007 (Joosten, 2007).
In Zambia, fresh vegetable & cut flowers grown on smallholder farm rose from 6 million US dollar in 1994 to more than 33 million US dollar in 2001 (Diao and Hazell, 2004).
The export gains from horticultural crops signify a progressive and remarkable result, replication to other cereal crops requires appropriate technological investment.
This staple food crops are important for the growing population food security and is the major source of income for rural smallholders in Ethiopia.
Besides, more than 70 % of the rural population in Ethiopia produce cereal crops (Shiferaw and Teklewold, 2007).

9. However, the lack of technological change and market imperfection have often locked smallholders into subsistence production and contributed to stagnation of the sector.
Although the gov’t of Ethiopia working with extension agents distribute improved agricultural inputs to smallholders, studies indicates that farmers use of main input such as HYVs is less than 5 % (CSA, 2013; Taffes et al., 2013).
Thus, increasing farmers use of new agricultural inputs by encouraging smallholders participation in farmers group and other rural associations increase the adoption behavior.
Moreover, farmers may access information from their neighbors. Thus, social learning can play an important role in the adoption of improved agricultural technologies.
Munshi (2007) due to sequential information flow from one farmers to the neighbor, social learning provides a natural explanation for the gradual diffusion of new agricultural technologies even in a homogenous populations.
(Gebreeyesus and Iizuka, 2012) learning does not takes place in isolation, rather it involves a range of actors and networks; both in the formal and informal way.
Either in a social or economic network, learning externality could increase new agricultural technology adoption and farmers participation in to output marketing.
Thus, encouraging smallholders participation in farmers groups, agricultural cooperatives and other rural associations increase farmers use of HYV and chemical fertilizer.
To this effect, there is initiatives through Gov’t and Non-Gov’t Orgn encouraging rural smallholders to participate in different association such as agricultural cooperatives, farmer's group, savings and microcredit cooperative, other groups.

10. 2. Research aim
Usually commercialization is discussed with respect to large scale farming systems and ignoring smallholders and poor farm households participation.
Besides, due to limitation of data about information on interconnectedness, empirical estimation of the impact of technology adoption on smallholders MP linked to rural social networks is rare.
However, enormous theoretical literature on social learning.
This study evaluate the impact of HYVs and chemical fertilizer adoption on marketed surplus cereal crop production under learning externality to contribute to the empirical literature that use micro-level data measuring the intensity of MP.
This could guide the type of intervention which is needed to facilitate the current commercial transformation of subsistence agriculture in Ethiopia.
Therefore, this study incorporate the issue of neighborhood effect in the adoption of HYVs and chemical fertilizer on smallholders marketed surplus crop production to guide the type of intervention which is needed to facilitate the current commercial transformation of subsistence agriculture.
The analysis also assess the issue of neighborhood effect in the adoption of improved agricultural inputs among rural smallholders.

11. 3. Methodology
We use household level data from Ethiopia Farmer Innovation Fund (FIF) impact evaluation survey collected by the World Bank in 2010 and 2012.
The survey covers 2,675 households drawn from two regions and fifty-one rural villages.
Since technologies are not randomly assigned into households, farmers decide whether to adopt or not to adopt and the decision to participate or not to participate into output marketing.
Thus, analyzing the impact of HYVs and chemical fertilizer adoption on smallholders market participation is subject to self-selection bias.
Since technologies are not randomly assigned into households, farmers decide whether to adopt or not to adopt and the decision to participate or not to participate into output marketing, hence adoption and market participation both are potentially endogenous

12. ...Contd
Besides, the heterogeneity of households in socioeconomic, demographic, institutional and information access, both the observed and unobserved factors affects adoption and market participation.
Moreover, the decision of households not to participate into output marketing may give rise to too many zeros (censored dependent variable case).
To address the self selection bias in technology adoption and market participation, we applied endogenous treatment effect (TE) model.
Heckman selection models is also employed to solve the problem associated with censored observations.
Regressions based on matching techniques are employed for robustness check.

13. ...Contd
Heckman selection models is applied to solve the problem associated with censored observations.
The traditional approach to evaluate data that yields a censored dependent variable is applying a Tobit model.
In the Tobit model, all households including censored values are included without considering the source of the zeros.
The two level decision of households in technology adoption and market participation (how much supply to the market) are taken to be the same in the Tobit model, while the explaining variables may not be the same.
We estimate the impact of HYVs and chemical fertilizer adoption on smallholders marketable surplus crop production through endogenous treatment effect (TE) model by taking into account the self selection bias in technology adoption and market participation.
To address the self selection bias in technology adoption and market participation, we applied endogenous treatment effect (TE) model to evaluate the impact of HYVs and chemical fertilizer on smallholders marketable surplus crop production.
After we estimate the impact of HYVs and chemical fertilizer on smallholders marketed surplus, a sample selection model Heckman (1979) is measured to address the problem associated with the censored observations generated by households decision not to participate in the output market.

14. 4. Empirical analysis
The estimation results presented in this section shows the two main aims of the paper.
First, the analysis evaluates the impact of HYVs and chemical fertilizer adoption on smallholders marketed surplus crop production and emphasis the r/s b/n technology adoption and market participation.
Second, the analysis identifies the main factors that determines technology adoption and market participation (sales volume).
Guo and Fraser (2014) indicated that since the development of the sample selection model; econometricians have formulated many new models and estimators.
One of the important developments is applying the sample selection model to estimation of treatment effects in non-experimental data.
Regression based on sample selection model is measured to understand the factors that determine households which do not participate in the output market (zero values).
Both endogenous treatment effect model and sample selection model allows variables to freely vary in the second stage than variables in the first stage.

15. Table 1. Adoption rate of the two inputs in the data
Guo and Fraser (2014) indicated that since the development of the sample selection model; econometricians have formulated many new models and estimators.
One of the important developments is applying the sample selection model to estimation of treatment effects in non-experimental data.
Regression based on sample selection model is measured to understand the factors that determine households which do not participate in the output market (zero values).
Both endogenous treatment effect model and sample selection model allows variables to freely vary in the second stage than variables in the first stage.

16. Table 2. Endogenous treatment effect estimation result
Guo and Fraser (2014) indicated that since the development of the sample selection model; econometricians have formulated many new models and estimators.
One of the important developments is applying the sample selection model to estimation of treatment effects in non-experimental data.
Regression based on sample selection model is measured to understand the factors that determine households which do not participate in the output market (zero values).
Both endogenous treatment effect model and sample selection model allows variables to freely vary in the second stage than variables in the first stage.

17. Table 3. Matching techniques
Guo and Fraser (2014) indicated that since the development of the sample selection model; econometricians have formulated many new models and estimators.
One of the important developments is applying the sample selection model to estimation of treatment effects in non-experimental data.
Regression based on sample selection model is measured to understand the factors that determine households which do not participate in the output market (zero values).
Both endogenous treatment effect model and sample selection model allows variables to freely vary in the second stage than variables in the first stage.

18. From the sample selection model result, the main exogenous factors that positively and significantly affect smallholders market participation intensity are:
Access to modern inputs, cereal crop price, farm size, availability of labor, infrastructure facility and asset ownership (Livestock)
Access to credit and training fosters smallholders improved agricultural inputs use.
However, we are unable to witness the effect of neighborhoods (access to information from neighborhoods, number of persons contact and membership of local associations) on smallholders sales volume.
Guo and Fraser (2014) indicated that since the development of the sample selection model; econometricians have formulated many new models and estimators.
One of the important developments is applying the sample selection model to estimation of treatment effects in non-experimental data.
Regression based on sample selection model is measured to understand the factors that determine households which do not participate in the output market (zero values).
Both endogenous treatment effect model and sample selection model allows variables to freely vary in the second stage than variables in the first stage.

19. 6. Conclusion
The empirical estimation result indicates that adoption of improved cereal crop varieties and chemical fertilizer helped smallholders to produce marketable surplus production.
This indicates targeting intensification through new agricultural technologies can have far reaching poverty reduction implication especially in rural areas farming is the major source of income and food production.
The findings show improved technologies is the priority requirement for changing rural households agricultural production and market participation system.
Importantly, elements of institutional service such as infrastructure development, efficient input distribution system and appropriate economic incentives such as farmers taking advantage from output price must be present.
The overall finding indicates the development of a more dynamic and competitive agricultural sector in Ethiopia requires appropriate rural institutions that respond effectively the changing agricultural technology adoption and market conditions.

20. 7. Policy implications
First, intervention to expand the distribution system of HYVs and chemical fertilizers should be the government strategy to increase smallholders marketed surplus crop production.
Second, the increased productivity gain from modern inputs use may face significant price risk, to reduce the adverse effect of sunk cost and price risk, the availability of risk management schemes such as support price, searching market through negotiation with private companies and cooperative unions tackle the risk of output price.
Third, encouraging the participation of farmers in training centers is the priority requirement that public organizations need to focus increasing smallholders new inputs adoption behavior.
Recommended policy interventions to improve rural smallholders agricultural technology adoption and market participation includes:
due to the bulk ration, farmers may not take in to advantage from marketed surplus crop production, thus increasing access to credit by prolonging the credit repayment period until the seasonal crop prices rise is crucial

21. Thank You