• No results found

Assessment of farmers’ willingness to pay for mechanization: a study on small-holder sesame farmers in Humera, Ethiopia

N/A
N/A
Protected

Academic year: 2021

Share "Assessment of farmers’ willingness to pay for mechanization: a study on small-holder sesame farmers in Humera, Ethiopia"

Copied!
47
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

I

Assessment of farmers’ willingness to pay for mechanization: a

study on small-holder sesame farmers in Humera, Ethiopia

MASTER THESIS

Author: Fabian Kohler Student number: S3245829

Contact: fd.kohler@gmail.com

Study program: M.Sc. International Economics and Business Faculty: Faculty of Economics and Business

University: University of Groningen

Supervisor: Dr. Bartjan Pennink, Dr. Clemens Lutz Co-Assessor: Dr. Tarek Harchaoui

(2)

II

Table of contents

Table of contents ... II List of figures ... III List of tables ... IV List of abbreviations ... V Abstract ... VI 1 Introduction on improving the use of mechanization in agriculture in

Ethiopia ... 1

1.1 Mechanization in Humera - a project with Selet Hulling PLC ... 2

1.2 The study area of Tigray region in Northern Ethiopia ... 3

2 Theoretical background: the role of mechanization in Africa ... 4

The role of mechanization in agriculture ... 4

3 Literature review on mechanization and willingness to pay in the agricultural sector of developing countries ... 6

3.1 Mechanization in Ethiopia ... 6

3.2 Contributions on willingness to pay ... 8

3.3 Key facts and resulting research hypotheses ... 11

4 Methodology ... 13

4.1 Contingent Valuation Method ... 13

4.2 Probit Model ... 18

4.3 Multinomial Probit Model ... 20

4.4 Analysis and results ... 21

5 Conclusion ... 31

5.1 Implications for government policies and private mechanization companies ... 32

Appendix ... 33

Bibliography ... 36

(3)

III

List of figures

Figure 1: Agriculture value added per worker in 2016 in constant 2010 USD...1

Figure 2: Map of Ethiopia and Location of Tigray Region...3

Figure 3: Land Productivity in Cereal Yield (kg per hectare) 1961-2014...5

Figure 4: Nonlinear probability curve of the probit model...18

Figure 5: Summary of dependent and independent variables...22

Figure 6: Explanation for negative willingness to pay results...22

Figure 7: Distribution of family labor used for farming...23

Figure 8: Main bottlenecks of machinery use...24

Figure 9: Average marginal effects of the probit model...25

Figure 10: Classification of probit model for WTP...27

Figure 11: Frequency of the multinomial dependent variable...28

Figure 12: Marginal effects of the multinomial probit model...28

(4)

IV

List of tables

(5)

V

List of abbreviations

A B C

AGRA = Alliance for a Green Revolution in Africa BDT = Bangladesh Taka

CVM = Contingent Valuation Method G K N P S U W

GDP = Gross Domestic Product Kg = Kilogram

NGO = Non-governmental organization PLC = Public Limited Company

(6)

VI Title: Assessment of farmers’ willingness to pay for mechanization: a study on small-holder sesame farmers in Humera, Ethiopia

Year: 2018

Author: Fabian Kohler

Examiners: Dr. Bartjan Pennink, Dr. Clemens Lutz, Dr. Tarek Harchaoui

Abstract

(7)

1

1 Introduction on improving the use of mechanization

in agriculture in Ethiopia

Ethiopia, the second most populated African country with 102 million inhabitants (World Bank 2016), has experienced the third fastest economic growth in GDP per capita worldwide since 2000. (International Monetary Fund 2017) In contrast to developed countries, Ethiopia’s agricultural sector still exhibits a key role in the country’s economic growth, accounting for 2.6 percentage points in the country’s 10 percent GDP growth in 2014/15. However, this could even underestimate the relevance of the sector. With 80% of the working population being employed in the agricultural sector producing 90% of the country’s total exports, the sector is not only of high economical importance, but also of utmost importance for the country as a whole. (Alliance for a Green Revolution in Africa (AGRA) 2016; Abrar et al. 2004; World Bank Group 2016 )

When compared to other countries however, Ethiopia has certain bottlenecks in the area of agriculture. The country has one of the lowest value added rates worldwide and also underperforms in total factor productivity in comparison to other sub-Saharan countries especially during 1986-2006. (Yu & Nin-pratt 2011) Figure 1: Agriculture value added per worker in 2016 in constant 2010 USD. (The World Bank 2016)

With a value added per worker in agriculture of only 485 USD in 2016, Ethiopia ranks only 124th out of 137 countries worldwide. The world average lies at 17,182 USD and developed European countries reach a value added of more than 80,000 USD per year. Research shows that the main explanation for this is a lack of mechanization, since most of Ethiopia’s farm power is coming from draught animals, mostly oxen, followed by manual labor (Clarke, L., & Bishop 2002) whereas even Sudan with its ongoing civil wars for example has a higher share of mechanization. (FAO/UNIDO 2008)

(8)

2

The fact that Ethiopia has a relatively strong economy in comparison to other African countries, but a lower productivity in agriculture due to a lower grade of mechanization, makes the assessment of introducing machinery in Ethiopia worthwhile to investigate. Agricultural machinery is positively associated with agricultural productivity and thus inhibits a key role in increasing the country’s overall productivity. (Alliance for a Green Revolution in Africa (AGRA) 2016) Despite all the economic reasons to use advanced technology a major role is, whether farmers want to have a change towards mechanization or not. Field research in Ethiopia revealed that a lack of finance is a key issue for farmers and that many can’t afford mechanization. Farmers’ willingness to pay however, is a key prerequisite for mechanization to take place. (Sims et al. 2016) As a consequence, the introduction of machinery does not make sense, if farmers are not willing to pay for it.

Four research questions are therefore investigated, focusing on farmers’ general willingness to pay, the influence of socioeconomic factors, such as age, income or education on willingness to pay, and differences in willingness to pay when farmers are having the choice between different machinery packages. In order to get the necessary information to conduct the econometrics, I will directly speak to farmers, collecting data in sample surveys and measuring environmental factors, which might influence farmers in their decision-making. Following on the introduction, I will shortly introduce a mechanization project in Ethiopia, which aims on introducing machinery. Second, the region in which the research takes place is examined, before giving further insights into the major areas of this paper mechanization and assessing relevant existing literature about mechanization and willingness to pay, in order to give an overview about what has been researched in this area before. The main part discusses the used methodology and why it has been chosen for the research of the thesis, followed by the results and whether they stand in line with results of previous research. In the end a conclusion is given, with an outlook on further steps needed to realize mechanization in the given project.

(9)

3

meaning an organization that is lead by farmers and represents the interests of local small-scale farmers. After harvesting, Selet processes the sesame in its factory in Addis Ababa and ships it to different locations worldwide. (Selet Hulling PLC 2009)

Together with another master student, I worked for Selet Hulling in a project, which aims at introducing a machinery rental service in the Tigray region in Ethiopia to sesame farmers. We created a business plan for a machinery rental service and therefore conduct research on the current farming process of local farmers and draw up cost calculations about current sesame farming techniques in comparison to mechanized sesame farming. Furthermore, we

compare different machinery packages, in order to find the optimal farming solution for the local farmers. The aim is to sustainably increase productivity in farming as well as increase yields per hectare, while searching for possibilities to reduce costs.

The business plan itself does not only focus on profit maximization for Selet Hulling, but follows a social perspective as well. Instead of sourcing sesame from large-scale farmers, Selet Hulling wants to promote local small scale farmers with land sizes between two and twenty hectares. The company wants to pre-finance machinery for farmers, to build up a rental company for the machinery, and to give local farmers the opportunity to buy shares of the rental company from the money made through mechanization. In this model, farmers can, over the years, become the only shareholder of the rental service, giving them the opportunity to own machinery themselves. Additionally, knowledge transfer and capacity building play a major role, in order to increase yields for the farmers and to create new jobs in areas of machinery driving and maintenance as well as management of the new company.

1.2 The study area of Tigray region in Northern Ethiopia

Figure 2: Map of Ethiopia and Location of Tigray Region (Ethiopia Black History Heroes 2010)

(10)

4

400 to 800 mm each year, with the rainy season lasting from June to August whereas the rest of the year is characterized by drought period. (Teka et al. 2012) Besides sorghum, wheat and teff, which is mainly used for local production (Teka et al. 2012), sesame is one of the main crops planted in the region, due to its high oil quality and export demand. (Gebremichael & Parzies 2011) Furthermore, it is the second largest source of export revenues after coffee and therefore has a major importance for the local and national economy. (Olthaar et al. 2017) The economy of the region highly depends on agriculture as it accounts for 40% of the regions GDP followed by tourism as the second largest income source. (Ministry of Finance and Economic Development 2015) The exact study location of the thesis is a region called Humera, which is in the North West of Tigray, directly bordering Eritrea and Sudan. As in the rest of the region, agriculture is the main income source with sesame being the most-grown crop. Sesame farmers are partly independent and sell their yields to the Ethiopian Commodity Exchange, or are as mentioned before, organized in so-called cooperatives. (Olthaar et al. 2017) Since Selet Hulling is working together with cooperatives, the research will focus on farmers from those cooperatives as well.

2 Theoretical background: the role of mechanization in

Africa

In this chapter, I explain the role of mechanization in agriculture and its importance for Sub-Saharan Africa (SSA) and especially Ethiopia, which is the reason for choosing the research on willingness to pay for mechanization.

The role of mechanization in agriculture

Mechanization in an agricultural context is defined as tools, implements and powered machinery, which are essential inputs to agriculture. (Clarke 2000) Furthermore, the level, appropriate choice and subsequent proper use of mechanized inputs into agriculture have a major effect on land productivity, labor productivity, profitability, sustainability and the environment of agriculture. (Olaoye & Rotimi 2010)

(11)

5

Figure 3: Land Productivity in Cereal Yield (kg per hectare) 1961-2014 (The World Bank 2015)

As displayed in figure 3, the Sub-Saharan region performs the worst in land productivity. Whilst regions such as Asia experienced a so called Green Revolution following the introduction of modern agricultural practices and equipment (Rashid et al. 2007; Evenson 2003), most farmers in SSA still rely on human or animal power and mechanized sources of power are rare (Sheahan & Barrett 2014). Ethiopia in particular underperforms especially from the 1960s until the 2000s, but experienced an increase in productivity since then. Compared to its neighboring country, Kenya, Ethiopia closed the productivity gap from recent decades and shows that it is on the right track. However, it still underperforms in comparison to all other non-African regions.

Research from the International Food Policy Research Institute and the Ethiopian Development Research Institute have shown that to a large amount, the lack of productivity in Ethiopia in comparison to most other African countries, is due to a lower use of mechanization in agriculture. (Yadene 2017) The same can be seen in Yu and Nin-Pratt’s paper about Agricultural Productivity and Policies in Sub-Saharan Africa, where Ethiopia’s productivity through technological change accounts for 1.001 points, with the index 1 standing for zero growth. (Yu & Nin-pratt 2011)

The low productivity is particular problematic as the development in the agricultural sector is not only an important factor for the reduction of poverty in the rural areas of the region, but also to reduce the prevailing food shortages. (De Janvry & Sadoulet 2010)

(12)

6

3 Literature review on mechanization and willingness

to pay in the agricultural sector of developing

countries

This literature review covers two main segments, the literature on a specific sector: “willingness to pay” (Par 3.2) and also literature on mechanization in Ethiopia (Par 3.1). Literature on mechanization is assessed to give a close insight into the situation of mechanization in Ethiopia. With the background of chapter 2, I will assess the need for mechanization in Ethiopia, and display potential bottlenecks in mechanization of the country in order to find out about the possible willingness to pay.

The existing literature on willingness to pay, which is the main focus of this literature review, is divided into two main sections: environmental studies and health care studies. A variety of research has been done in both areas, for this paper however, I will focus on environmental studies, especially the farmers in development countries, to have a direct comparison to the topic of my own research. For a full comparison of methodological approaches, I will also introduce papers which use methods for assessing willingness to pay, which are not mainly used in the context of environmental studies.

3.1 Mechanization in Ethiopia

Baudron et al. (2015), analyze in their paper the need for mechanization in Southern and Eastern Africa, as well as the possibilities to solve these needs. They don’t apply econometric models, but compare results and techniques from different papers targeting various countries in the region of East and South Africa. Their combined results are that for the case of Ethiopia, especially the farm power available per area of agricultural land lags behind. The number of tractors increased up till 2005, yet amounting to 14 tractors only per 1000 ha. Ethiopian farmers instead mainly use draught animals for farming, which explains the farm power constraints. In addition, due to the lack of mechanization, 53% of farmers hire additional labor for farming activities. The lack of mechanization leads to high labor drudgery making farming unattractive to the youth. Baudron et al. (2015) suggest, that by the use of small, multipurpose and inexpensive power sources, appropriate mechanization in Ethiopia and other parts of East and South Africa may be re-examined, mainly because most farming land is owned by smallholder farmers, which makes the use of large machinery difficult. Introducing new technology however, would need the help of NGOs, the government and donor-supported projects, to facilitate linkages between smallholder farmers and the private sector and develop working business models for mechanization.

(13)

7

ministries, as well as cooperative unions, machinery importers and farmers. Second, they used quantitative data including import data from the United Nations and household data from the Ethiopia Socioeconomic Survey and the Ethiopian Feed-the-Future Program. Results reveal, that even though the number of tractors imported in the country was five times higher in 2014 than in 2009, the uptake in mechanization in Ethiopia is still low, with less than one percent of farming land being plowed with machinery. This is a loss for farmers, as machinery usage correlates with higher yields. The increase in imports of tractors is further associated with increasing labor and animal traction costs. However, most of the newly imported machinery is only used by large commercial farmers and state farms. Smallholder farmers rely heavily on machinery rental services, as owning machinery is not efficient nor affordable for farmers with small fields. There are physical constraints to mechanization, linked to the type of soils. This results in farmers from certain areas relying on large machinery, since small tractors don’t have the required power to operate in heavy soils. Land fragmentation and small farm plots complicate the use of big machineries though, which suggests that policies should promote land consolidation and rental services to solve this issue.

(14)

8

3.2 Contributions on willingness to pay

Main literature on willingness to pay

Uddin & Gao (2016) analyze farmers’ willingness to pay for agricultural extension services in Bangladesh, where? the extension services experience a funding crisis. It is one of the most recent studies on willingness to pay. The contingent valuation method (Par. 4.1) was used here for data collection, and tobit and logit model for analysis. The logit model was designed to capture farmers’ willingness to pay in a binary model and finding out about influencing socioeconomic factors; the tobit model was used to find out about the influence of extension visits. The latter was selected as the dependent variable was only incompletely observed and the tobit model makes it easier to come to results in that case. They research the influence of socioeconomic factors (age, income, education etc.) on the farmers’ WTP. Their main findings are that the price plays an essential role and the farmers are willing to pay a maximum of 50 BDT. In addition, education, income and farming experience were revealed as the most important factors for WTP.

The combination of CVM and the logit model is a popular format in WTP studies. Other authors making use of this technique are Ahmed et al. (2015), analyzing farmers’ willingness to pay for a planned adaptation program for addressing climate issues in Pakistan’s agricultural sectors and Bakopoulou et al. (2010), who study farmers’ WTP to use recycled water for irrigation purposes in Greece. Both studies use similar socioeconomic factors as explanatory variables. Ahmed et al. (2015) showed that education, income, household size, firm size and motivation were positive coefficients for WTP. For Bakopoulou et al. (2010) education and gender positively correlated with WTP, whereas income and farm size are negatively correlated with WTP for recycled water. This makes sense, as if one has more money available, one would prefer using fresh water for irrigation over recycled water.

(15)

9

Bozorg-Haddad et al. (2016) examine farmers’ willingness to pay to avoid water shortages. However, other than most papers, the authors try a different method. Counter to the prevailing approach, the authors opt for a probabilistic optimization method to estimate WTP instead of CVM.

A nonlinear programming model was formulated to illustrate water use and estimate WTP in cases of water shortages. Furthermore, the authors distinguish between single farmers and group of farmers and a Monte Carlo simulation method is applied. It generates the means and variances of uncertain model parameters. Results suggest that with no water rationing, water use declines in decreasing water prices. With water rationing, farmers’ WTP and water use are affected by the water price and willingness to pay for water to avoid shortages decreases, with higher prices.

Another method is introduced by Ali (2013), who assesses farmers’ WTP for insurances in rain-fed areas in Pakistan. Using surveys from over 500 Pakistani farmers, Ali applies the Gustaffson-Wright model of WTP, which is as follows: WTP =!(#$, #&, ', (, ), *, +). ! stands for the maximum amount a person is

willing to pay in order to avoid a certain risk; #$, #& are the levels of utility with

and without insurance; L stands for household assets; Z denotes the vector of households and farm characteristics; ) represents the risk probability; * is the risk aversion and + represents the (other) unobserved factors. Results show that income, age, education, tractor ownership, credit availability, access to extension services and landholding had a positive effect on WTP for crop and food insurance, whereas crop diversity and non-farm participation had a negative influence on WTP.

Kpadé et al. (2017) published one of the most recent studies, using a partly different methodology than the previously introduced literature. When analysing cotton farmers’ willingness to pay for pest management services in Northern Benin, a double bounded contingent valuation survey is used for data collection and an interval regression model for data analysis. The double bounded contingent valuation survey differs from the standard CVM, as follow-up questions are asked depending on the answer of the individual, to get more detailed information. Answers are organized as left censored if negative, or right censored if positive. The interval regression then estimates the probability that a latent variable exceeds one threshold but is less than another threshold, meaning, it estimates the probability of the latent variable lying within a certain interval. 95 percent of farmers were willing to pay for pesticides, with four factors mainly affecting the farmers’ WTP decision, namely an increase in land size having a negative effect on WTP, and higher level of education and higher percentage of cotton area in the field having a positive effect on willingness to pay.

(16)

10 name already suggests, it is used to measure willingness to pay at auctions. Skiera & Revenstorff (1999) use this technique to measure students’ willingness to pay for mobile phone contracts, while testing to sell contracts via an auction during a university lecture. (The method is, that) at a vickrey auction, all participants are bidding for an article at the same time. The bidder with the highest bid gets the acceptance of the bid, however they only have to pay the price of the second highest bid (that was) submitted. This has the aim of participants bidding at the true valuation of the good. The authors aim on finding out how reliably the vickrey auction method determines the willingness to pay. The results are, that the vickrey auction method produces fairly accurate predictions in terms of willingness to pay in relation to their preference data as well as correlation of prices for different contract models. In their opinion, the vickrey auction model is a promising method to measure prices of willingness to pay at auctions, which has been neglected in most research so far. However, the vickrey auctions may suffer from overbidding bias, which has to be taken into account when using this method. (Breidert 2006)

Last, I want to have a look at another auction form, the Becker, DeGroot and Marshak (BDM) method, used by Wertenbroch & Skiera (2002). In BDM, individuals give an offer for a product at the same time. The sale price is then randomly drawn from a distribution of prices. Bidders with bids higher than the drawn sale price (then) receive a unit at costs of the drawn sale price. Those who bid below the sale price do not receive the good. Similar to the vickrey auction, BDM bid only determines whether an individual has the right to purchase a good., The price however, is set by another mechanism and is always below the participants’ bid. (Breidert 2006) Using this method, Wertenbroch & Skiera (2002) find out, that differences in WTP estimates arise from the incentive constraint in BDM or vickrey auction, in comparison to using cognitive effort required methods like CVM and surveys.

Willingness to pay studies in Ethiopia

Shiferaw & Holden (2002) examine peasants’ willingness to pay to sustain land productivity in Ethiopian highlands, using data sets from 1994. The authors focus on the influence of market imperfections and poverty on willingness to pay, using the contingent valuation method. The main result of their paper is, that peasants are willing to pay only a small fraction (max. 3.5 percent) of external on site costs of their own soil-degrading practices. Also, poverty and liquidity restraints reduce the willingness to pay for land conservation and cash liquidity has a high positive effect on willingness to pay for all three named services.

(17)

11 in numbers of visits by an extension agent to the farmers in twelve months. Further independent variables are farmers’ walking distance to the milk group, education, the number of crossbreed cows milked, the number of local cows being milked and two dummy variables depending on which association the farmers belong to. The dependent variable, again, is the farmers’ willingness to pay for extension services. Using data from 168 Ethiopian households, Holloway & Ehui’s research showed the following results using the probit model: distance had a negative effect, all other factors including extension visits had a positive influence on the willingness of farmers to pay money for additional extension visits, hence leading to extension visits themselves being a potent catalyst stimulating entry into emerging milk-markets.

Asrat et al. (2004) examine farmers’ WTP for soil conservation practices in Ethiopian highlands. Asking 100 randomly selected households, the authors use the contingent valuation method to collect data and the logit model for assessment of the data, to find out about the WTP in relation to a number of explanatory variables including age, education, sex, income, family size, attitude of farmers, awareness about technology and land size. Results of the study are, that 63 percent of households were willing to pay for soil conservation practices, with literacy having a huge impact on willingness to pay. Whereas 76 percent of literate farmers are willing to pay, it is only 16 percent of illiterate farmers willing to do so (total amount of literate farmers being 54 percent). Furthermore, education, perception of soil-erosion hazard, land size, attitude towards soil conservation, family size and the slope of the farmland owned influenced farmers’ willingness to pay most. Last but not least, farmers stated to be less willing to pay in cash for services, but are rather willing to spend a higher amount of working time as payment.

3.3 Key facts and resulting research hypotheses

General willingness to pay for mechanization

(18)

12 Based on those facts (mentioned above) and the fact that farmers in Humera are organized in cooperatives, which makes land clustering possible, I assume that farmers in Tigray region are willing to pay for machinery under the right circumstances. Hence my first hypothesis is as follows:

H1: If tractors which work on the heavy soils in Humera are accessible in a rental model, smallholder farmers would be willing to pay for it instead of using manual labor.

Impact of socioeconomic factors on willingness to pay

When assessing the willingness-to-pay studies, different variables influence the decision making of farmers. However, some factors had a positive influence in almost all willingness to pay papers. Both, income working experience and education had a positive influence in almost all papers, which tested those variables, with the exception of Bakopoulou et al. (2010), where farmers with higher income rather used fresh water than recycled water. In the same paper, gender is tested positively for WTP, with female candidates being more willing to pay for recycled water than male candidates. Land size correlated positively for Ahmed et al. (2015), Ulimwengu & Sanyal (2011) and Asrat et al. (2004) , but negatively for Kpadé (2017). Age had no major influence on WTP when tested by Uddin & Gao (2016) and Asrat et al. (2004), but a positive impact in encountered in the study by Ali (2013).

Drawing from those results, I expect the following socioeconomic factors, to have a positive influence on willingness to pay: Gender, age, education, farming experience, land size, and income. My second hypothesis is therefore:

H2: The socioeconomic factors gender, age, education, farming experience, land size and income positively correlate with farmers’ willingness to pay for mechanization.

In addition, even though it was not part of any of the reviewed studies, I want to test the impact of family labor on willingness to pay. As from my experience, employing family members for farming activities is quite common among farmers in the Tigray region, I want to test whether those farmers are less willing to pay for mechanization, as this could decrease employment possibilities for family members. Hence, my third working hypothesis is:

H3: Family labor has a negative impact on farmers’ willingness to pay for mechanization.

(19)

13 farmers are given the choice of differently priced machinery packages, or whether all farmers would go for the cheapest option. The fourth and the last hypothesis is:

H4: The socioeconomic characteristics gender, age, education, farming experience, land size and income and family labor also impact the decision making on different machinery packages.

4 Methodology

When measuring willingness to pay, one has different methods to choose from. Taking into account results from previous studies, I decided to use the contingent valuation method for data collection, and the probit model for data evaluation. CVM is in my eyes the best method, as it focusses on assessing the value of environmental goods, and is said to be the best method when assessing WTP in environmental areas, making it the most relevant method for my research. Other methods, like the vickrey auctions method or BDM model focus more on how much an individual is willing to pay for a product, using auctions. I however, want to find out the individuals’ general willingness to pay for a good, and not the price at which it is willing to pay for the product. This is important as no market in the research area exists at the moment, hence an auction would not make sense in our case.

For evaluating CVM data, two econometric models can be used, either the logit model or the probit model. Both are binary outcome models and both are used when having a dichotomous dependent variable, hence both estate the probability that an outcome variable takes the value 1. The two models produce almost identical results and both can be applied for evaluation of the CVM. In this study we use the probit model, as the logit model is said to fit better for studies, where coefficients can be interpreted in terms of odds ratios, which is mainly the case in health care studies. In addition to the probit model, I use a multinomial probit model for evaluation of machinery packages, preferred by farmers, which is further explained in par. 4.3. The multinomial probit model is based on the probit model.

4.1 Contingent Valuation Method

(20)

14 It is mostly applied for non-market values1 , non-use values2 or both. Furthermore, CVM is popular when researching individuals’ willingness to pay for preserving a resource base for future use (option value) or measuring the willingness to pay for preserving a particular resource merely for the sake of its existence. (Venkatachalam 2004)

CVM uses sample surveys (questionnaires) to elicit individuals’ willingness to pay for generally hypothetical projects or questions. Its name refers to the values of respondents, which are contingent upon a simulated market presented in the questionnaire. (Portney 1994) The survey usually contains a description of the hypothetical project, that is asked about and which is evaluated in the questionnaire. The description should give the respondent a clear picture of the good that he is asked to value. Second, the survey must contain a mechanism to elicit value from the respondent. Those can be open-ended questions, yes or no questions, as well as option-questions (Which would you be most willing to pay for? A,B,C,D?), which ask the participant about their willingness to pay for a certain good. Last, CVM uses socio-economic characteristics such as age, income, gender et cetera, as well as environmental attitudes of the participant. (Portney 1994) Whereas the willingness to pay questions represent the dependent variable in the survey, the socioeconomic as well as environmental attitudes represent possible dependent variables, which influence the participant’s choice.

Even though CVM is the most used economics survey approach (Gregory 2000), the model is criticized for different reasons. The main criticism lies within the validity and the consistency of the CV results, with validity referring to the accuracy of CV results, and reliability referring to the consistency of CV, meaning that the same results are difficult to be reproduced in the same way again. (Venkatachalam 2004) In terms of validity for example, Freeman et al. (2014) argue, that there is no true market value to certain environmental factors, and it is not possible to measure their willingness to pay, as costs are unknown, or participants of the survey don’t know the value of certain factors. (Freeman et al. 2014)Damigos & Kaliampakos (2003) furthermore criticize, the hypothetical nature of the CVM as well as the format of the questions asked. In their opinion, results depend on information given to the respondents about the project in the questionnaire, since they don’t have a direct comparable model, as it is a hypothetical scenario.

Loomis (1989) researched in his paper only the reliability of the CVM, by re-surveying the same households, nine months after the original CVM data collection. Using a paired t-test, he found no statistical differences between the first and second reported willingness to pay, supporting the contention that the

1 Non-market values: environmental goods and services that are not traded in a market (Champ et al. 2003)

(21)

15 contingent valuation method provides reliable results when assessing individuals about their willingness to pay. These results are supported by Kealy et al. (1990), who also analyze the reliability of CVM. Using CVM to find out about students willingness to pay for a particular product from the food sector, reproducing those results two weeks later, they confirmed the reliability of the CVM.

Even though, the method has experienced critics, I still decided to use it for the following reasons: the factor of the hypothetical nature of the test will be the same, even when using a different test and it is the point of the research, to find out whether farmers are willing to pay for modern technology, that they don’t have yet. Difference to other research is however, that farmers already use few, rather old mechanization techniques, and know about benefits and bottlenecks that come with mechanization. Even though there is currently no market for mechanization in the tested area of Humera, farmers still have certain knowledge about possibilities of mechanization, therefore the critics of participants not knowing about the value and hence having validity issues don’t apply in this case.

CVM Model

The contingent valuation model applied for this thesis is as follows: My aim is to find out about small-scale sesame farmers’ willingness to pay for modern machinery, meaning new tractors as well as implements such as ploughs, row planters, weeding machinery and harvesters. We want to find out about the farmers’ cost-benefit assessment for their own farms, mainly whether farmers are willing to spend money for an expected yield increase through new machinery. Currently, farmers in Humera have few 50-60-year-old tractors and old ploughs, nothing else. There is currently no modern machinery rental available in the region, hence we talk about a non-market value. It is a hypothetical approach, as farmers never used modern machinery before, and it is also not sure whether modern machinery will be introduced into the region. For the evaluation, I created a questionnaire, based on the CVM approach. The questionnaire was divided into three parts.

First I ask about the following socioeconomic factors:

gender, age, education, average yearly income, years of farming experience, owned land size, and how much family labor is used for farming activities. Those variables will be used as independent variables for the study, to find out whether they influence the WTP or not.

(22)

16 Table 1: Coding of the variables in the research survey

Gender the value 0 if female, the value 1 if male

Age the actual value of the farmers age, e.g. 42 if the farmers is 42 years old

Education the value 0 if the farmer didn’t attend school, the value 1 if primary school was attended, the value 2 if secondary school was attended

Income the actual value of the farmers farming income per year in Ethiopian Birr, e.g. 10.000 if the farmers earned 10.000 Birr per year

Farming experience the actual value of the farmers experience in years, e.g. 10 if the farmer has 10 years of farming experience

Land size the actual value of the farmers land size in hectares, e.g. 5, if the farmer owned 5 hectares of land

Family labor the value 1, if the farmer doesn’t use any family labor, the value 2, if the farmers uses up to 20 percent family labor to prepare his fields, the value 3 if the farmers uses 20 to 50 percent family labor and the value 4 if the farmers uses more than 50 percent family labor

In the second part, I asked farmers about current machinery use and knowledge, namely:

1. Do you currently use tractors or farming machinery? a. If yes, is it sufficient for your needs?

b. If you don’t use machinery, why not?

2. Do you think new tractors and implements will increase your sesame output?

Aim of the second part is to find out about expectations of farmers towards mechanization, and whether they are familiar with machinery already or not. The third part is about the farmers’ willingness to pay, however I took two approaches at the same time.

First I asked them about their pure willingness to pay for mechanization, with a fixed price of 6500 Birr per ha. with a yes or no answer possibility. This aims on finding out on farmers’ general willingness to pay for mechanization. In the next question, I let farmers choose between different machinery packages for different prices, and different implements in the package. Aim is to see which package farmers choose, with the following options.

(23)

17 2. Tractor + field cleaning implement + deep plough + row planter + weed control, costs of 4900 Birr per ha. This package offers the best quality in terms of yield increase, however is the second most expensive option. 3. Tractor + field cleaning implement + deep plough + row planter + weed

control + harvester, costs of 6500 Birr per ha. This package offers both, best quality in terms of yield increase, as well as mechanical harvesting. Currently, harvesting is done by manual labor, and therefore expensive for local farmers. In my opinion, even though it is the most expensive option, it is the best in terms of cost-benefit aspects. However, at this moment, mechanical harvesters for sesame are not available in Ethiopia. 4. Tractor + deep plough + weed control + harvester, costs of 4200 Birr per

ha. This solution mechanizes all steps which are currently very labor intensive and therefore expensive for farmers. If someone only aims to reduce farming costs, this solution would be the most interesting one. 5. None of the above

Prices were calculated using price information from tractor and equipment manufacturers, including costs for drivers and running costs as well as interest costs for a possible bank loan for purchasing machinery in the first place. These costs were then transferred to costs per hectare, since this is the unit of how farmers currently rent their old machinery.

In addition to the questions about willingness to pay, I asked whether willingness to pay would be influenced by different payment methods, what expectations farmers have from mechanization, and whether or not they would be willing to rent machinery without having seen it before. All exact questions and answer possibilities can be seen in the questionnaire in Appendix 1.

The whole questionnaire was translated from English into Tigrayan, the local language in Tigray region. Before filling out the survey, farmers were explained about mechanization, the machinery that is talked about in the survey, as well as previous results from trials, regarding yield increase, and time needed for the farming activities. One challenge when filling out the interviews was, that many farmers were illiterate. In that case, the questionnaire was read by a translator from the company Selet Hulling to the farmer, and filled out by the translator, according to the farmer’s answers.

(24)

18 The socioeconomic data from part one of the questionnaire are used as independent variables. I want to find out, whether those independent variables influence the decision, on whether farmers are willing to pay for mechanization and whether it influences which package the farmers are willing to pay for. In order to analyze the questionnaire from the CVM, we use two different models, depending on which dependent variable we examine. We use the probit model to find out about the pure willingness to pay and the multinomial probit model to examine the package option of willingness to pay.

4.2 Probit Model

The model

The probit model is a binary outcome model, meaning that the dependent variable can only take on two values, zero or one. This is the case for the willingness to pay variable in this study, when farmers are asked whether they are willing to pay 6500 Birr per ha for mechanizing or rather continuing farming without modernized mechanization. The y variable, which is the dependent variable WTP is therefore 1, if they are willing to mechanize for that price, or 0 if not. The following function displays this probability of the binary outcome model, where the probability of y=1 is calculated.

- = -/ 0 = 1 2 = 4(2´6)

Figure 4: Nonlinear probability curve of the probit model (Hill et al. 2011)

(25)

19 and later increases at a much slower rate. In this model we calculate the probability that the independent variable y takes the value 1.

The probit function represents this curve and is as follows: 4 2´6 = Φ x´6 = 9 : ;:

<´= >?

The formula means that, the probability of y=1 given x´6 is 9 : ;: with 9 : being the probit function (cdf of the standard normal distribution).

Interpreting the coefficients of the probit model

The probit model estimates the maximum likelihood method, meaning that an increase in the independent variable x, increases/decreases the likelihood of y=1, hence making the outcome more or less likely. In the basic probit model, only the sign of the result is interpreted, not the magnitude however. This is, as different models have different scales of coefficients, which makes it impossible to interpreted the magnitude of the coefficients. This becomes clear when the probit model is compared to the standard OLS model, where 6probit ≅2.56OLS. This means that the coefficients of the probit model are about 2.5 times the coefficients of the OLS model. As a result, it can be only said whether the probability is more or less likely, depending on the sign of the result.

Marginal effects of the probit model

In order to solve the problem, of interpreting the magnitude of our results, the marginal effects will be calculated. From those results, we can see the probability of our dependent variable being one (y=1), given a one-unit change in our independent variable x.

The marginal effects for the probit model are calculated as follows:

A-A2B = Φ(2´6)6B

This expression shows the effect of an increase in x on the probability p. The latter j stands for the jth independent variable of the model. In our case we have seven independent variables, equaling seven coefficients which are estimated. The effect depends on the slope of the function, which is shown by Φ(2´6) and the magnitude of the function, which depends on 6B. Φ(2´6) is a probability density function, meaning that its value is always positive. Hence, the sign of CD

C<E

(26)

20 has to be estimated at a specific value of x, either the mean or the average. In this thesis, the marginal effects are calculated at the average marginal effect. This is due to independent variables such as gender, where the mean marginal effect is not useful.

The average marginal effects are calculated as follows:

A-A2B =

∑Φ 2´6 6B J

with n being our sample size. With the probit model, we can calculate partial effects of discrete variables. In our case for instance, we can calculate the change in probability of y=1, depending on whether our participant is male or female. This is calculated as follows:

4 (6K+ 6$<$+ 6& M + 1 − 4(6K+ 6$<$+ 6& M )

In this formula, k represents the discrete variable. (e.g. is the participant male or female?). The variable k takes the value 1 for males in the first part of the formula 4 (6K + 6$<$+ 6& M + 1 and 0 for females in the second part 4(6K+

6$<$+ 6& M ). 6 depicts the estimated coefficients in our model. The difference

of the two functions show the effect of a change from male to female for the probability of y=1. Here, the magnitude can be interpreted, meaning that an increase in x increases/decreases the probability of y=1 by the percentage shown in the result of the formula. Both the magnitude as well as the sign of the result are of importance.

4.3 Multinomial Probit Model

The model

The multinomial probit model differs from the ordinary probit model, as its dependent variable is not binary, but it can take on more than two values. In this research, this is the case when we ask farmers about their preferable choice of machinery packages, where they can choose between five different options. Important for the multinomial model is, that the individual can only select one alternative, the different choices (j) are coded numerical (j=1,2,...,m) and those numbers are only used as codes, and their magnitude cannot be interpreted. As these numbers are only codes, the means or average cannot be used to summarize the dependent variable, however their frequency can be used instead. As our dependent variably y can take different values, it is referred to as y=j.

In the multinomial probit model, we estimate the likelihood, that individual i, chooses alternative j. The probability function for this is as follows:

(27)

21 The functional form of 4B is used to assure that the probabilities lie between 0 and 1. The sums of all j’s are therefore always 1. Hence, the coefficient interpretation for j is as follows: when we have an increase in the independent variable, the choice of our dependent variable j is more or less likely in comparison to the base alternative.

Independent variables of the multinomial probit model

For the multinomial probit model, two kinds of independent variables exist. Alternative-invariant and alternative-variant regressors. We only focus on the first one, as this is the case in this study.

Alternative-invariant regressors only vary over the individual i, and not over the choice j. This is the case for our independent variables (age, education, income, etc.). These differ for each individual, but do not change depending on which package of dependent variable the individual chooses.

Marginal effects

Similar as for the binary probit model, we again can calculate the marginal effects for the multinomial probit model. However, in the multinomial model, the marginal effect is an increase in the regressor on the probability of choosing alternative j.

A-OB

A2O = -OB(SB− ST)

This means, the effect of a change in the independent variable 2O on the

probability that individual I chooses alternative j is -OB(SB− ST). In this term, ST

represents an average of all coefficients and SB depicts the chosen alternative. As it is unknown whether SB or ST is bigger, we do not know the sign of the

coefficient. Hence, unlike in the binary model, in the multinomial model, the marginal effects don’t always correspond in sign to the coefficients.

The interpretation of the marginal effects is as follows: a one-unit increase in the independent variable increases or decreases the the probability that the individual chooses alternative j by the result of the marginal effect. This is expressed in percentage points.

4.4 Analysis and results

(28)

22 Figure 5: Summary of dependent and independent variables

Variable Mean Std. Dev. Min Max

WTP 0.5058 0.5029 0 1 Gender 0.7882 0.4109 0 1 Age 46.2705 14.5327 19 75 Education 0.2352 0.4266 0 1 Income 25100 20188.81 5000 110000 Experience 14.3058 11.7848 1 50 Land Size 8.7529 8.6118 1 50 Family labor 1.7058 0.8139 1 4 CurMachine 0.9529 0.213 0 1 Bottlenecks 2.0471 0.3417 1 3 Incr. Out 1 0.1543 0 2 Pay. Method 1.964706 0.1856 1 2 Reason 1 0 1 1 Testing 1 0 1 1 Observations 85

As can be seen from figure 5 above, the dependent variable willingness to pay (WTP) has a mean of .5058. This means that 50.8 percent of farmers asked, are willing to pay for machinery and 49.2 percent are not willing to pay. This shows no clear trend on farmers’ willingness to pay.

Figure 6: Explanation for negative willingness to pay results

Variable Mean Std. Dev. Min Max

Expensive 0.9047 0.2971 0 1

Observations 42

Farmers who answered if they are willing to pay for mechanization with no, were asked if the reason was, that the price for mechanization is too expensive. Figure 6 displays the results. The number 0 means that this is not the reason, the value 1 was crossed in the survey if this is the case. As can be seen from figure 6, 90 percent of farmers who were not willing to pay for machinery, stated that the reason is that prices are too expensive for them.

(29)

23 attended secondary school can be explained with the fact that in rural areas, no secondary schools exist yet. Only in the city Humera itself, students have the possibility of attending secondary school, and those having done this try to work in companies rather than having their own small farm.

The mean income of farmers is 25.000 Ethiopian Birr per year, which is about 760 Euro. However, there are big differences in income (min. 5000 Birr / 150 Euro, max. 110.000 Birr / 3340 Euro) as well. Those differences can be explained through different land sizes (mean of 8.7 ha, min: 1 ha, max 50 ha) and differences in productivity and experience (mean farming experience 14 years, min: 1 year, max: 50 years).

Asking about income and farm size is a sensible issue for farmers, as some are afraid, that information will be given to tax authorities for comparison. However, as questionnaires are filled out anonymously and farmers were not asked for names, farmers didn’t have any problems with stating those numbers and also the reliability of the numbers are ensured.

When looking into farming techniques, and how much of the labor is done by family members, the distribution is as follows:

Figure 7: Distribution of family labor used for farming

Amount Fam.labor Freq. Percent Cum.

0% 43 50.59 50.59 0-20% 15 29.41 80 20-50% 16 18.82 98.82 >50% 1 1.18 100 Total 85 100

As figure 7 shows, the majority of 51 percent uses no family labor, 29 percent of farmers use up to 20 percent family labor, for 19 percent, up to half of their workers are relatives and only 1 percent of farmers has more than 50 percent family labor.

To find out more about the current situation of farmers, they were asked about current machinery use, described by the variable CurMachine.

As can be seen from figure 5, the results are that 96 percent of farmers are already using machinery at some point for farming activities. That means that most farmers know about benefits of machinery, in terms of being faster and saving costs for manual labor. But they also know about problems that occur with machinery regarding costs for reparation and downtime.

(30)

24 Figure 8: Main bottlenecks of machinery use

Bottlenecks Freq. Percent Cum.

Affordability 3 3.53 3.53

Accessibility 74 88.24 91.76

Other reasons 7 8.24 100

Total 85 100

Only 3.5 percent of farmers say the machinery is not affordable, 88 percent stated that the accessibility to machinery is insufficient and 8 percent stated other reasons as the main bottleneck. This is additional information that does not directly influences the probit model, however it shall help to understand the situation and motivation of farmers’ willingness to pay and understand why no modern machinery is being used so far. This is especially interesting, that even though half of the farmers asked would not be willing to pay for machinery because it is too expensive, a majority sees accessibility still as a bigger problem than affordability. This is of big importance when thinking about introducing a machinery rental model in the region.

When asked about their expectations of machinery use, all farmers expect from modern machinery mainly an increase in yields but not a decrease of farming costs. This is important as the main reason for mechanization indeed is a yield increase and if farmers would want to lower their costs by mechanizing, this would mean that they have a wrong idea of the advantages and disadvantages of mechanization. Selet Hulling performed a trial in 2017 where two areas next to each other were compared. One field was prepared mechanically, using tractors for ploughing and a row planter for seeding, the second field was prepared with manual labor as usual. The field which was prepared with machines had yields of eight quintals per hectare, whereas the manually prepared field had yields of only five quintals per hectare. This showed that under the same conditions, using mechanization can lead to a massive yield increase, making mechanization worthy in the first place.

Furthermore, farmers were asked about their preferred payment method for machinery, with the value 1 representing cash payment and the value 2 standing for payment in product. 96 percent of farmers stated that their preferred payment method is paying in product instead of cash (fig. 5). All farmers stated that their main reason for wanting to mechanize is an expected yield increase (fig. 5) and not the expectation of lowering costs. This information also has to be taken into account when thinking about introducing machinery to Humera. Farmers should be given the possibility to pay for machines in product instead of cash.

(31)

25 to whether farmers initially trust external companies to bring in good machinery to the region. On the other hand, as machinery is rented per hour, farmers don’t have a high risk in trying out new machinery without having seen it before. Results of the probit model

To better interpret the results of the probit model, the average marginal effects were calculated.

Figure 9: Average marginal effects of the probit model Probit Marg. Effect WTP Gender -0.1666 (-1.40) Age -0.0055 (-1.49) Education -0.7413 (-0.65) Income 0.0000148** (3.03) Experience 0.0005 (0.12) Land Size -0.0035 (-0.31) Family labor -0.1762** (-2.88) CurMachine 0.3241 (1.32) Bottlenecks -0.0387 (-0.26) Payment 0.1103 (0.42) Observations 85 Pseudo R2 0.3876 t statistics in parentheses ="* p<0.05 ** p<0.01 *** p<0.001"

The Pseudo R2 replaces the R2 of the OLS model. The McFadden pseudo R2 has its optimal value between 0.2-0.4, indicating a very good model fit. We have a pseudo R2 of 0.39. This means that our model has a very good fit. The pseudo R2 ranges from 0 to 1.

(32)

26 10 percent level. An additional year of age also has a small negative effect on WTP of 0.55 percent, meaning that the older the person is, the less likely it is that he is willing to pay for machinery, but results are not relevant at the 10 percent level with a significance of P>0.140. statistically significant, same goes for education. Income has a small positive effect and each additional unit of income makes farmers 0.0014 percent more likely to be willing to pay for the proposed machinery. Even though experience and land size are positively correlated with willingness to pay, results are not statistically significant. Problems with significance are due to the number of observations being only 85. During the time of the research, farmers were in the middle of harvesting season, which lead to very little availability of farmers for questionnaires. In addition, as I explained to each farmer details about the mechanization program, in order to have valid results, and the fact that all questions had to be translated by a translator, questionnaires were very time consuming. Hence, even though I took several weeks for my research in Ethiopia, it wasn’t possible to perform more than 85 questionnaires with farmers.

Family labor makes farmers less likely to be willing to pay for machinery by almost 18 percent. The results are highly significant. This can be explained, as machinery replaces manual labor and therefore the family labor is replaced with more high-skilled jobs such as drivers and mechanics which are needed for mechanization. However, as neither, most of the farmers, nor their families have these skills, the employment of the own family would decrease through mechanization. Another reason for the effect of family labor is, could be low opportunity costs, which lead to the refusal of mechanization. Current machinery use, the statement of bottlenecks and the preferred payment method are again not statistically significant.

Concluding, it can be said that income and the amount of family labor are key elements of farmers’ willingness to pay, whereas the factors gender, age, education, experience and land size have no significant impact.

(33)

27 Figure 10: Classification of probit model for WTP

Classified Willing to pay Not willing to pay Total

+ 32 6 38 - 11 36 47 Total 43 42 85 Sensitivity Pr (+ | Willing to pay) 74.42% Specificity Pr (- | Not willing to pay) 85.71% Positive predicted value Pr (Willing to pay | +) 84.21% Negative predicted value Pr (Not willing to pay | -) 76.60% False + rate for true "Not willing to pay" Pr (+ | Not willing to pay) 14.29% False - rate for "Willing to pay" Pr (- | Willing to pay) 25.58% False + rate for classified + Pr (Not willing to pay | +) 15.79% False - rate for classified - Pr (Willing to pay | -) 23.40% Correctly classified 80% Classified + if predicted probability for willingness to pay >= 0.5

As can be seen from figure 10, 32 values were correctly predicted as willing to pay, whereas 6 were wrongly predicted as willing to pay. Furthermore, 11 values were falsely predicted as not willing to pay and 32 were correctly predicted as not willing to pay. In total, the model has a correct prediction of 80 percent which is still a good fit.

Results of the multinomial probit model

In the multinomial probit model, we have four different dependent variables, listed as 1,2,3,4. They stand for the different options farmers can choose from: Table 2: Machinery Options for farmers

1. Basic Machinery Option Tractor, Rowplanter, 1.400 Birr 2. Advanced Machinery Option Tractor, Field cleaner, deep plough,

row planter, weed control, 4.900 Birr 3. Full Package Tractor, Field cleaner, deep plough,

row planter, weed control, harvester, 6.500 Birr

4. Cost Saving Package Tractor, deep plough, weed control, harvester, 4.200 Birr

Those four packages are our dependent variable and I aim on finding out in how far our independent variables from before, influence the choice. As explanation to the four options:

(34)

28 Option 2 includes all steps for proper farming, including field cleaning, ploughing, planting and weed control and promises the best results in terms of quality and quantity in farming.

Option 3 adds mechanical harvesting on top of the equipment of option 2. Harvesting is very expensive currently, however, it is very difficult to mechanize, hence some farmers might not want it as it is still too complicated to handle. Option 4 mechanizes only the most labor intensive steps of ploughing, weed control and harvesting. Farmers who only aim on reducing manual labor costs will choose this package.

Figure 11: Frequency of the multinomial dependent variable

Package Freq. Percent Cum.

Option 1 for 1.400 Birr 34 40 40 Option 2 for 4.900 Birr 11 12.94 52.94 Option 3 for 6.500 Birr 29 34.12 87.06 Option 4 for 4.200 Birr 11 12.94 100 Total 85 100 The frequency of the dependent variables can be seen in figure 11. Clearly, option 1 and 3 are the most preferred ones, with 40 respectively 34 percent of farmers choosing them, whereas only 13 percent preferred option 2 and 4. This means that farmers choose either the cheapest model or the most expensive one, but don’t drive for the in-between solutions.

Figure 12: Marginal effects of the multinomial probit model

Option 1 Option 2 Option 3 Option 4

Gender 0.1384 -0.1718 0.0641 -0.0307 (1.01) (-1.22) (0.45) (-0.30) Age 0.0033 0.0028 -0.0077 0.0014 (0.65) (0.74) (-1.5) (0.45) Education 0.0787 0.0262 -0.1185 0.0135 (0.52) (0.21) (-0.79) (0.13)

Income -0.0000122* 5.58E-0.6 9.04E-06 -2.38E-06

(35)

29 Payment – – – – – – – – Observations 85 z statistics in parentheses * p<0.1 ** p<0.05 *** p<0.01"

When looking at the marginal effects of the multinomial probit model (figure 12), we have similar results as we have in the earlier used probit model.

For option1, income is significant at the ten percent level and family labor is significant at the five percent level. Whereas a higher income has a negative impact on willingness to pay, family labor has a positive influence. This makes sense, as farmers with higher income, most probably favor a more expensive option, which is likely to produce better farming results. On the other hand, in the cheapest option, only few steps are mechanized, and labor intensive work such as harvesting or weeding still needs to be done by manual labor.

Both, option 2 and option 4 didn’t show any significant results due to the low number of farmers choosing this option.

For option 3, we have a slightly different result towards the normal probit model. Land size has a positive effect and is significant at the 10 percent level. The result shows that with each additional ha land, farmers are around three percent more likely to be willing to pay for option 3. As before in the probit model, family labor is also a significant variable, having a high negative influence. With each unit increase in family labor, the likelihood of farmers’ willingness to pay decreases by 30 percent. This makes sense, as already explained in the probit model before, farmers don’t want to replace their own family by machines. There was no fit for the variable payment, as all iterations kept running with non

concave.

Figure 13: Predicted probabilities of the multinomial probit model

Variable Mean Std. Dev.

Option 1 for 1.400 Birr 0.3954 0.2571 Option 2 for 4.900 Birr 0.1312 0.1037 Option 3 for 6.500 Birr 0.3432 0.2788 Option 4 for 4.200 Birr 0.1301 0.0755 Observations 85

(36)

30 Comparison of the results to the hypotheses and to previous results of the literature

Going back to our four hypothesis, which resulted from previous literature, we have the following results from our own research:

H1: If tractors which work on the heavy soils in Humera are accessible in a rental model, smallholder farmers would be willing to pay for it instead of using manual labor.

Only 50 percent of farmers asked in this survey are willing to pay for mechanization. Therefore, the first hypothesis has to be rejected, as this number is too small to confirm the hypothesis. This does not mean that mechanization is not wanted in the region, but rather that there is no clear majority who is willing to pay.

H2: The socioeconomic factors gender, age, education, farming experience, land size and income positively correlate with farmers’ willingness to pay for mechanization.

The second hypothesis also has to be rejected. Only income had a positive effect on willingness to pay, all other variables were not significant.

H3: Family labor has a negative impact on farmers’ willingness to pay for mechanization.

The third hypothesis can be approved. Results are statistically significant and confirmed that an increase in family labor decreases the likelihood of farmers’ willingness to pay for mechanization.

H4: The socioeconomic characteristics gender, age, education, farming experience, land size and income and family labor also impact the decision making on different machinery packages.

Referenties

GERELATEERDE DOCUMENTEN

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of

De wortels van de planten op de bakken die bij het bedrijf waren ontsmet (behandeling 2, hoge druk met 1,6% formaline) had- den ook verkleurde wortels, maar vertoonden

Daarnaast wordt er ook pas na de training een samenhang verwacht tussen externaliserende gedragingen en de mate van mindfulness, doordat de jongeren voorafgaand aan de training

In het archief zouden dan niet alleen de officieel naar het stadsarchief overgebrachte digitale archieven opgenomen moeten worden, maar ook alle semi-statische archieven die bij de

Despite NIBP devices being accurate on delivery by the manufacturer, subsequent regular recalibration is Figure 2: Mean arterial blood pressure measurements—Bland–Altman ratio

Joseph Conrad identity postcolonial space third space colonial discourse race culture gender... 4 CONRAD: POSlSlONERlNG VAN IDENTITEIT

Gedacht wordt aan een set kwaliteitscriteria vanuit cliënten- en familieperspectief voor de zorg en ondersteuning van ouders en verzorgers rond voedingsproblemen bij jonge

Met dit formulier kunt u iemand machtigen om namens u bezwaar te maken tegen een beslissing van Zorginstituut Nederland. In dat geval moet u het machtigingsformulier