Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Kenya smallholder farmer education and farm productivity Mbwika, James M. 1990

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1990_A6_7 M38.pdf [ 6.02MB ]
Metadata
JSON: 831-1.0098394.json
JSON-LD: 831-1.0098394-ld.json
RDF/XML (Pretty): 831-1.0098394-rdf.xml
RDF/JSON: 831-1.0098394-rdf.json
Turtle: 831-1.0098394-turtle.txt
N-Triples: 831-1.0098394-rdf-ntriples.txt
Original Record: 831-1.0098394-source.json
Full Text
831-1.0098394-fulltext.txt
Citation
831-1.0098394.ris

Full Text

KENYA SMALLHOLDER FARMER EDUCATION AND FARM PRODUCTIVITY By James M. Mbwika B. A. (Econ) (University of Nairobi) 1986. A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES AGRICULTURAL ECONOMICS We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April. 1990 © James M. Mbwika, 1990 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Agricultural Economics The University of British Columbia 6224 Agricultural Road Vancouver, Canada V6T 1W5 0.1 Abstract This research was undertaken to study the effect of education on small farm revenues and profits in Kenya. Schooling (denned as the number of school standards completed by the farm operator) was used as the most important source of education. It was hypothesized that schooling has a positive effect on farm revenues and profits. The effect of other sources of information viz; extension contact, demonstration attendance and baraza attendance on farm revenues and profits were also investigated. The research was done using regression analysis where these variables and other farm activity relevant variables were fitted in regression equations. The choice of these vari-ables were based on economic theory, Kenya small farm characteristics and the objective of the study. Several factors would qualify as supporting evidence for the argument that educated farmers are more productive. We expect educated farmers to be more informed in terms of use of new production technologies. Education as a source of human capital also enhances the productive abilities of human beings and also enables those who have invested in education to use their resources more efficiently as well as adjusting to new "ways of producing more efficiently". In the current study we find that schooling of the farm operator is positively related to level of expenditure on farm purchased variable inputs. This indicates that education enhances adoption of new technologies and innovativeness. Further it was shown that farmers with more education earned more value added per acre from their farm business compared to their less educated counterparts. On the overall farm activity, farmers with eight or more standards of schooling earned upto 80.2% in value added per acre compared to those who had no schooling. The regression estimates were done on a stepwise procedure where farm specific enter-prises were estimated separately and then aggregated and estimated as one farm sector. n Thus a crop equation, a livestock equation and a total farm output equation were esti-mated. This model was then developed into a variable profit function. A simple linear function procedure was used in the regression analysis. In all the estimated value added equations the schooling coefficient was positive and significant at 5% level two tail t-test. As we move from farm specific activities to a farm aggregate output model and lastly to value added model the schooling coefficient increased in size confirming the positive role of education in allocative effect. These results show that schooling plays an important role in allocation of other purchased inputs and also choice of crop mix and input selection. The estimated marginal return to schooling of farm operator in the profit function was Kshs.281. In an earlier function where schooling of the farm operator was fitted into a total farm income equation the estimated marginal return to schooling was Kshs.778.89. When schooling of the farm operator is allowed to interact with extension service the estimated interaction variable coefficient is negative showing the two act as substitute sources of knowledge, and the schooling coefficient increased in size showing that those who had both schooling and extension service earned comparatively more farm revenues. The role of other educative factors like extension service, demonstration attendance, and baraza attendance in influencing agricultural production was investigated. Regres-sion results showed that extension contact had a negative and significant effect on farm revenues and profits. Demonstration and baraza attendance had similar effects on farm revenues and profits. In the value added function hired labour variable was fitted as the cost of hired labour per day. The estimated coefficient for this variable was positive and significant at 5%. The estimated coefficient for this variable shows hired labour is not optimally used, and farmers can increase their farm profits by hiring more labour. When this variable was fitted as the wage rate paid to hired labour per day the estimated coefficient was positive iii and significant. These results indicate that cost of hired labour depends on its quality. In the sales function hired labour was specified as mandays of hired labour per year and the estimated coefficient which reflects the shadow price of labour was higher than average hired labour wage rate implying that this factor is underemployed. In the sales function the estimated coefficient for the value of purchased inputs variable indicates that there is an element of underutilization of these inputs. This variable is fitted in value terms and in profit maximizing conditions the estimated coefficient is expected to be no different from unit. However, the estimated coefficient for this variable is approximately 2.5 showing a shilling spent on purchased inputs will bring forth 2.5 shillings. Thus an increase in the use of purchased inputs will increase farm revenues. Results show evidence of regional differences in farmer productivity and utilization of purchased inputs in favour of Central province. The study is based on the 1982 CBS-IDS-World Bank Household Survey of Rural Kenya data set. iv Table of Contents 0.1 Abstract ii List of Tables vii Acknowledgement viii 1 Introduction 1 1.1 Background 1 1.2 Problem Statement 7 1.3 objectives 9 1.4 Outline 10 1.5 Methodology 11 2 Literature Review 12 2.1 The Small-Holder Farmer in Kenya 12 2.2 Human Capital 18 2.3 Education and Farm Productivity 19 3 Model Specification. 27 3.1 The Sales function approach 30 3.2 Value Added Approach 34 4 Empirical Implementation 38 4.1 Data Sources 38 4.2 Definition of Variables . . 40 v 4.3 Functional Form . 46 5 RESULTS 51 5.1 Aggregate Value of Farm output 60 5.2 OfF-farm and Total Family Income 67 5.3 Value Added Equations 72 6 Summary, Recommendations and Limitations 81 Bibliography 89 Appendix A 92 A. l The Livestock and Crop Output Functions 92 Appendix B 1 0 Q B. l Land Use by Level of Education 100 B.2 Yield levels by Type of Extension Service and Crop Adoption by Level of Education 100 v i List of Tables 1.1 Farm production by education and secondary occupation of farmers . . . 6 4.2 Summary Statistics Mean, Standard Deviation 49 5.3 Correlation between education and other variables 56 5.4 Expenditure on Purchased Inputs Including Hired Labour by Level of Ed-ucation 58 5.5 Value of Farm Output Per Acre and Value Added on Farm Output Per Acre by Level of Education of Farm Operator 59 5.6 OLS Regression Results for Total Farm Production 62 5.7 OLS regression. Results for Total Family Income(TFY) and Off-Farm In-come(OFI) equations 69 5.8 OLS Regression Results For Various Specifications of Value added Equations 75 5.9 OLS Regression Results For Various Specifications of Value Added functions 78 A. 10 OLS Linear Regression Results For'Crop(VCO) and Livestock (VLS) Sectors 97 B. l l Land Use by Level of Education 101 B.12 Value of Farm Output/Acre by Type of Extension Service 102 B.13 Distribution of Crop Adoption by Level of Education - .—102 vii Acknowledgement This thesis would not have been possible without the help of many people. First I would like to extend special thanks to my committee members who made their time available amidst very tight schedules. My advisor, Dr. R. Barichello made enormous contributions and I was able to use his wealth of knowledge on developing countries' economies, not to mention his articulate knowledge of the role of education in agricultural production. Prof. R.C. Allen helped a lot in shaping my regressions,. While Dr.G. Kennedy made insight contributions on the policy issues involved. I am also indebted to R. Robacheau for helping with setting up the data set, Dr.C. Short whose help led to acquistion of the data set, The Kenya Government for funding my studies, Diane Richie always made sure my stipend arrived on time. My fiancee Mwende was very supportive throughout the entire period of working on this thesis. I would also like to thank Dr. A. Bigsten for sending me a copy of the data set. Lastily I dedicate this thesis to my mother Mary Nziva who encouraged me to put up with school and financed my schooling since my childhood. vm Chapter 1 Introduction 1.1 Background ^ Kenya is largely an agrarian economy deriving most of its livelihood from this sector. Agriculture is the largest employer as well as the major foreign currency earner. Further more, over 80 % of the population live in the rural areas and derive their livelihood from this sector. On average agriculture accounts for 30 to 40 % of the GDP per annum compared to manufacturing which on the average contributes about 12 % to GDP per annum. It's contribution to export earnings and wage employment has been on the average of 55 % and 67 % respectively (Kenya: statistical abstracts various issues). The growth rate in Kenya's agricultural output in the first decade after independence (1963-1970) was very impressive averaging about 4.7 %, while the overall GDP grew by about 6.2 % per annum. This growth rate was mainly due to expansion in the cultivated land acreage, yield increase and increased production of high value crops such as coffee and tea1. In the sessional paper No.l of 1986 of economic management for renewed growth 2 the target growth rate of the agricultural sector is set to be 5.3 % per annum. In this policy document the importance of agriculture to the economy's growth is underlined. It says that if the goals of economic development are to be achieved, agriculture will have to:-i) provide food security for 35 million people by the year 2000. 1see World Bank report on Kenya:growth and structural change 1983 p. 325 2 This paper is a government policy document written in 1986 to focus on long term economic policies and act as a basis on which the 5 year development plans will be based. ' 1 Chapter 1. Introduction 2 ii) generate farm family incomes that grow by at least 5 % a year for the next 5 years iii) absorb new farm workers at the rate of over 3 % a year for the next 5 years iii) supply export crops sufficient for a 150 % increase in agricultural export earnings by year 2000, and iv) stimulate the growth of productive off-farm activities in the rural areas, so that off-farm jobs can grow at 3.5 % to 5 % a year. Empirical forecasts by the Kenya Long Range Planning Unit 3(KLRPU) show that the first target is unachievable and that the country will have to import 15 % of its food requirements by the year 2001 and about 30 % by the year 2011. These findings paint a bleak picture of the country's ability to feed itself in the future, and policies have to be designed to address these problems. Although Kenya has a land area of about 575 thousand square kilometres only 7 % can be described as high quality land with adequate rainfall and good soils, 11 % is classified as medium quality and an. additional 4.5 % is arable, but is subject to periodic drought and crop failure. The remaining land is of marginal agricultural productivity and suitable only for ranching. The agricultural sector is composed of a large farm sector of a few plantations operated mainly by large multinational companies, and specializing in growing of cash crops 4 and a small-scale sector which produces for home consumption as well as for commercial purposes. There are about 1.7 million smallholders in Kenya accounting for about 95 % of total farming population. The average size of these holdings is about 2.3 hectares, and 75 % of these are under two hectares. The most basic need in the rural areas is food sufficiency from the family farm. Despite consuming most of their output, smallholders 3A Project set up in 1985 under the Ministry of Planning and National Development, through a bilateral agreement between the governments of Canada and Kenya to develop a long term quantitative multi-disciplinary model of the Kenyan economy and integrate the findings of, the model into a policy making process. 4mainly tea, wheat, barley and coffee Chapter 1. Introduction 3 play a crucial role in commercial production of crops for domestic and export markets. They produce 90 % of maize 5 , 60 % of coffee, 35 % of tea, 45 % of sugar, and almost all marketed production of rice, cotton, and pyrethrum. They also produce most of the marketed milk, beef and poultry. Overall this sector contributes about 75 % of total agricultural output in the country and 85 % of total agricultural employment (Kenya: statistical abstracts various issues). Coffee and tea are second and third after tourism in terms of contribution of foreign exchange earnings. One of the critical problems facing the country currently is the economy's ability to feed the ever increasing population. As noted above the small-scale sector is the major food supplier but its ability to supply food is being constrained by increasing population. Another major problem is that of unemployment, and agriculture being the mainstay of the economy is looked upon as the immediate solution because the industrial sector is not growing fast enough to absorb the increasing labour force. It appears that these problems can only be tackled through increased productivity in the small-holder sector because of it's significant role in the economy (World Bank 1983). Policies that increase agricultural productivity have to be designed with reference to this sector6. An increase in agricultural productivity will lead to rising rural incomes and provide opportunities for easing rural unemployment as well as a slow down in rural urban migration. To feed the ever increasing population, now growing at a rate of about 3.8%, productivity in agriculture will have to grow at a rate that can sustain this population growth. Research, extension service, and marketing policies have been the key policy issues taken by the government to promote agricultural productivity. 5known as corn in north America 61983 World Bank report concludes that "..Kenya agriculture will have to grow faster than it has in the past twenty years, and much faster than it has in the past five years, if there is to be significant economic progress in Kenya during the rest of this century. Moreover, this must be done without many of the conditions which favoured growth in the past, i.e. an expanding supply of high potential land, the rapid growth of several high-value crops whose production had previously been suppressed, and substantial increased yields in the principal food crop, maize Chapter 1. Introduction 4 Elsewhere in the U.S., Canada and even in the developing countries, several studies have indicated that farmer education is a major factor determining farmer productivity and that returns to education in agriculture are high (Barichello 1979; Dunlop 1986; Fane 1972, Khaldi 1975; Petzel 1976, Pudasaini 1983). In 1973 Moock conducted a study on the effect of education on maize yield in a small division of western Kenya and found that those farmers with an education level above 4 years of .schooling produced up to seven more bags of maize per acre than farmers with no education. He argued that, "farmers with higher educational attainment would have an expanded capacity to understand, interpret in light of specific conditions, and apply on their own farms the lessons offered by extension agency". These results are encouraging especially considering that his study was based on a small locality, on one crop, and only one person out of the 152 sampled had an education level of four years of secondary. This suggests that education could play a big policy role in increasing agricultural productivity in Kenya. Education levels among the Kenyan population have been rising over time and more people are getting educated. In 1963, there were 891,553 pupils enroled in primary schools representing less than 50 % of the children of school age, by 1978 this figure was about 3.3 million representing about 85 % of children of school age, and by 1985 about 4.7 million pupils were enroled in primary schools. Secondary school enrolment rose from 30 thousand students in 1963 to about 350 thousand in 1978 and about 500 thousand in 1985. The percentage of secondary school age going children enroled in secondary schools rose from 39 % in 1970 to about 74 % in 1980. Apart from formal schooling adult literacy classes have also been established to increase literacy levels in the country especially in the rural areas. In 1985 a UNESCO report put the average literacy rate at 59.2 % (male average was 69.4 % and female average was 49.2 %).(Africa south of sahara: survey and reference book; 1989). Chapter 1. Introduction 5 This study is designed to investigate the contribution of education to farm productiv-ity, revenues and profits in the small farm sector in Kenya. Education is defined for the purpose of this study as the number of school standards completed by the farm operator 7 . Other educative factors that are investigated are extension service, demonstration atten-dance and baraza attendance. The role of education in farm decision making is considered as an important factor affecting the way farmers choose what to produce, how to produce it, use of other inputs and choice of product mix and allocation of other productive inputs among competing uses. The small scale farmers in Kenya are involved in production of different types of crops which are either intercropped, or grown in separate plots. Some of these crops are grown for home consumption and the surplus is sold, while some crops are grown entirely for sale. The cash crops require more purchased input out lays compared to the subsistence crops. These farmers are also involved in keeping livestock and poultry. Small scale farmers have also benefitted from government policies of expanding small scale cash crop production of tea and coffee. As farmers move from production of subsis-tence crops to production of market oriented crops they face new and more complicated technologies. These technologies include use of new inputs, gathering of market informa-tion for input and output prices. It is within this context that education plays a role in helping the farmer adjust to these new conditions. Educated farmers wil l have an upper hand over their less or non-educated counterparts in adjusting to these new conditions. It is argued that education increases the farmers process of gathering and use of infor-mation about new production technologies, markets (input and output prices) and also increases the use and allocation of other inputs among competing needs to maximize returns. Put another way the study hypothesises that educated farmers are more pro-ductive than their uneducated counterparts. Thus it is expected that those farmers with 7farm operator is the person reported as the head of the household irrespective of sex Chapter 1. Introduction 6 Table 1.1: Farm production by education and secondary occupation of farmers measures of output No. of school grades completed secondary occupation farm 0 1-4 5-10 10+ none labour other crop prod.(K.shs) 1437 1610 1976 3806 1589 989 1969 per acre 248 224 318 647 273 336 259 per cropped acre 474 419 661 1136 524 585 521 livest.prod.(Kshs) 771 911 1153 1515 775 1943 1060 per acre 133 127 186 258 133 661 140 Total farm prod.(Ks) 2208 2521 3130 5321 2364 2932 3030 per acre 381 351 504 905 406 997 399 total acreage 5.79 7.19 6.21 5.88 5.82 2.94 7.59 number of farms 670 300 421 35 1022 68 406 % of total farm products crop production 65.1 63.9 63.1 71.5 67.2 33.7 65.0 Livestock production 34.9 36.1 36.8 28.5 32.8 66.3 35.0 total 100 100 100 100 100 100 100.0 source:ILO/University of Nairobi Household Survey (1974): cited in Anker and Knowles page 234. higher education will have higher productive abilities than those with lower education: The study is aimed at determining if education can be used as a policy tool in increasing agricultural productivity. Statistics in table (1.1) show a tendency for both farm productivity and total farm production to increase with the level of education. Value of crop production per cropped acre and the value of livestock production per acre is about 40 % higher for the farmers with 5-10 years of schooling completed than for the farmers with no formal schooling. However, these data show that some education level may lead to reduced productivity (1-4 years of schooling), this supports Moock's findings for maize production in Vihiga division (Moock:1973). Higher education 10+ years of schooling have even more dramatic results. This sample had only 35 farmers with 11 or more years of schooling (Anker & Chapter 1. Introduction 7 Knowles). Overall, those farmers with non-farm secondary occupation produce less per acre than all other farmers. This means that these farmers find the non-farm activities to have higher returns and therefore devote most of their time there at the expense of their farms, put another way, the high pay-off in non-farm sector has deprived their farms managerial labour. The thesis uses the 1982 CBS8-IDS9 World Bank Survey of Rural Kenya data set. The purpose of this survey was to collect data to study the effect of the coffee boom of late 1970s on tree crop adoption in Kenya, it collected data on small farm and household characteristics from Central and Nyanza province. The two provinces cover some of the most productive and densely populated regions in Kenya. The major agricultural activities in these regions include cash crop and food crop production as well as livestock activities. As in most parts of developing countries family labour is the single most important input in the small-scale farm, however, in those areas where cash crop forms a significant part of the farm output, hired labour, fertilizer, pesticides and other chemical inputs are used, oxen and tractor services are also common in the Kenyan small-farm sector especially the former. The other common feature is the seasonality of labour demand, with peak period being in the harvesting and planting season(Wolgin 1975), this can cause a considerable strain especially in the tea and coffee growing areas. 1.2 Problem Statement Some of the major problems facing the Kenyan econonvy are, falling rural per capita incomes, unemployment and food sufficiency (Sessional paper No. 1 of 1986, KLRPU 8CBS:-Central Bureau of Statistics 9IDS:-Institute of Development Studies (University of Nairobi) Chapter 1. Introduction 8 1988, World Bank 1983). Productivity growth in agriculture and in particular the small-scale sector is looked upon as the immediate solution to these problems (Sessional paper No.l 1986). Kenya's efforts to increase farm productivity have mainly centred on input and output pricing and marketing policies, research and extension. It is not currently clear how farmer education affects farm revenues and profits in Kenya. As a result of this information gap policy makers, farmers and prospective entrants are not aware of the returns to education in farming. Previous studies on education and farm production in Kenya have mainly focused on maize (Moock 1973; Hopcraft 1974). Moock's study shows a negative effect on agricultural productivity for low levels of education 1-3 years of education and positive effect for higher levels of education (4+ years of school completed) (Moock:1973). Hopcraft estimated production function for maize, tea and livestock using the value added approach and his results show that family labour, extension visit, and years of experience10 were the most important determinants of farm profit. Years of off-farm employment had negative and significant effects on farm profits, while education had positive but insignificant effect. These results are contradictory to those of Moock (1973) in that Moock's had positive and significant coefficient for high levels of education. Hopcraft's results contradicted those of Moock in that he found education above three standards to have negative effects on maize production. These studies concentrated more on specific crop production other than overall farm enterprise and so could not effectivelly capture the whole return to farmer education in farm production. Moock's study though based on more detailed data concentrated on maize production ignoring other farm activities while acknowledging that most of these crops were intercropped, thus this study could not give the full return to education in farm production. The fact that these studies done on Kenya were based on data sets collected in small regions and 10years of experience were denned as the number of years the farm operator worked in off-farm employment Chapter 1. Introduction 9 also for a crop that is basically grown for home consumption, and the fact that the results of the two studies contradict each other justifies undertaking another study. The data set used in this study covers a much larger area and takes into consideration all the farm and off-farm activities of each sampled household. The author believes this is the best way to capture the effect of education in the farm production decision making as it allows the farmer all the available production options. I propose to employ the value added approach in measuring the effects of education on farm profits and allocative effect. My data set also includes a larger sample, 748 observations compared to Moock's 152 observations. The fact that Moock found positive effects of education using a small sample from a small locality with a small proportion of the people having education suggests that a more diversified sample could have even better results. Further more no such study has been done using the 1982 IRS511 CBS-IDS World Bank Survey of Rural Kenya. The study is designed to address the following questions; -is investing in education beneficial to those who wish to take a career in farming? -is the extension service more or less beneficial to those who have invested more in education? (do the educated benefit more from extension service compared to the less educated)? -what are the marginal returns to education for those farmers who have invested in secondary education (8 4-years of schooling)? 1.3 objectives Specifically the study is meant to estimate the returns to education in the smallholder farm sector and to measure the effects of education on small-scale farm allocative effect in Kenya. The research will be done in the following procedure. 1 1 Integrated Rural Survey 5 Chapter 1. Introduction 10 i) Review of the literature on smallholder farm sector in Kenya to evaluate it's impor-tance in the economy. ii) Discuss the main factors influencing the smallholder farm productivity in Kenya, including farmer education as a possible factor influencing farm productivity and profits. iii) Build an econometric model to measure the importance of each of the inputs and in particular education in increasing farm productivity and profits. iv) Determine the correct functional form and the relevant variables for estimating the returns to schooling and farmers efficiency. vii)To interpret this measurement of the effect of schooling on farm revenues and profits and evaluate it in the context of decisions made by farm operators and agricultural policy makers. 1.4 Outline Chapter II is a review of the theoretical framework of the thesis. In.this chapter relevant studies and findings are reviewed and differences and similarities in the approach noted. A brief review of the smallholder farm sector in Kenya is also presented in this section. In chapter III the model is presented and discussed with particular attention focused on the profit and production function approach. Chapter IV is the empirical implementation, in this section the data requirements and sources are discussed. The functional form of the model is also presented. Results are presented and discussed in chapter V. Chapter VI is the summary and conclusion and the policy implications of the study. The limitations of the study are also discussed in this chapter. Chapter 1. Introduction 11 1.5 Methodology The study begins with a review of the small farm sector in Kenya and particular attention is focused on it's importance in the Kenyan economy, as foreign currency earner, food supplier, source of employment, and a major contributor to the GDP. A review of factors affecting smallholder production is made and the problems facing the sector are also reviewed. The study is then designed to address the issue of small-farm production problem and particular attention is focused on investigating the role of schooling and information (extension service) in the small farm production. The study begins with the sample data analysis spefically designed to investigate the relation between farmer education and farm productivity, this is then followed by regression analysis. The linear functional form using the ordinary least squares procedure is employed in the regression analysis. The regression analysis began with estimation of particular farm activities; viz livestock and crop sectors were estimated separately, then these were aggregated into one farm activity and a separate model estimated. Off-farm income was also included in the farm output revenue and a different model estimated, this was meant to capture the role of education when the farmer is faced with off-farm income opportunities. The model was then developed into a profit function with the aim of capturing the role of education in farm profit optimization. The model was estimated with education categorized in dummy form for different schooling categories so that the effect of different levels of schooling on farm revenues and profits could be captured. Chapter 2 Literature Review 2.1 The Small—Holder Farmer in Kenya For the purpose of this study the small-holder is defined as in the Integrated Rural .Survey series, and includes all those farms covered under the 1982 world bank survey. The typical peasant household combines the consumption and labour supply roles generally performed by households with managerial and production roles commonly per-formed by firms. The most interesting theoretical question is to what extent decisions taken in each of these areas affect and are affected by decisions taken in other areas. The role of decision making is important because it has direct effects on the farm production. For example a farm operator has to decide what to produce, how to produce it and when. These decisions have direct effects on his labour supply and allocation, choice of quality and quantity of other inputs as well as the way he allocates these factors of production among different competing activities. For example he could determine whether to employ labour for production of certain types of crop or livestock activity, and also how total family labour should be allocated for different activities. In a country like Kenya farmers also have to make decisions on what crops to grow and in what mix. High farm risks and uncertainity in Kenya increases the value of decision making. These risks arise from irregular rainfall, uncertain market conditions, and changing production technologies. World wide fluctuations in prices of permanent cash-crops such as coffee and tea increase the risk of long term investment in these crops. 12 Chapter 2. Literature Review 13 Lack of credit facilities also limit the farmers ability to apply appropriate technology, and in particular lack of credit to acquire inputs at the right time and in convenient packages. Even where certain inputs like fertilizer and chemicals are available knowledge of how to use these inputs is also an important component of their effect on productivity. Small scale producers have been accused of improper use of inputs such as fertilizer and chemicals, especially in the coffee and tea growing areas. This misuse ranges from over application of these inputs to underapplication. Over application has been attributed to farmers lack of proper use knowledge and inability to read the application instructions, while underapplication is attributed to lack of credit to acquire enough inputs or lack of these inputs in appropriate packages. Another aspect of peasant household behaviour affected by risk is the determination of the farm household's product mix. Thus risk averse farmers may choose to produce one product over another depending on the perceived risks. Another choice is to spread these risks by growing a variety of crops instead of specializing in production of a particular crop. This practice is very common among the small-scale farmers and more so to those living in marginal areas1, as in most cases the rainy seasons are very short. The question of timing is also important if rain seasons are perceived to be short, this means farmers have to sow their crops earl}' enough before the beginning of the rainfall. Crops that take a short period to mature are most preferred if the rain season is perceived to be short. These risks affect the product mix of crops grown by the farmers and also make production for home consumption more preferred by most farmers than production for the market (Anker and Knowles, 1983). Prices affect the farmers' decisions depending on whether the crop is for domestic consumption or for export. Prices for domestically consumed crops are announced by the government prior to planting season, while export crop prices depend on the world 1areas that recieve little and unpredictable rainfall Chapter 2. Literature Review 14 market, therefore, they have more price risks. In most cases this price control is not effective unless the produce is sold to a marketing board. But if produce is sold through a different outlet, either direct to the consumer or a private middle man the price is bargainable and in most cases it is above the government set price. This motivates farmers to sale outside the controlled markets. Another factor that motivates farmers to sale outside regulated markets is the delay in payments by the marketing boards for the crop delivered. The choice of technology depends on price and production risks involved. Growing of cash crops or hybrid maize, for example means great outlays on purchased inputs like fertilizer, pesticides, chemicals, seeds and other inputs. It may also involve high investments on farm machinery and equipment, than it would if the farmer only grew subsistence crops. Thus growing of these crops exposes the farmer more to the input markets, where information on prices and inputs quality may be scarce and expensive to acquire. The same problems arise in the case of improved livestock over the traditional livestock, in that improved livestock require more outlays on input expenditures like feed and veterinary services. Most of the farmers under the current study fall in this category as production of cash crop and hybrid maize is prevalent in the two provinces from which the sample is taken. However, growing of these crops in small scale does not require a lot of expenses on farm machinery as the processing is done by farmers' co-operative society factories. But the farmers have to purchase fertilizer, pesticides and other forms of chemicals, either through a co-operative loan or from their own savings. As noted earlier in the introduction small-scale farmers are also engaged in non-farm employment at a considerable distance from home. Farms belonging to such migrant workers may suffer from reduced managerial inputs. This could lead to a lower marginal product or reduced utilization of other inputs. On the other hand, access to non-farm income could increase acquisition of capital which can be invested in the farm and hence Chapter 2. Literature Review 15 increase productivity. The population density has also been rising in the rural areas and so the average household holding size has been declining, so to maintain the levels of production or improve them it has been necessary to adopt methods that improve land productivity (world Bank p. 327). To improve land productivity requires adoption of new production technologies and this exposes the farmer to more risks. These new technologies require more purchased inputs and improved management, than the traditional forms of produc-tion. They are dynamic unlike the traditional methods in the sense that new methods are being introduced through findings from research and the farmer has to keep abreast with this new information. The above sentiments show that the Kenyan small-farmer produces under conditions in which decision making in the process of production is very important. The importance of the small scale sector is understood when you consider its role in the economy. Agriculture contributes about 40 % of GDP, and the small scale sector contributes about 75 % of the total agricultural output and provides about 85 % of the total agricultural employment. Considering the land shortage and the falling acreage, population ratio, if the country is to avoid future food shortages and economic stagnation serious steps in terms of designing policies aimed at increasing land productivity should be taken. Nowhere is the need for these policies more apparent than in the small-scale sector. The only way the government can hope to have meaningful increase in agricultural output is to design policies that increase productivity in this sector. Traditionally production has been geared towards household food sufficiency, thus commercial farming was not considered a priority. Natives were also not allowed to grow cash crops (coffee and tea) (Hazlewood 1979, Maitha 1974). These crops were mainly grown by the large scale farmers. Today modernization and increasing demands for cash income to purchase industrial goods as well as to meet other financial obligations have Chapter 2. Literature Review 16 forced small farmers to diversify their production to include cash crops (World Bank 1983). It is also along the government policy to encourage small-scale production of cash crops especially coffee and tea. The cultivation of these crops entails use of new technologies which require more outlays in purchased inputs. There is increased use of hybrid maize, improved livestock, fertilizer, pesticides and other chemicals as well as methods of erosion control. These factors indicate to what extent the small farmer in Kenya has modernized his farming methods. Although a lot of research has been carried on at various research stations and also by economists on the factors that determine agricultural productivity in Kenya, very little attention has been directed towards education as a productive factor in small-scale sector. Most studies have focused on conventional inputs like land, hired labour, capital, family labour and fertilizer (Etherington, Maitha, Wolgin), as the determining factors of agricultural production. These studies mainly estimate production functions for particular agricultural products other than aggregate production functions, although most crops are intercropped and it is difficult to discern what portion of a particular input went into production of a particular crop. For example when one tries to estimate a production function for maize alone, when in reality it is intercropped with beans2 and other crops the results will have a bias in input use, because it is not possible to determine the exact proportion of the inputs that went into production of maize alone. However, a gross production function has less bias as it considers most of the crops grown by the farmer. Going by Moock's (1973) findings and statistics in table (1.1) educated farmers appear to produce progressively more per acre than the less educated. The use of purchased inputs among the small-farmers in Kenya is also positively related to education level, . 2Moock's(1973) analysis showed that 70.4 % of farms had maize intercropped with beans, only 14.5 % had maize planted alone 5.7 % had maize beans and peanuts interplanted 5.7 % had maize beans and millet interplanted and 3 % had maize and peanuts interplanted Chapter 2. Literature Review 17 with educated farmers using progressively more purchased inputs like fertilizer, chemicals, and sprays. Small-holders grow a variety of crops and keep livestock as well. Most of these crops are interplanted, maize for example is commonly interplanted with beans and other forms of legumes, although this is basically done to reduce the risks of crop failure as well as to ensure a high yield per unit of land, it has been found to be biologically useful in maintaining soil fertility. Inter-cropping also helps in cases where length of rainfall season is not predictable, because incase the season turns out to be short the farmer can still get some yield from those crops which need a short period of rainfall. Cattle, sheep, goats, and hogs are is also kept. Cattle are mainly kept for milk production, as well as provision of other services like ploughing. Statistics in table (1.1) lend support for the argument that education has a positive influence on agricultural productivity. Kenyan farmers are exposed to a considerable amount of extension service which is offered in different ways; it could be through a visit to the farm by an extension officer, a seminar attendance by the farmer or through public baraza 3 . Other forms of extension service available are demonstration attendance, and farmer training centre seminars. Most of these extension services are for a particular crop and not agriculture in general. It is important to note that extension service is the main vehicle through which the government and other organizations pass their research findings to the farmers. It is the purpose of this study to investigate how the extension service influences farm productivit}' and if it acts as a substitute or a complementary to education. 3A kiswahili word for public gathering normally convened by local government adminstration officers to pass some important government announcements or discuss local development issues Chapter 2. Literature Review 18 2.2 Human Capital Economic theory has been extended to make room for human capital as an integral part of capital theory. A wide array of human skills are essential in fuelling the dynamics of development. These skills are useful in increasing the productivity of other inputs, as well as for managerial purposes both in the manufacturing sector as well as in the agricultural sector. The principle activities that contribute to acquisition of human capital are child care, home and work experience, schooling, and health (Schultz:1980). The value of this well-being raises the levels of labour productivity by increasing entrepreneurial ability in acquiring information and adjusting to disequilibrium inherent in the process of modern-ization. For the purpose of this study attention is focused on schooling as a contributor to development of human capital. This does not imply that the author underplays the importance of other sources of human capital like experience or health as contributing factors in agricultural production. Another factor of human capital that is considered in this study is extension service contact. Schultz has argued that schooling is more than a consumption activity, in the sense that it is not undertaken solely to obtain satisfaction or utility while attending school, on the contrary private and public costs of schooling are incurred deliberately to acquire a productive stock, embodied in human being, that provides future services in production or income earning activities. Education and health standards among the Kenyan popu-lation have increased over the years, this has been matched with increased government expenditures in these two areas. This rise in educational standards has led to growth of human capital, and hence skilled (educated) labour among the Kenyan population. This growth in human capital has led to a remarkable growth in the industrial productivity in Kenya. The main source of this human capital is formal schooling. Although schooling Chapter 2. Literature Review 19 in Kenya is seen by the parent and the student as a means to acquire skills necessary for off-farm employment, it's contribution to farm productivity cannot be underestimated. The ability of farmers in low-income countries to perceive, interpret, and respond to new events in a context of risk is an important part of the human capital of these countries. "This ability is treated as the entrepreneurial ability of these farmers" (Schultz 1980). According to the findings of Kenya Long Range Planning Unit of the 7.9 % of the per capita income growth for the period 1970/72-1980/82, 0.7 % came from increase in human capital per capita growth. These results show that the growth in human capital in Kenya which has resulted from increased education has led to positive effects on the Kenyan economy. The gains from investing in human capital which is primarily schooling depend on the environment and opportunities available to the investor in the region in which the investment is made. These gains will be higher if the environment is modernizing but lower, negative or zero in traditional environment. In the case of the Kenyan small-farmer we have seen indications that there is increased use of modern methods of production which exposes the farmer to new technologies and hence disequilibria, thus investment in schooling for these farmers is expected to have positive returns. 2.3 Education and Farm Productivity The relationship between education and agricultural productivity has been a subject of study by many agricultural economists in both developed and developing countries. In most of these studies education has been found to have positive effects on agricultural output (Barichello, Cunningham-Dunlop, Moock, Pudasaini). Other researchers have estimated the effect of education on farm production in different ways, Khaldi (1975) and Fane (1972) use simillar approach, Khaldi estimates the effect of education on optimizing Chapter 2. Literature Review 20 the use of purchased farm inputs, while Fane uses the cost approach by estimating the effect of education on farm cost minimization. Petzel (1976) uses a different approach by estimating the effect of education on farm acreage adjustments, while Huffman estimates the effect of education on off-farm labour supply decisions. In all the above studies education is found to have the expected effects. However, Shultz argues that these effects depend on whether the environment is modernising or traditional, a theory later confirmed by Jamison et. al (1980) by reviewing several research papers which have been written on..this kind of research. They established that education has a negative or zero effect on agriculture if the product being studied is produced traditionally. However, in the modernizing sector the effects are positive. Educational effects on agricultural productivity is in the most a measure of alloca-tive efficiency under disequilibria conditions. This disequilibria is assumed to arise from changing technology in the agricultural production, uncertain input and output prices, availability of different kinds of inputs at different prices and uncertain weather condi-tions. Research works which are constantly being undertaken generate new technologies and farmers have to keep abreast with these developments and so need for new in-formation increases and this leads to increase in the value of education. Problems of determining product mix and allocation of time between farm and off-farm employment exists among the small farm operators. Thus faced with these situations the farmer has to take decision on what to do and how to do it. It is within this context that education plays a role. In the industrial sector the importance of education can be seen by looking at the rewards in form of high wages paid to the highly educated employees, they also stand higher chances of job promotion (Gintis,H 1971). Thus wage levels have been seen as depending on the level of education, this indicates the extent which employers associate education with productivity. On average higher education is seen as an indicator of level Chapter 2. Literature Review 21 of productivity of an individual. The importance of education in promoting human value is also seen in numerous research findings that have been made possible due to education. However, in the agricultural sector to understand the effect of education on farm pro-ductivity one needs to follow Welch (1970). Welch has argued that education contributes to farm production in two ways, the "worker effect" that is effect of education on ordinary farm labour derived from given resources including information (or simply the marginal effect other variables held constant) and "allocative effect" that is effect of education on entrepreneurial or management decision making due to information processing, learning, and disequilibrium adjustment (depends largely on the dynamic response to technical or market circumstances). He argues that the worker effect is an underestimation of educa-tion effect on farm productivity, it measures the effect of education on farm production holding other factors constant, thus it is a measure of marginal effects. A measure that considers both worker and allocative effects, is a better estimate of education effect on farm productivity as it takes into account the contribution of education in allocation of other inputs to farm productivity in a dynamic environment where information is changing, scarce, and costly to obtain. If education is treated like any other input then it's contribution to farm output will be measured by the marginal effects that is the worker effect. In this case given a production function of the form; Q = g(X,E) where Q is a measure of output, X is the level of inputs used in production of this output and E is a measure of the level of education. The contribution of education (E) to production will be given by the partial derivative of the function as below 8Q/6E =8q(X,E)/6E. This gives the marginal product of education which is the "work effect", but does not include allocative effect as the farmer is dealing with only one product and so questions Chapter 2. Literature Review 22 of allocation do not arise. Research has shown that education does not just affect farm production like any other conventional input, but its effect goes further by influencing the level of use and choice of other inputs (Barichello, Cunningham-Dunlop, Huffman, Shultz, Welch). So in this case a measure of the marginal effect will be an underestimation of the contribution of education to production. It also improves the quality of human beings in terms of increasing their skills and therefore making them more productive as labour units and as users of other factors in the production process. Thus the productivity of other factor inputs is higher if it is combined with highly educated labour force other than less educated labour force. Khaldi (1975) has argued that in a dynamic agriculture, research activities, both basic and applied-augment the stock scientific knowledge and yield new and improved inputs and as producers face a less than perfect familiarity with the new things, they at any point in time will be making production errors. Hence, productive research activities create and perpetuate disequilibrium in farming. Education improves the decision making process and thereby influence the level and/or the composition of other inputs, and in a country where mixed cropping is common it can influence the choice of crops to grow or the crops to intercrop. Under these conditions an estimate of the work effect of education represents an underestimation of the total effect of education on farm production. Thus, in an environment where new inputs are appearing and relative prices are changing, the presence of disequilibria yields excessive costs and creates an incentive for decision makers to learn and adjust their activities. The extent to which education may enhance farmers' ability to respond to these new growth opportunities implies that, for this input, there does exist an allocative effect in agricultural production. Pudasiani,P.S. (1975) on the effects of education in agriculture in Nepal found edu-cation to have a higher pay-off to productivity in a modernizing environment than in Chapter 2. Literature Review 23 a traditional one. His work is based on Welch's theory of differentiating the worker ef-fect from allocative and input selection effect. His results show that education is more effective in a modernizing compared to a traditional sector, while at the same time con-firming Welch's theory. He found that rate of productivity generally declined with a rise in education in both regions, suggesting that diminishing marginal productivity applied even to education. Huffman (1973) on a study on to determine the role of education on decision making on adjustment of midwestern U.S. farmers to the changing optimum quantity of nitrogen fertilizer in corn production, found rate of adjustment to be positively related to education of the farmers, availability of information (agricultural extension) and scale incentive to be informed (acres of corn). His results like those of Moock (1973) and Pudasaini (1983) show education and extension as substitute sources of allocative efficiency. When the productive arts remain virtually constant over many years, farm people know from long experience what their own effort can get out of the land and equipment. In allocating the resources at their disposal, in choosing a combination of crops, in deciding on how and when to cultivate, plant, water, and harvest and what combination of tools to use with draft animals and simple field equipment-these choices and decisions all embody a fine regard for marginal costs and returns. These people also know from experience the value of their household production possibilities; in allocating their own time along with material goods within the domain of the household, they too are finely attuned to marginal costs and returns. In short they are efficient and there are no gains to be had from investing in education. On the other hand in modernizing environment farmers deal with a sequence of changes in economic conditions, which are in general not of their own making because they originate mainly out of the activities of people other than farm people. These changes are endogenous and they originate predominantly from the useful contributions Chapter 2. Literature Review 24 that flow from organized agricultural research and from improvements in the inputs that farm people purchase and use in agricultural and household production. The demand for the ability to deal with the new and better production possibilities is in large part de-termined by organized agricultural research and non-farm firms that produce the inputs that farm people purchase. Furthermore, it takes time to reallocate resources in arriving at a new equilibrium. Moreover, additional changes occur even before the reallocation called for by the preceding change has been completed. Hence, the implication is that full efficiency is kept beyond the reach of farm people. Most of the Kenyan small farmers under the current study fall under this category. We have seen that these farmers produce under uncertain conditions, are faced with new technologies as they move from production of subsistence crops to commercial crops which require more outlays in form of purchased inputs. Growing of these crops and especially cash crops requires technologies different from those used in production of subsistence crops. In 1973 Moock estimated the effects of education on farmer productivity in Kenya using sample data of 152 households in Vihiga division in western Kenya on hybrid maize production. His findings show that a certain level of education 1-3 years of schooling led to negative effects on farm productivity, while education level above four years of schooling had positive effects on production. Only one farmer had completed four years of secondary school (11 years of schooling) and 32 % of the sample had no education. He also included an extension service variable, which had a positive and significant effect on maize production. His results show that a 10 % increase in extension service was associated with a 0.2 % increase in maize yield ceteris paribus. Farmers with Four or more years of schooling produced 2 % more maize per acre, while farmers with 1-3 years of schooling produced 11 % less maize than those with no schooling. Independent effect of schooling was Chapter 2. Literature Review 25 strengthened when he allowed extension to interact with schooling in the production function, while the coefficient of the interaction of education and extension service was negative showing that they acted as substitute sources of technical knowledge. Hopcraft's (1974) study whose major concentration was on maize production suggest that labour is less important than land as an input in maize production. Purchased inputs were found to be statistically significant especially in hybrid maize production. However, his study show that formal schooling above three years had consistently negative effects on maize production. This relationship was confined within cultivation of local varieties, which means the effects on the hybrid variety could be positive. For the livestock production function Hopcraft found that only land and livestock capital were significant inputs in livestock production function. Formal schooling and labour had the same effects as in the maize production function. For the tea production function the schooling coefficient was still negative and significant. For aggregate function the coefficient was negative and insignificant. These results for Kenyan study give contradicting results and further study using a different data with a wide variety of crops and covering a wide area is found desirable. This allows education to play an important role in determing the product mix, input allocation among competing uses, input selection effects4 and then specifying a value added function so that we can capture the "full" effect of schooling on farmer profits. There are potential gains to be had by considering all the crops the farmer produce so that we can capture his product mix choice and allocative abilities. By considering all the farm activities of the farmer we remove the bias of underestimating the return to education in farm profits. Between the time Moock and Hopcraft's study were done and the time the data to be used in this study was collected is a period of about 10 years 4allowing more crops expands the farmers horizon of input choice for the various crops. Inclusion of the livestock in the function is expected to increase this impact. The plausible expectation is that the education coefficient will increase in size as more alternatives are added in the function Chapter 2. Literature Review 26 and within this period the levels of education in Kenya have risen significantly, and it is expected that this data set has a higher level of educated farmers. This implies the sample has a higher proportion of educated farmers more appropriate to be used for estimation of return to education in farm production. Chapter 3 Model Specification. Having highlighted the Kenyan small-farm sector production characteristics and prob-lems, we now turn to the model specification that relates levels of output and profit to the particular physical and non-conventional inputs used in the production process. The foundation of this study is based on the neo-classical production theory and the human capital theory. The neo-classical theory is used to give the basis of the production re-lationship between physical inputs and levels of output, while the human capital theory is incorporated to determine the economic value of human abilities generated through schooling in the production process. To this end two approaches will be used; a sales function and a restricted profit function. Sales function is defined as the total value of farm products (crops and livestock inclusive) as recorded in the survey sample. The sam-ple recorded production information for 27 different crops. For the livestock and poultry sector we have sales on eggs, milk, hides and skins plus sales of livestock (sheep, goats, pigs, cattle, other). A modefied definition of the sales function is arrived at by including off-farm income to the total farm sales, this gives the total family income and this is estimated as function of the levels of conventional and non-conventional inputs. The neo-classical economic theory attributes the contribution of a factor in the pro-duction process to its marginal value in the use of production of that product. However, for education it is not clear that the direct contribution to physical production accounts for the total contribution to revenue. The contribution of education to total revenue is more than the marginal product as normally defined; that is, the ability to enable the 27 Chapter 3. Model Specification. 28 worker to accomplish more with the resources at hand, the worker effect, which gives a measure of increase in output per unit change in education holding other factor quanti-ties constant. Increased education may enhance worker's ability to acquire and decode information about costs and productive characteristics of other inputs and also how to chose his product mix. As such, a change in education results in a change in other inputs including, perhaps, the use of some new factors that otherwise would not be used, it could also lead to a change in the product mix. So a model is needed that lets education influence allocation and choice of other inputs, farm operator time allocation, as well as choice of product mix. We start with a simple Value added function expressed as the difference between the gross value of farm output minus the cost of purchased variable inputs Qi=pq{Xi,Z,E)-VxXi (3.1.1) p is commodity price, q is the physical product, and is a function of purchased inputs, X, farm operator education (E) and inputs supplied by the farmer, Z. The price of X is px, and both px and p are assumed exogenous to the producer. Maximization of Q with respect to X gives; 6Q/8X =p(6q/5x)-Px = 0 (3.1.2) which is the marginal productivity theory, that is, in equilibrium the value of the marginal product of X should equal its price. This assumes perfect knowledge and so there are no gains from investing in education. But, again assume that farmers are not equally adept at assessing productivity and that the quantity of X purchased is a function of education. In this case, the marginal product of education is, SQ/5E = (p5q/Sx -Px)dX/dE (3.1.3) To illustrate how the total effect of education on production can be measured using a value added function Welch uses an example based on two products, qr and q2, where each is a function of three inputs: 1): education, E; 2):other supplied inputs by the farm, Chapter 3. Model Specification. 29 Z; 3): purchased inputs, X. If we assume a value added function specified as below;1 Q = P i 9 i ( z i , 2 i , £ i ) +P2q2{^2,Z2,E2) -p2X (3.1.4) where E - El + E2; Z° - zx + z2\ X = xl + x2 (3.1.5) and . E i , z i , xa are the levels of the three inputs used in production of good one, the ones used for good 2 are subscribed with a 2. Totally differentiating 3.1.5 with respect to education (E) gives, 1 = dE1/dE+dE2/dE; 0 ='Azx\'dE+dz2/'dE; and dX/dE = dXl/dE+dx2/dE (3 Finally assuming value added is a function of the total quantities of education and farm supplied inputs, Q = f(E,Z°) and partially differentiating 3.1.4 with respect to education and substituting in 3.1.6 gives us the marginal product of education as, Sf/SE = p28q2/8E +{Pl8gi/8E-P28q2/8E)dE1/dE +\p18ql/8z-p28q2/8z)dzJdE + (Pi8qi/8x — p28q2/8x)dxi/dE + (p26q2/8x — px)dX/dE (3.1.7) The first term above is the own value of marginal product of education, the worker effect, the next three terms refer to the gains from allocating the respective factors, education, supplied inputs, and purchased inputs, efficiency between competing uses. The last term refers to the allocative gain from selecting the right quantity of purchased inputs, X. If this were a sales function dX/dE will equal zero and the last term will be lost. A model specification of this nature is fitting for the current study, noting that we are dealing in situation where the farmer has several crops to grow either separately or intercropped, he also has to allocate his labour time and that of the hired labour among competing farm and non-farm activities. The purchased inputs can also be allocated in different proportions in production of different farm output. For these reasons most of 1this example is taken from Welch(1970) in Education and production p.45, it also used by Pudasaini in Effects of Education in Agriculture (1983) p.509-515 Chapter 3. Model Specification. 30 the emphasis is placed on this method of estimation in this study. 3.1 The Sales function approach In the sales function gross value of agricultural output is expresed as a function of con-ventional and non-conventional variables. Gross value of farm output is given a loose definition due to several factors. One, not all the farm output is sold, this means we need to value the farm output that is consumed within the household. However, for the maize it is not easy to compute the total farm output since farmers start consuming the green maize while in the farm without keeping records. So the only maize output records that were available were those of dry maize that was harvested. This means the maize value is an underestimation of the total maize produce value. Also data is not available for the domestically consumed milk, which gives another source of farm output underestimation. Most of the farm produce waste is used as feed for livesctock or left to decay in the farm to provide manure. The difference between the sales function and the restricted profit function is that the later assumes profit maximization. Another source of the difference is, where as physical levels of variable inputs are included as explanatory variables as in the sales function, only prices of these inputs are included as right hand variables in a profit function, consequently dE/dX equals zero in the sales function and so does not capture the input selection effect. Using sales or the profit function allows us to aggregate across different products so that the data contains all the possible alternatives open to the farmer, this is desirable if gains to management decision making from choosing an efficient output mix are to be captured. On the other hand, aggregation across different commodities leads to model mis-specation unless it is assumed that all individual production functions are identical and linear (Griliches 1957). However, this is not a major problem if the Chapter 3. Model Specification. 31 farms under study are not very different, in terms of their production technologies (De Boer and Chandra 1978). As most of the farms under study grow almost homogeneous crops this problem is not expected to be a major one. Neo-classical theory assumes perfect knowledge in the production process and this imposses a costly restriction in the current study, so we need to assume that knowledge is scarce and costly to acquire. This assumption is necessary if we are to estimate the role of education in the production process properly. In using this method to estimate the contribution of education to agricultural productivity, the value of purchased and family owned variable inputs, capital service flow, education of farm operator or education of the spouse, hired and family labour are considered as independent variables. The advantage of this specification over the engineering function is that the engi-neering function merely gives the technical relationship between production of product Q with particular inputs without taking into consideration allocation effects. Thus the marginal product of education in an engineering function will only capture the worker ef-fect. However, by using the sales production function we can do better than just estimate the technical relationship of output and the inputs. Sales function allows aggregation over several products. This step increases our ability to measure the contribution of education to farm productivity, in that it allows the education coefficient to pick up any returns to the decision maker from re-allocating his array of inputs among the products, and selecting his product mix. We begin with a simple technical production^  function, and assume that the farmer produces crop Q with a set of variable inputs X and a set of fixed inputs Z and has an education level E. We assume that education enters the function explicitly like any other factor. In this case the technical production relationship between physical output Q and X, Z, and E can be expressed as below, Q = q(X,Z,E) (3.2.1) Chapter 3. Model Specification. 32 In this case the marginal product of education is 5qjSe and refers only to worker effect or the ability to accomplish more (physical output), given the resources at hand. In this case there is no room for allocative ability, since questions of allocation do not arise as he is growing only one crop and no other alternatives are entertained in the function. Now let us assume the farmer grows two crops and q2 and let us assume these are functions of the input vector X. And let us assume X is given but its allocation among competing uses Xi and x2 is not. X\ and x2 are proportions of X used in the production of good qi and q2, respectively, such that the sales production function is expressed as below. Q=Piqi(x1,E,Z)+p2q2(x2lE,Z) (3.2.2) Here technical efficiency refers to being on the product transformation frontier, that is, of maximizing qx, given q2 and X, and does not correspond to maximization of sales, Q, given X. Where pa and p2 refer to the prices of the two commodities qi and q2 respectively and are assumed exogenous to the farmer. To maximize Q, we have, 6Q/Sxi = pi8qi/8xi — p28q2/Sx2 = 0... (3.2.3) as the first order conditions. Maximization of sales requires technical efficiency and that the marginal value product of X be equated between its competing uses, further, the marginal values of X\ and x2 should equal their respective prices. Now suppose we assume that productive capacities of factors are not equally understood by all farmers and that allocation of X among alternative uses is a function of education, that is, Xi = X\(E). The sales function is, Q = PiqiME), Z) + p2q2{x2, E,Z) (3.2.4) Then in this case the marginal product of education is, 6Q/6E = (pifyi/fo! - p28q2j8x2)dxxjAE (3.2.5) and is positive if education enhances allocative ability. Thus in gross value produc-tion function if the allocation of inputs among alternative is not an explicit part of the Chapter 3. Model Specification. 33 function, and education is treated as a factor in the function, we have the inference that the marginal product of education includes gains in allocative efficiency as well as the worker effect. It is expected that farmers with more education will have better access to information and are better acquainted to the use of available information on production technology and markets. Thus farmers with higher educationn are expected to produce more com-pared to their counterparts who are less educated. This is to be reflected by positive and significant coefficient of education in the sales function. For the sales function approach different equations will be estimated. In particular, a crop equation, livestock equation, and a total sales function will be estimated. The dif-ferent equations are estimated to capture the effect of education on different agricultural enterprises. This is based on the assumption that education has different effect on crop and .livestock sector. Different specifications of the following sales function will be estimated; Qi = f(Fml, Hrl, Ld,mc,mt, Ed 1, Ed2, Lv, Ext,piec,piel,piea )..(3.2.6) Qi> the gross sales plus value of domestically consumed farm output2. Fml:-family labour hours/mandays per year spend on the farm Hrl: hired labour mandays Ld: acres of land Mc: machinery services (includes value of farm machinery, equipment, value of build-ings and building improvements over 1975-1982 period) Mt: mature tree crops Edl: education of the farm operator Ed2: education of farm operator's spouse 2the defination of Qi will depend on whether a livestock, a crop, a gross sales or gross farm income model function is being estimated Chapter 3. Model Specification. 34 Ext: extension (includes all form of extension contact) piec: value of crop related purchased inputs piel: value of livestock purchased inputs Lv: Value of livestock (Cattle,sheep, goats, pigs, poultry) piea: value of total farm purchased inputs (piel-fpiec) 3.2 Value Added Approach Value added (restricted profit) is defined as value of farm output less value of purchased inputs, with operator education, family labour, and farm capital considered as fixed inputs. The importance of classifying the fixed is not in the fixed per se, but rather in the flexibility of allowing the first order conditions not to necessarily apply (Barichello 1979). This is the firm's thoery of profit maximization and it allows the marginal product of an input to differ from its price. The essential difference between the restricted profit function and the unrestricted profit function is that the later assumes perfect knowledge. This means the first order conditions hold and that the marginal product of the inputs is equated to their respective prices. But we need to estimate a function that assumes profit maxization and at the same time does not impose the neo-classical assumption of perfect knowledge. This is necessary inorder to allow education to play a decision role in the production. We expect the coefficient of education to become correspondingly bigger as we move from the sales function to a profit function. The other step which we need to take inorder to use this approach is to aggregate capital stock into one fixed variable. This reduces the number of variables in the function and at the same time allows education coefficient to pick up any returns to the decision maker from re-allocating among these capital components (Barichello 1979). The edu-cation coefficient is estimated holding constant only the level of total capital, allowing Chapter 3. Model Specification. 35 the individual components of capital to vary. The education coefficient is expected to be larger in the equation with the highest aggregation of capital. Another step we have to take is to let the level of purchased inputs used X, be a function of education. The level of the variable inputs is no longer included as a right hand variable as in the sales function. So in the value added function we exclude the level of purchased inputs and include an array of their prices as right hand variables with the level of fixed capital and family supplied inputs. Welch (1970) has stated that "if value-added function, based on multiple products, is estimated which specifies the quantity of supplied inputs Z, but does not specify allocation among competing uses, and if purchased inputs are omitted, the marginal product of education will contain the worker effect, selection of the quantity of other inputs, and the allocation of these inputs among their competing uses." In this case you get closer to estimating the full effect of education on productivity, this should be reflected by the coefficient of education being larger in the restricted profit function than in the sales function. We begin with the familiar concept of production function, and let the function be defined as, Y{ = f{X,Z, ) (3.3.1) where Yi is the level of output of product i, whose price is given as pi, then the total revenue is given as the sum of Y\pi. Z is the level of fixed capital inputs (farm machinery and equipment, land, buildings, livestock,tree crops), including family labour. Growing of tea and coffee in Kenya is done through government licensing and so for these reasons these crops will be treated exogenously. If the level of purchased non-labour variable inputs is given as Xi and the price of the variable inputs is given as qi, then the restricted profit function is given as, II = YiPi - Xiqi = Pif(X,Z) - Xiqi (3.3.2) Assuming that farmers maximize profits then the first order conditions (marginal Chapter 3: Model Specification. 36 productivity) for the variable inputs are Sf(X,Z)/SX =qi (3.3.3) by solving these first order conditions we get the optimal quantities of the variable inputs A'", as a function of the (normalized) variable input prices and the fixed input quantities, and obtain the following equation; X' = X(qi,Zj) (3.3.4) substituting 3.3.4 into 3.3.2 we obtain the variable profit function, where profit (vari-able) depends on quantities of fixed inputs (Z), education (E) and prices of variable inputs (qi). n = n ( g i , z i , J B) :.(3.3.5) This function is then estimated with the following arguments included as right hand (independent) variables, the level of education of farm operator and that of the spouse in number of school standards completed. Other variables are capital service flow (livestock, tree crops, farm machinery, land, buildings), the price of variable inputs including that of hired labour. Also included are a number of extension variables. The estimated profit function will take the following general form, II = U(Px,ld,mc,Lv,Ed 1, Ed2, Ext, Mt, Fml, Rd )...(3.3.6) where, Px: array of input prices K : capital (land, buildings, machinery ) Lv: value livestock (including poultry) Edl: farm operator education in standards Mt: number of mature tree crops Ed2: education level of spouse in standards completed Ext: extension contact Variable (baraza, demonstration attendance, extension visit) Fml: Family labour mandays Chapter 3. Model Specification. RD: regional dummy variable (0=Nyanza; l=Central) Chapter 4 Empirical Implementation 4.1 Data Sources In chapter three the model specification was presented and it is fitting at this point to define the data for the variables to be used in implementing the model and their sources. The data used for this thesis is from the 1982 IDS-CBS World Bank Survey of Rural Kenya (IRS5). The survey collected data on farm and household characteristics from Central and Nyanza provinces of Kenya. This survey is the fifth of a series of National Integrated Sample Survey Programme, known as Integrated Rural Survey (IRS5). These surveys provide a broad spectrum of the social-economic factors dominating the small-scale agricultural households with holdings of less than eight hectares. This is the first time a thorough and well collected data set on rural small farm sector in Kenya is available. The availability of this data set makes it possible to estimate a variety of production functions for the small farm sector. The following definitions are used in the survey; household:-comprises a person or group of persons generally bound by ties of kinship who normally reside together under a single roof or under several roofs within a single compound and who share the community of life in that they are answerable to the same head and share a source of food. holding:-is defined as all land and.livestock used partially or completely for agricul-tural purposes and operated by a single holder. It is assumed that a household and holder 38 Chapter 4. Empirical Implementation 39 are the same person. holder:-is a person or persons who have the control and ability to make decisions relevant to the agricultural activities on the holding. The two provinces have favourable climatic conditions compared to most of the coun-try and have some of the most densely populated regions in the country. Central province has a land area of about 1,318 thousand hectares of which 909 thousand are classified as high potential, 15 thousand as medium potential, 41 thousand as low potential and 353 thousand classified as other. Nyanza province has a land acreage of about 1,252 thou-sand, of which 1,218 thousand is classified as high potential and 34 thousand as medium potential. These classifications are based on the annual amount rainfall recorded as fol-lows: high potential 857.5mm or more, medium potential between 735mm and 857.5mm and low potential 612.5mm.(statistical abstract. 1980). Agricultural activities in these regions include cash-crop and food crop cultivation as well as raising of livestock (dairy cattle, goats, sheep, poultry, pigs etc.). Except for poultry none of the other forms of livestock classified will be kept by a majority of the farmers sampled. For example very few people keep pigs. Cattle are kept mainly for milk production most of which is consumed at home. However, not all the farmers keep cattle, and among those who do, the majority will have only one or two head of cattle. In central province most farmers keep improved type of cattle, while in nyanza province most people keep the traditional type of cattle. The main cash crop grown in these regions include tea, coffee, sugar cane (Nyanza only), pyrethrum, and cotton (Nyanza only). Other crops that are grown basically for domestic consumption include maize (both hybrid and local), beans, millet, peas, veg-etables of different kinds. The surplus of these products is normally sold in the local markets or to respective marketing boards set up by the government. Chapter 4. Empirical Implementation 40 Like most rural areas in developing countries, family labour is the single most impor-tant input in the family farm. Children contribute a lot in labour supply to the farms, and so have to allocate labour time between the family farm and school. The other family members may also be involved in other non-farm activities, like off-farm employ-ment or working in communal activities. Off-farm employment may involve migration to urban areas or may be obtained locally. A considerable number of farmers in Kenya also operate non-farm businesses, like owning small grocery stores, a bar or running a matatu1 business. In this case they have to allocate their time between the farm and other non-farm activities. 4.2 Definition of Variables The following variables have been identified as useful in estimating the sales and the profit functions. The choice of the variables is based on economic theory, earlier research work and my understanding of the factors that influence farm production in Kenya. These variables are also chosen on the basis of the hypothesis being tested in this study, namely to investigate the role of education in small farm production in Kenya. Thus for this reason, emphasis is placed on this variable. Capital (K) To estimate this variable certain resources that are in the region under study that qualify as capital inputs had to be considerd. These include livestock, farm machinery, farm equipment, buildings, tree crops and land. Machinery and equipment owned by farmers range from oxen drawn ploughs, wheel barrows, handtills, spades, bicycles, to tractors, but the tractors are owned by very few farmers2. *a Kiswahili colloqual word for a mini bus 2in most cases farmers who own tractors have relatively large farms as it is uneconomical to farm a small farm with a tractor. It is common practice for farmers who do not own tractors to hire tractor service from those who own one Chapter 4. Empirical Implementation 41 Since these inputs are fixed the best estimate will be their service flow rather than their aggregate value as they are not used completely in one particular year. One of the problems with this variable is inability to get an appropriate valuation of tree crops. The only information we have for this input is the number of mature and immature trees owned by each sampled farmer. Trees here include coffee and tea. This means we cannot aggregate the tree crops with the other capital stocks/so we have to measure the tree crops separately from other capital inputs. The capital stock variable is defined as the aggregate of the service flow of farm ma-chinery, buildings, land and livestock, these values are weighted by respective discount weights. The survey asked farmers for the current value of their machinery farm equip-ment and buildings purchased between 1975 to 1982 and so we do not have the value of machinery and buildings acquired before 1975. This will understate the value of the capital variable, but this is not expected to be serious because of the high depreciation of the simple typical building and farm equipment and machinery. This assumption cannot be levied on some buildings which are biult of stone and have a longer life on average, but these are owned by a small cross section of rural farmers. The service flow of machinery and buildings is calculated at 15%3. Land is either used in physical terms (as number of acres) or in value terms. Tree Crops (MT) The tree crops for the purpose of this study is defined as the number of mature crop trees reported for each individual family farm in the survey. These trees include coffee trees and tea bushes. This variable is considered exogenous as tea and coffee growing is practised under government regulation, and farmers cannot alter the number of trees in their farms without government permision. 3this figure is rather conservative, assuming interest rates of about 30% and 15% depreciation, the reason for the high interest rates is due to rural informal money markets which charge on average 40% to 60% nominal interest. The inflation rate in 1980 -1984 averaged about 20% Chapter 4. Empirical Implementation 42 Family Labour (FML) There are two ways in which this variable could be used in the two functions, either in man-hours or mandays per year for family members who resided and worked on the farm for a period of 12 months prior to the survey. This information is directly recorded in the survey. This variable is considered to be very important as in most developing countries family labour is the most important input in the family farm. The total supply of family labour maybe considered fixed, but can be varied by reallocating labour time from other activities to the farm activity when need arises. The operator farm labour (HFL) is also specified separately from the famil)- labour as it is considered that the number of days the he/she spends on the farm will have a direct effect on his managerial performance. Hired labour (HRL/W/C1) Hired labour is measured in mandays per year and the wage rate as Kenya shillings per day. In the profit function the wage rate for hired labour is used as a right hand variable. Total cost of labour (CL) is calculated by multiplying the wage rate (W) and mandays of hired labour (L). Expendtiture on labour is used in the variable profit function instead of hired labour in mandays, this is necessary if differences in hired labour quality is suspected. Total mandays of hired labour is used as independent variable in the revenue functions. Hired labour is mainly used during the peak season, especially planting and harvesting season, this means a high marginal value of labour during these two seasons compared to the off peak season. A better estimate for this variable would have been to estimate by season. But seasonal hired labour data was not available. The available data was that of hired labour for the whole year prior to the survey. Education(EDl/ED2) Education variable is defined as the total number of school standards completed by the head of the household (Edl) or the spouse (Ed2). At the time the survey was conducted there were seven years of primary education, four years of secondary education, two years Chapter 4. Empirical Implementation 43 of high school and at least three years of university education. Previously there was an eight year primary education system which was dropped in the 1950s, and some of the people interviewed had gone through this old system. This measure has a bias in that in Kenya school class repetition is common and the survey did not seek information as to whether certain people had repeated classes. This study will therefore assume that respondents only spend one year in each class. The other bias expected is the inability to judge the material content of the education received, by different individuals, but this is not considered a serious problem as there is a universal syllabus for the entire education system in Kenya. As an alternative measure, schooling was specified in dummy form for various educa-tion levels; four dummy classifications were used viz; dl = l for those with eight or more standards of schooling and zero otherwise, d2=l for those with four to seven standards of schooling and zero otherwise, d3=l for those with one to three standards of schooling and zero otherwise and lastly d4=l for those who have had schooling and zero otherwise. The purpose of these classifications was to estimate any differential effect of different schooling levels on farm revenues and profits. Extension contact Extension contact in Kenya is of different forms, it could be through a visit to the family farm by an agricultural extension officer or the farmer might have attended a public gathering (baraza) where an extension officer was available to educate them, or he could have attended a farmers seminar. Demonstration is also used as a form of extension service in Kenya. Another form of extension service used is farmer's attendance of a field daj'4. Only six respondents out of the whole sample of 748, reported to have attended a field day. This number is considered to be a very small proportion of the 4field-day entails a farmer's visit to a demostration farm but does not necessarily mean the farmers will be shown practically the growing methods used in the farm Chapter 4. Empirical Implementation 44 whole sample but nevertheless it is used in the regression. The extension service is measured taking into consideration these classifications. For the purpose of this study three categories of extension service are investigated; baraza attendance, demonstration attendance, and extension service visit. The choice of these extension variables was based on data availability. It is not clear when the farmers received these service, thus the expectation of their effect on farm production is uncertain. For example if the farmer received extension service when his crop was almost ready for harvest there is very little he can do to alter the output, on the other hand if he received extension service several years ago it is usefulness may have become obsolete and no longer useful. All the 44 respondents who reported to have attended baraza were from central province. Demonstration was attended by 57 respondents, while 226 respondents had extension contact. Interaction of education and extension contact (Edext) It is not clear on a priori basis whether education and extension contact are substitutes or complements. If schooling increases the ability to utilize knowledge obtained through extension service the two will act as complements. In most previous studies extension contact has been found to act as a substitute to education. The schooling extension contact interaction variable is measured as a multiple of number of school standards completed by the farm operator and the frequency of extension contact. ., Fertilizer (fert) The value of fertilizer used for each crop was recorded in the survey and this value is used as a right hand variable in the sales function. For the restricted profit function this value is used in calculating the variable profits. Neither the price nor the quantity of fertilizer was recorded in the survey so the price of this input could not be used in the restricted function as a right hand variable. Pesticides and Sprays (PS) Total expenditure on pesticides and sprays used for each crop was recorded in the Chapter 4. Empirical Implementation 45 survey. Thus in the sales function this value is used as a right hand variable. For the restricted profit function this variable is used in calculating the variable profit. The price of this input was not recorded so it could not be used in the variable profit function. Seed In the survey the value of seed both purchased (Psd) and household owned (Osv) was recorded for each particular crop that was grown in the family farm. However, neither the price nor the quantity of the seed was recored. As in the above cases this variable could not be included in the variable profit function. Livestock (Lv) Livestock in this study is classified as cattle, sheep, goats, pigs and poultry. These are reported at their current value (1982 Kenya shillings). For the cattle we have two different classification of cattle; improved 0 and the traditional cattle. The improved livestock has a higher value compared to the traditional cattle. In small scale farming it is normally kept for milk production. On the other hand, oxen 6 are used as draft power for ploughing. In consideration of the above factors this" variable was entered in value terms other than in physical terms so that aggregation could be possible. Feed This variable is a measure of total expenditure on feed in Kenya shillings. The quality of feed used was not specified in the survey, so it is assumed that the expenditure on the feed will give a clear representation of both quality and quantity of the feed used by each farmer. Sloth This variable is a measure of total expenditures in Kenya shillings on veterinary, dipping and other livestock costs. 5 cattle with foreign ancestry Gmale castrated cattle Chapter 4. Empirical Implementation 46 Off-farm Income (OFI) This variable is a measure of all the income earned by the farm operator from other activities other than none farm activity for the last 12 months prior to the survey. It is used to compute a new dependent variable. Two equations emerge from this variable, one using the off-farm income as the only dependent variable and another using the aggregate of farm output revenue plus off-farm income as dependent variable. It is calculated as the sum of profits from own non-farm businesses, salaries and wages earned from non-farm work and income earned from working on other people's farms. Research In Kenya agricultural research funding is financed by different bodies and correct data for this variable is simply not available. Research is funded by the government through the ministry of agriculture, ministry of livestock development, government agencies, and other international agencies and governments. The other reason for not including this variable is that, if it is included it will enter at the same value for all the observations and this will not make sense for regression analysis. Also the data set used is cross- sectional data and not time time series. Regional Dummy (Rd) Since the data used in this survey was recorded from two distinct regions it is con-sidered appropriate to estimate regional equations for comparison purposes. For this purpose a dummy variable will be included. It will be introduced as l=central; and 0=Nyanza. 4.3 Functional Form Both the linear and the Cobb-Douglas functional forms were tried. However, the linear function had a better fit and good coefficients for the crop, aggregate production and the Chapter 4. Empirical Implementation 47 value added functions. The Cobb-Douglas function was superior to the linear function in the equation for livestock farms. The linear function was used mostly in the estimation of the crop and aggregate farm equations. This step was necessary in order to take into account those farms which had zero values for certain inputs. Linear functional form had a better fit for the value added and the profit functions compared to the Cobb-Douglas functional form and it was used in the estimation of all the value added and profit functions. Various forms of the following linear function were estimated for the revenue function; Yj = ao + o-iXi + biZ{ + CiEde 4- d{Ext + e; where Yj is defined as, crop output revenue, farm output revenue, farm output rev-enue, livestock output revenue, or total farm income depending on the equation being estimated. A'^  is the level of purchased inputs and hired labour in mandays. Zi is the level of farm supplied inputs which include, family labour, land, machinery and buildings, livestock and number of mature tree crops. Machinery and buildings are aggregated into one variable, livestock is estimated in value terms and it is composed of the sum of the value of cattle, goats, sheep, pigs and poultry. The choice of these variables is based on chapter 2 and 3, theory and earlier research work on agricultural production. Ed, is the education variable measured as the number of school standards completed by the farm operator, in some equations it is specified in dummy form for different school categories, where d l= l if farm operator has completed eight or more standards of education and zero otherwise, d2=l if completed four to seven standards of education and zero otherwise, d3=l if completed one to three standards of education and zero otherwise, d4=l if no schooling and zero otherwise. The sample had 51.07% of farm operators with no formal schooling, 11.9% had one to three standards, 25.4% had four to seven standards and 11.63% had eight or more standards of education. The purpose of specifying the school-ing variable in dummy form is to investigate the effect of different school categories on Chapter 4. Empirical Implementation 48 farm revenues and profits. This approach makes it easier to identify within a regression analysis which of the schooling categories has the highest effect on farm revenues and profits. Ext, is a measure of different forms of extension extended to the farmer, these in-clude baraza attendace, demonstration attendance, and extension contact, all specified in dummy form. In some equations the intercation between education and extension service is investigated by fitting a variable that is a multiple of education and frequency of extension service. These dummy variables are fitted to investigate the effectiveness of different forms of extension on farm revenues and profits. The relevance of these vari-ables is based on the fact that they are used as important policy tools by the Kenyan government for the agricultural sector, a^ , fc;, c;, d{ are estimated coefficients, et- is an error term assumed distributed normal with mean zero. Different forms of value added and profit functions with the following general from were also investigated; II = a 0 4- diPxi + bilnZi + CiEd, + d{Ext + e,-/nil = a 0 + bilnPxi + CilnZ + diEd, + eiExt + e; where Zi is the level of farm supplied inputs as defined above, Pxi is price of purchased inputs, including land rental rates and hired labour wage rates. This variable is used in the profit function along with level of fixed forms of farm capital and family labour. The capital component includes the sum of service flow of livestock, farm machinery and buildings and land. Capital is introduced in component and aggregate form in the profit function to examine whether the allocation of capital component within an aggregate generates a higher return to education. II stands for either variable profit (value of farm output minus expenditure on purchased inputs including cost of hired labour) or value added. As in the sales function the effect of educational factors is also investigated, including the schooling of the farm operator and different forms of extension services. To Chapter 4. Empirical Implementation 49 Table 4.2: Summary Statistics Mean, Standard Deviation Variable Mean Standard Deviation value of crop output (VCO) 5740.9 10481 value of livestock (LV) 4825.3 6438.3 sales of livestock products (LVS) 423.56 1851 value of total farm products (FS) 6164.5 10846 off-farm income (OFI) 3475.1 7004.2 value added (VA) 5338.7 12275 variable profit (PRT) 5044.9 11952 schooling of farm operator (edl) 2.77 3.36 schooling of farm operator's spouse (ed2) 2.05 3 value of land (ldv) 84612 213880 value of farm machinery, equipment, buildings and expenses on building renovations (mc) 3050.9 9068.5 family labour (mandays/year)(fml) 453.85 382.87 farm operator labour(mandays/year)(hfl) 166.04 100.47 hired labour (mandays/year) (hrl) 26.41 97.36 hired labour (wage rate/day)(w) 10.45 5.6 age of farm operator (agel) 49.1 14.46 expenditure on crop related inputs (piec) 739.3 1610.4 expenditure on livestock related inputs (piel) 96.11 429.37 land rental rates (rent/year) (rent) 372.79 184.1 land acreage (Id) 5.87 6.1 piea (piec-|-piel) 825.77 1708.3 number of mature trees (mt) 625.2 1831.7 investigate the role of education in allocating hired labour the cost of labour is subtracted from the value added equation and a new equation is estimated without the hired labour variable. Other ways of investigating the differential effects of education investigated are at regional level, by disaggregating the sample into two on regional basis and also at sector level by disaggregating the data by farm sector, viz; crop, livestock and aggregate. Out of the whole sample of 748, 673 of the farmers kept some form of livestock, 274 farmers used hired labour and only 38 farmers farmed on rented land. There were 226 farmers who reported to have had extension contact, 44 had attended a baraza, 57 Chapter 4. Empirical Implementation 50 had attended a demostration and only 6 had attended a field day and five of them had attended a baraza and non of the other activities. This led to strong correlation between baraza and field-day variables and because of the small size of the people who attended field day this variable was dropped from the regression analysis. Chapter 5 RESULTS The objective of this study was to investigate the effect of education (denned as the number of school standards completed by the farm operator) on farm production and profits. This was to be carried on through regression analysis where the education variable was fitted in sales and variable profit functions. It was hypothesized that education has a positive effect on farm production and profits. These hypotheses were confirmed within the regression analysis. Both the Cobb-Douglas and linear estimation procedures were tried. Much emphasis was put on linear functions because of zero values for some of the variables in most obser-vations which made it difficult to transform them into logarithm form without reducing the sample size. Different forms of sales and profit functions were estimated. The starting procedure was to estimate separate equations for livestock and crop sector. This was meant to give a picture of the role of education in the two agricultural activities. It was expected education was more effective in crop production compared to livestock production. This expectation was based on the fact that crop production requires more input outlays com-pared to the livestock production and this gives education a greater role. The other reason is that the crop activity involves several crops and therefore gives the farmers a wider range of choices for the kind of crops to grow and crop mix, and this implies that decision making is more important here than in other agricultural activities. Technolog-ical changes are also more frequent in the crop sector compared to the livestock sector 51 Chapter 5. RESULTS 52 and based on the Shultz theory, we expect benefits of acquiring education to be more in a changing (modernizing) sector compared to the non-changing sectors. Problems were however, encountered in estimating these separate equations as some of the variables were only available for the entire farm sector and it was impossible to separate them for individual farm enterprises. These variables are the family farm supplied labour, hired labour and livestock variable. These equations were nevertheless estimated and results presented in the appendix. Total farm sales model was estimated by aggregating the value of crop output and the value of livestock products into one variable and estimating one equation with this variable as the dependent variable. The schooling coefficient as expected was correspondingly larger in this new aggregate equation compared to that of separate activity equations. This indicates the positive role of education in selection effect. Further improvement was made by incorporating off-farm income into the production process and this meant giving the education variable a greater role, this means we could not only capture the marginal return to education in farm activity, but also its marginal return in off-farm activity. The effect of education in the decision .process of allocating both the farm and non-farm supplied inputs was investigated within variable profit functions. This was done by aggregating and holding fixed farm supplied inputs, and using the price of purchased inputs as right hand variables instead of the physical levels of these inputs in the profit function. Price data was available for only two variables, wage rate and land rental rates. The estimation was carried on a stepwise procedure, where value added was first defined as value of output less the cost of purchased inputs. This estimation only captures the effect of education in choosing and allocating the variable inputs and farm supplied fixed inputs, but does not include education effect in allocating hired labour. To capture the effect of education in allocating hired labour the cost of labour was subtracted from the Chapter 5. RESULTS 53 value added. This means the new equation is estimated without the labour variable in the right hand side and it's effect is moved to the left hand side. In this new equation we have the intuition that hired labour is held constant and estimation captures the allocation of this labour input among competing activities. The main obstacle in investigating effect of education on farm production within either the value added or the profit function framework was the inability to obtain data for the prices of purchased inputs. The only available price data was that of labour wages and land rental rates. The effect of a number of educative factors on productivity were investigated, these include, formal schooling, measured as the number of school standards completed by the farm operator (head of the household), extension contact; this was further disaggregated into extension visits by the extension officer to the family farm, farmers' attendance of a demonstration, or a public baraza attendance. When the data was split into two subsamples for the two provinces, and a new equa-tion estimated for each of the two provinces, the education coefficient was still positive and significant at 10% (see equations 4 and 5 table 5.7), but it was bigger for central province subsample compared to Nyanza province. This result is not surprising because in Central province farmers are more inclined to remain in the farming business even when they get educated and obtain off-farm employment, while in Nyanza province which is far from the national capital people tend to migrate after acquiring education and this makes it difficulty to monitor their farm business as closely as those in the Central province who are much closer to the nation's capital. Thus this is expected to be one of the reasons for the difference in the coefficient size between the two provinces. The other possible reason is that farming is more commercialized in Central than in Nyanza province. The extension service coefficient sign for Central province was negative and insignif-icant while it was positive and insignificant for Nyanza province. This tends to suggest Chapter 5. RESULTS 54 that either there is a difference in the way extension service is administered in the two provinces or that people percieve it differently in the two provinces. The hired labour coeffiecients were bigger for Central province compared to Nyanza province, reflecting the difference in cost of labour between the two provinces. This is not surprising considering the fact that Central province is close to Nairobi and has a comparatively more commercialized agriculture. Considering the whole sample size the education coefficient remained positive and significant at 5% level in all the estimated equations. And it increased in size as we moved from the farm specific activity functions to aggregate farm revenue functions and then to profit functions. This result confirms the hypothesis stated earlier concerning the effect of education on allocative effect. Thus as the farm operator is exposed to more farm and off-farm activities his education becomes more valuable. These regression results are discussed below under different subheadings. Before regression analysis was done the following statistical analysis was done using the sample data; correlation coefficients, yield data, land use and expenditure on pur-chased inputs per acre statistics were calculated for different levels of farmer education, the results are presented in tables 5.3, 5.5, B.ll and 5.4. respectively. Statistical tests were done on table 5.4 statistics and these differences in levels of output by level of schooling were found to be true at 1% level. Further data analysis showed that the level of farmer education was associated with innovation and adoption of new technology. For example about 60% of farmers with four and above standards of schooling were growing hybrid maize, while about 50% of those with no schooling or upto three standards of schooling were growing hybrid maize. About 25% of farmers with four or more standards of schooling were growing local maize, while 83% of those farmers with one to three standards of schooling were growing traditional maize and 37% of those with no schooling were growing traditional maize. 32% of those Chapter 5. RESULTS 55 with eight or more standards of schooling were growing coffee, while only about 20% of the others were growing it. These statistics indicate that educated farmers are more receptive to new technologies. The positive correlation between level of education of the farm operator and expen-diture on purchased inputs show that education influences adoption of new technologies. These purchased inputs are not akin to traditional methods of production and their use indicates that farmers have moved away from traditional ways of farming and are now using more modern methods of production. This exposes them to the obvious risks and costs of acquiring new information and education plays a greater role in adjusting to these new ways. The educated farmers also tend to pay higher wage rates, as indicated by positive correlation between education and wage rates paid to hired labour, this is expected especially if the wage rates reflect labour quality implying that educated farmers seek more experienced labour. This should be reflected by a positive coefficient for hired labour wage rates. The positive correlation between education level of farm operator and value of farm output is not surprising as we expect those farmers who have invested in education to be more productive compared to less educated counterparts. The negative correlation between levels of education and baraza attendance is not surprising as most educated people do not usually attend these meetings as much as their less educated colleagues. The farm operator labour is positively correlated to levels of education, this was not expected as those with higher education are expected to devote most of their time to higher paying off-farm labour activities. However, as the correlation between education level and off-farm income is strongly positive, it shows that educated farmers can supply more farm labour and still be productive in the off-farm sector. This indicates that educated farm operators are good time managers. Chapter 5. RESULTS 56 Table 5.3: Correlation between education and other variables variable correlation coefficient. value of crop output (VCO) 0.123 total off-farm income (OFI) 0.251 value of livestock output (VLS) 0.064 variable profit (PRT) 0.232 value added (VA) 0.235 education in standards of spouse (ED2) 0.386 extension officer visit (EXT) 0.084 public baraza attendance (BARZ) -0.130 attendance of a field day (FLDY) 0.015 expenditure on fertilizer (FERT) 0.163 value of buildings, machinery & expend, on new buildings(MC) 0.139 value of land (LDV) 0.081 i expenditure on feed (FEED) 0.144 other livestock costs (SLOTH) 0.009 expenditure on pesticides and chemicals (PS) 0.108 expenditure on purchased seeds (PSD) 0.056 hired labour wage rate (W) 0.115 hired labour man-days (HRL) 0.066 head of household labour man-days (HFL) 0.042 value of livestock including poultry (LV) -0.001 family labour man-days (FML) -0.006 age of farm operator (AGE1) -0.428 land rental price (RENT) -0.035 Note: Only the ED2 coefficient is significant from zero at 1%. The other estimated coefficients are not significant from zero. Chapter 5. RESULTS 57 Statistical analysis show that those who had eight or more standards of education spend Kshs.260.97 per acre on purchased inputs, those with four to seven standards of schooling spend Kshs.155.37 per acre on purchased inputs per acre, those with one to three standards of schooling spend Kshs. 174.17 per acre while those with no schooling spend about Kshs.99.92 per acre on purchased inputs. These results indicate that those who are educated are more inclined to apply modern technology than the less educated. When the data was disaggregated into particular inputs and the same analysis done, the above results were confirmed for all forms of inputs, and especially expenditure on fertilizer and feed. These statistics are reported in table (5.4). A test of significance of the statistics in this table shows that at 20% significance level, farmers with eight or more standards of education spend more on fertilizer than those with four to seven standards of schooling and those with no schooling. At 20% test level those with one to three standards of schooling spend more on purchased inputs than those with no schooling. Considering all the purchased inputs, although highly educated farmers spend more on purchased inputs than the less edu-cated these differences are not statistically significant. On average the highest educated group employed more labour per acre and paid the highest wage rate compared to other groups. Those with one to four standards of education employed the least amount of labour per acre and paid the lowest wage rate. Further statistical data analysis was done to demonstrate the relationship between education and agricultural productivity. The results are presented in table (5.5). These statistics show value added, expenditure on purchased inputs, and farm production per acre by level of education. These statistics show that higher education is positively related to farm output and value added per acre in both the crop and livestock sectors. For example in the sample 51.07% of the farmers had no education and they produced only 42.04% of total crop output although they had 52.85% of the total crop land area. While those who had one Chapter 5. RESULTS 58 Table 5.4: Expenditure on Purchased Inputs Including Hired Labour by Level of Educa-tion no educ. 1-3 4-7 8+ stds stds stds expenditure on fert/acre 60.62a 113.186 92.49c 231.28d (crop land) expnditure on feed/acre 11.74 39.21 67.31 99.65 (all land) expenditure/acre on all 99.92e 174.17' 155.37* 260.97* inputs excld. labour hired labour 4.01 3.73 3.37 9.81 man-days/acre expenditure on hired 36.80*' 24.63J' 36.79fe 138.08' labour/acre average wage rate/acre 9.17 6.6 10.92 14.07 expenditure on crop related 178.19 287.79 260.91 421.56 inputs/acre(crop land) expenditure on livst. related 25.89 93.61 84.12 149.71 inputs / acre Note:- The following statistics are significantly different at 20%; a and b; a and d; b and d; e and f; e and h; g and h; i and 1; j and 1; k and 1. Chapter 5. RESULTS 59 Table 5.5: Value of Farm Output Per Acre and Value Added on Farm Output Per Acre by Level of Education of Farm Operator Measures of No education 1-3 4-7 8+ Total output stds stds stds crop prod.(Kshs.) 1482.48 2104.84 0 1 2146.32 a 2 2737.45 per acre % of total crop value 42.04 12.63 27.24 18.1 100 value added crop sector 1304.29 1817.05 1885.41 2315.89 per acre Livestock prods, sales 204.2 303.31 b l 336.45 b 2 796.71 per acre % of livst. prod. v l . 33.77 . 15.58 27.92 22.73 100 Value added/acre livst. 178.3 209.7 252.32 645 Total value of 828.37 1152.53 c l 1189.27 c 2 1663.28 output/acre all sectors Value added per acre 691.65 953.73 d 1 997.12 d 2 1246.23 all sectors average age of 54.69 48.47 41.62 41.48 farm operator % of farmers In sample 51.07 11.9 25.4 11.63 100 /:- ay and a2 are significantly different at 1% level /:- bi and b2 are significantly different at 1% level /:- Cj and c 2 are significantly different at 1% level /:- di and d2 are significantly different at 1% level Chapter 5. RESULTS 60 or more standards (48.93%) of education produced 57.96% of total crop output from 47.15% of crop land. For the livestock sector the differences were even greater with those educated producing 66.23% of the total livestock products value. Comparing the value added by education group, those who had one to three standards of schooling had a value added 27.48% greater than those with no schooling. Those with four to seven standards of schooling had value added 4.35% higher than those 1-3 standards of schooling, while those with 8 or more standards of schooling had value added 19.99% higher than those with 4-7 standards of schooling. The average distribution of land ownership was even across different education clas-sifications. Those farmers who had no education had the highest average land acreage per farmer of 6.06 acres, while those with four to seven standards of education had the lowest average of 5.57 acres. A l l the farmers allocated most of their land to crop farm-ing compared to livestock activities. On average the highest educated group allocated a higher percentage (55.5%) of their land to crop activity compared to other groups and the smallest percentage to livestock activity (17.72%), the other land was occupied by homestead and unused land. Those farmers with one to three standards of education allocated 31.72% of their land for livestock enterprise and 50% to crop activity (50%). These results are reported in table ( B 5 . l l ) in the appendix. 5.1 Aggregate Value of Farm output This stud}7 aims at measuring the effect of schooling on farm revenues and profits, thus it is important that we take into consideration all the farm activities open to the farmer, because it is along these lines that he makes his decisions. A n d as schooling is hypoth-esized to enhance the decision making process, if we restrict the farmer to a single farm enterprise then we shall be underestimating his decision making ability. For example a Chapter 5. RESULTS 61 farmer who starts off by growing maize in one season and the next season expands his enterprise to include cash crops like tea will no doubt feel inclined to weigh the pros and cons of allocating his resources among the production of the different crops. In this case decision making becomes important and the role of education becomes more important compared to the case where only one enterprise was operated. Thus by aggregating the total farm output value and fitting it in one equation we have the intuition that we are opening more decision avenues to the farmer and we expect his education to become more valuable. Taking into consideration the above arguments the total value of crop output and live-stock products for each household was aggregated and a different equation was estimated for the aggregate farm output. There are several ways in which decision making process is useful in this case. First we note that by aggregating the crop and livestock activity we allow the farmer to allocate within these two different activities, at the same time we also take into consideration that the crop activity involves different crops, which can either be grown separately or intercropped. Thus the role of decision making becomes important, and we expect those who have invested in education to perform better compared to their less educated counterparts. The regression results were as predicted for this aggregate farm output function. The estimated schooling coefficient for this function allowing schooling and extension interaction shows a marginal return to schooling of Kshs.233.13 and is significant at 5%. This shows that an extra year of schooling of the farm operator will increase total farm revenues by about Kshs.233. The estimated family labour coefficient was positive and significant at 5% level, but smaller than that estimated for hired labour. The reason for differences in the size of these two coefficients is because hired labour is usually hired during the peak period Chapter 5. RESULTS Table 5.6: OLS Regression Results for Total Farm Production independent dependent variable(FS) variable (1) (2) (3) (4) constant 1060.6 974.31 1061.2 1382.9 (1.665) (1.486) (1.781) (1.532) edl 197.880 233.13 179.96 345.2 (1.965) (1.962) (1.640) (1.837) Id 183.960 183.10 171.31 167.15 (2.076) (2.087) (2.326) (1.517) fml 1.344 1.330 1.262 3.085 (1.906) (1.884) (1.714) (1.488) hrl 15.890 15.968 13.844 20.150 (2.700) (2.721) (2.651) (1.269) lv 0.213 0.215 0.214 0.266 (2.392) (2.420) (2.614) (2.850) piea 2.721 2.735 1.145 (piec-f-piel) (5.632) (5.628) (3.431) mc 0.075 ' 0.117 (0.938) (1.144) mt 0.286 0.577 (0.619) (1.227) ext -2048.6 -1845.1 -2148.2 -4335.7 (-2.955) (-2.669) (2.675) (3.414) edext -37.086 (0.692) demo -1418.8 -1209.9 -1551.4 -2290.3 (2.098) (1.165) (2.260) (1.734) barz -126.76 3860.9 (0.136) (1.226) ddl 1.532 (2.537) dd2 3.082 (3.472) dd3 1.678 (3.067) dd4 2.664 (4.202) R2 .33 .33 0.34 0.25 R2 .33 .34 0.35 0.22 F 46.43 37.13 28.05 9.09 N 748 748 748 320 note: ddl=dl*piea; dd2=d2*piea; dd3=d3*piea; dd4=d4*piea Chapter 5. RESULTS 63 and at this time the marginal return to labour is high. Also hired labour has more opportunities compared to family labour whose major form of employment is working on the family farm. The estimated livestock coefficient was positive and significant at 5% two tail t-test, and it shows that an increase in the value of livestock by one shilling will increase the value of farm output by about Kshs.0.22 ceteris paribus. Livestock variable in this function is fitted as the value of livestock and the estimated coefficient reflects net interest rate of livestock capital. The extension contact coefficient was negative and significant at 5% two tail t-test, while the demonstration coefficient was negative and insignificant (equation (2)). This could be interpreted to mean that extension officers only visit those farms with prob-lems or if it is a client initiated visit only farmers experiencing problems seek extension help. Baraza attendance which is another form of extension service had a negative and insignificant coefficient. The results for these variables are surprising as extension service is meant to enhance farmer productivity. This indicates that either there are problems in the way extension service is administered or there exists a bias in the way extension service is offered, in this case implying that only farmers with problems get extension ser-vice. Hopcraft (1974) found similar problems especially when farmers had more extension service contacts and suggested that "may be these officers are harasing farmers". Number of mature trees and farm machinery, buildings and equipment do not have any significant effect on farm revenues. This is surprising since these factors are expected to have positive and significant effects on farm revenues. When education is allowed to interact with extension service, the interaction coeffi-cient is negative but insignificant while the education coefficient increases in size implying that educated farmers benefit more from extension service. The increase in size of the coefficient was about 17.81%. Chapter 5. RESULTS 64 The estimated hired labour coefficient is greater than the average wage rate. The coefficient equals wage rate only if cost of labour is embodied in the wage rate, when perfect capital markets or no risk exist. These assumptions do not necessarily hold and so the discrepancy between the estimated marginal return to hired labour and average wage rate could be a result of existence of imperfect markets, and risks. This coefficient reflects the shadow price of hired labour and it implies that farmers can increase their revenues by using more hired labour. Collier (1989) has found a similar result using Kenyan data and he argues that, the reason why farmers pay less for labour than what they get after using it is because these payments are normally made before any earnings from the use of labour are made. The other reason is because farmers are limited in credit facilities to purchase these services. I tend to agree with these two arguments, Wolgin (1975) found the same result and argued on the basis of risk and lack of credit. These arguments are synonymous because when a farmer makes payment for use of a factor before determing how much he will get out of its use there is always an element of risk involved. The coefficient for farm purchased variable inputs (the sum of all expenditures on livestock-related inputs and crop-related inputs) was positive and significant at a 5% two tail t-test. The size of this coefficient implies that an increase in expenditure on purchased inputs by one shilling will lead to an increase of farm output revenues by Kshs.2.74 ceteris paribus. Again this would appear to indicate that farmers are not optimizing the use of these inputs. Ideally in a profit maximizing situation a shilling spent on an input should earn a shilling in return. However, if a shilling spent on purchased inputs earns more than a shilling in return then the farmer can increase his profits by using more of that input. This is more possible in the manufacturing sector where production is continous and managers have more control of most of the factors that affect their production than in the farm sector where production occcurs seasonally and is difficult to adjust and there also exist factors outside the control of the farmer like weather that affect production. Chapter 5. RESULTS 65 Also in a country like Kenya where resources are limited, farmers use of these factors may be limited due to lack of credit facilities. Wolgin (1975) argues that the reason why farmers earn more returns per unit of purchased input than they spend on it is due to risk aversion and lack of credit facilities. In equation (3) a test of effect of education on optimization of use of purchased inputs is done. To accomplish this we disaggregate schooling into four categories; dl those with 8+ standards of education, d2 those with four to seven standards of education, d3 those with one to three standards of education and d4 those who had no schooling. These dummies are then multiplied with value of purchased farm inputs' (piea) to create four new variables (ddl,dd2,dd3,dd4) where ddl is the product of dl and piea, dd2 is the product of d2 and piea etc. The idea is to find out which of the four has a coefficient that is not significantly different from one, showing optimal use of purchased inputs. The results from this equation indicate all these classifications yield a coefficient that is not significantly different from one. However, those with eight or more standards of schooling have a coefficient that is closest to one. But there is no consistency on the size of the schooling-piea interaction coefficient as a function of schooling level. There are a number of reasons to expect farmers with more education to use more purchased inputs than those with less education. First having education increases their opportunity of obtaining off-farm work, which leads to extra income which they can use to purchase farm inputs. Having education also gives them extra exposure to capital markets and understanding of how these markets function, and this could lead to acquistion of borrowed capital. Also by virtue of their education these farmers have extra advantage of knowing which inputs are usefull for their different farm activities and also their approximate marginal returns. A different set of equations was estimated on the basis of provincial samples to test Chapter 5. RESULTS 66 regional differences in optimization of use of purchased inputs. Not surprisingly, regres-sion results came out in favour of Central province. This is because the use of purchased inputs is more common in Central than in Nyanza province. Also Central province is more developed and agriculture more commercialized here than in Nyanza or any other province. Central province results are reported in table (5.6) equation (4). Nyanza province results which are not reported here show a marginal return to a shilling spend on inputs of Kshs.3.98, showing considerably greater underutilization of purchased inputs by Nyanza farmers. This result provides strong support on the importance of access to credit and/or technical information. In a separate analysis where the sample was split into two on the basis of age, with one sample composed of farm operators who were 40 years or young, it was found that the marginal return to schooling was higher for the subsample of older farm operators, and negative and insignificant for young operators. There seems to be an indication of young farmers devoting more of their time to higher-paying off-farm activities, to the neglect of their farms. The reason why older folks have positive and higher marginal return is that apart from having less education on average, their age does not qualify them for as many off-farm opportunities as the young farmers, and so they use their acquired knowledge through education to improve their farm productivity. This hypothesis ought to be studied further. Livestock variable is expressed in value terms and the estimated coefficient for this variable indicates that it has a positive and significant effect on farm output revenues. This coefficient is significant at 5% level two tail test. The model explains about 33% of the variation in farm revenues as indicated by the adjusted coefficient of determination, a relatively high degree of explanatory power for individual farm data. Chapter 5. RESULTS 67 5.2 Off-farm and Total Family Income A considerable number of small-farm operators in Kenya are also engaged in non farm activities, and these activities have a direct effect on the farmers decision making. A rational farmer will allocate his time in a way that maximizes his returns. The decision to work off-farm will depend on his gain expectations. Thus, if returns from working off-farm are higher compared to farm income he will devote more of his resources/time to off-farm activities and the opportunity cost for his time and education will rise as these off- farm activities are allowed into the production function. The probability of earning higher off-farm income increases with his level of education, while his farm labour supply and that of his family members have a negative effect on off-farm family income. Hired labour is likely to be associated with high off-farm incomes. This is becuase hiring labour releases family labour to engage in high paying off-farm activities and also because hired labour also performs none-farm activities within the family business. Huffman (1980) has explored the effect of investment in education and information (extension) on the off-farm labour supply of farmers using U.S data. He finds that raising the education level of farmers and increasing the agricultural extension inputs increase the off-farm labour supply of farmers, implying that part of the return to education in agriculture arises from its effect on the reallocation of farmers' labour services between farm and nonfarm labour markets. The existence of off-farm activities raises a farmer's opportunity cost and the value of both his time and human capital increases. The role of schooling in decision making under these circumstances becomes important and it should be reflected by the schooling coefficient being correspondingly bigger in an equation where off-farm income is included in the dependent variable. The role of extension is not expected to change when off-farm income is added to Chapter 5. RESULTS 68 the farm income. This is because extension service is farm activity related and has nothing to do with off-farm income thus there is no reason to expect this variable to behave differently from the equation where only farm output revenue was considered as the dependent variable. What amounts to off-farm income in this study is the sum of net income from non-farm businesses, plus all salaries and wages earned by the family members from activities other than their farm output value. The off-farm equation behaved very much as predicted despite lack of sufficient rel-evant data. The estimated schooling coefficient was positive and significant, showing a marginal off-farm income return to schooling of Kshs.513.32. As expected, supply of family labour to family farm had negative effects on off-farm incomes. The estimated coefficient for this variable was negative and significant at a 1% level. A regional dummy variable (defined as RD=1 if Central province and RD=0 if Nyanza) was used to capture regional differences in off-farm incomes, and as expected the coefficient for the dummy was positive and significant at 20%, indicating higher off-farm incomes in Central com-pared to Nyanza province. As expected the schooling coefficient was much bigger when the off- farm income earned by the farm operator was added to the total value of farm output. For example comparing the schooling coefficient for equation (2) in the farm output model table (5.6) with the schooling coefficient in equation (2) in the total family income model table (5.7) we find that there is an increase in the schooling coefficient size by over 234%. This increase in the size of the coefficient captures the return to education from engaging in off-farm activities. Note the sum of estimated coefficients for schooling variable in farm output model and the off-farm income equation is less than the estimated coefficient for total family income model, the difference (Kshs.28.44) is the measure of allocative efficiency of education. Chapter 5. RESULTS 69 Table 5.7: OLS regression Results for Total Family Income(TFY) and Off-Farm In-come(OFI) equations independent variable (l)(OFI) (2)(TFY) (3)°(TFY) (4)b(TFY) constant 3315.8 3597.3 5241.8 3198.3 (5.545) (4.950) (4.274) (4.055) edl 513.13 778.89 850.30 581.78 (5.356) (4.815) (3.054) (3.271) W -.002c 168.00 22.737 207.13 (1.597) (1.992) (0.161) (2.481) hrl -47.130d 23.201 29.216 16.663 (1.001) (3.252) (1.736) (2.776) fml -2.147 -0.689 0.928 -1.178 (4.616) (0.796) (0.400) (1.233) lv 0.215 0.263 0.210 (2.461) (2.034) (1.802) piea 2.407 0.970 3.513 (piec-fpiel) (4.950) (2.652) (6.530) rd 853.28 (1.477) 635.20 (0.710) ext -1657.8 -4366.0 345.98 (1.855) (3.052) (0.309) edext -57.384 -35.204 -115.33 (0.744) (0.387) (1.195) demo 821.61 -2267.2 . 2484.3 . (0.534) (1.260) (1.234) R2 0.08 0.27 0.19 0.38 R2 0.08 0.26 0.22 0.40 F 13.13 27.05 7.34 24.99 N 748 748 320 428 Note:- a=Central province subsample; b=Nyanza province subsample, the dependent variable is total family income. :-c: variable expressed as value of land; d: variable expressed as wage rate for farm labour. Chapter 5. RESULTS 70 The family labour coefficient became negative and insignificant showing that, the proportion of off-farm income in the total farm income is higher than the farm sales revenue. This coefficient indicates that supply of farm labour has a negative effect on total farm incomes. The net effect of family farm labour supply on total family income is therefore insignificant. Hired labour coefficient increases in size capturing perhaps the role of this input in the off-farm activities in the family business. This is no surprise as in most cases domestic workers perform both farm and off-farm work within the family, and in most cases those who perform both activities usually are more "skilled" implying that they have a higher opportunity cost compared to those who just perform farm work only. The estimated coefficient remained significant at 5%. This result could also imply that as off-farm incomes increase there is greater demand for hired labour, indicating that this variable is not entirly exogenous. The purchased inputs coefficients were significant at 5% level two tail test in the two equations, but their size was lower compared to the model where only farm output value was considered. This is due to the fact that farm purchased inputs have no effect on off-farm incomes, which form part of the dependent variable. The size of livestock coefficient remained almost the same after off-farm income was added to the farm output value, and it remained significant at 5% level two tail test. The lack of change for the size of this coefficient indicates that it has no influence on off-farm income. This coefficient shows that an increase in the value of livestock by one shilling will lead to an increase in farm incomes by Kshs. 0.2 ceteris paribus. It shows the rate of return or net interest rate of livestock capital. When the sample was divided into two samples on provincial basis the schooling coefficient remained stable showing that schooling had positive effect on family income in both provinces. The estimated schooling coefficient was bigger for Central province Chapter 5. RESULTS 71 sample compared to that of Nyanza, this implies that schooling has a higher return for central province farmers compared to Nyanza farmers. This is not surprising as farming is more commercialized and advanced in Central compared to Nyanza province, as such the benefits for investing in education are more in this region. It is worth noting that by adding off-farm income to total farm product revenue the coefficient of schooling for Nyanza province becomes positive and significant compared to the case where only farm product revenues were considered. This shows that most of the returns to education in this region are derived from working off-farm. This is contrast to Central province where returns to education are positive for both farm and none farm activities. A possible explanation for this is the proximity of Central to Nairobi enabling people from this province to work both off-farm and at the same time be able to attend to their farms mostly over the weekends. People from Nyanza are disadvantaged in this respect as they are far from Nairobi and if they choose to farm as well as work off-farm as far as Nairobi then obviously their farms will experience a comparative neglect. The other possible explanation is the differences in development between the two provinces, where Central is more developed and based on the Schultz theory more likely to benefit from investing in education. Also hired labour had a higher shadow price for Central sample showing the oppor-tunity cost of hired labour is higher in this region. Family labour had no significant contribution to total farm incomes showing that, the number of mandays worked by the family members do not have effect on farm incomes, this could imply that most of the family income especially that part accounted for by off-farm income has very little to do with family labour. The estimated coefficient for the value of purchased inputs for Central province sample is close to unit, while that of Nyanza province is significantly different from unit showing that farmers in Central province are using purchased inputs more optimally compared to those in Nyanza province. The same reasons advanced for Chapter 5. RESULTS 72 the difference in the size of the coefficient for this variable between the two provinces in the sales function also apply here. The model explains 27% of the variation in farm incomes, as indicated by the adjusted coefficient of determination. 5.3 Value Added Equations The ultimate goal of any investment is not just the maximization of revenues, but rather profits. The small-farm operator in Kenya will rationally use his resources in a way that enables him to earn positive marginal returns. Economically any enterprise that continuously experiences losses will get out of business. Thus we expect farmers to behave the same way as any rational business person, that is, to operate in a way that maximizes gross margins or value added. We have seen evidence in the sales function that these farmers are not optimizing the use of purchased inputs but given the production risks and financial constraints under which they produce, there is no reason to argue that these farmers are being innefHcient and so we can assume they are maximizing their earnings given these conditions. The use of the value added approach is a superior measure of farmer effeciency in allocation of his factors of production in the farm business compared to the sales function approach. . We observe that the rational farmer will apply his factors of production in a way that yields maximum returns. In the value added approach we move all the purchased inputs from the right hand side to the left hand side by subtracting their value from the farm gross revenues. The dependent variable is now the value of farm revenues net of purchased inputs. We estimate this "variable profit" as a function of three classes of exogenous variables; price of purchased inputs, fixed-farm supplied inputs and educational variables. This assumes that the farmer allocates his purchased inputs in such Chapter 5. RESULTS 73 a way that maximizes the value added. The educational variables which are included as independent variables are expected to have positive and significant effect on the value added, showing positive allocative and selection effect of education. A set of value added or restricted profit functions were estimated. When estimating this function we express value added as a function of family labour, farm supplied capital inputs, price of purchased inputs and educative factors. Family labour and farm capital are held "fixed" and we have the intuition that, reallocation of these factors with com-peting farm activities enables the farmer to maximize his returns. The hypothesis was that educated farmers experience more profits in their farm business compared to their less educated counterparts. This is reflected by a positive and significant coefficient for the schooling variable showing schooling has a positive contribution to farm profits. A further step was undertaken to test the allocative effect of education through aggre-gation and holding constant certain factors so that education can allocate these "fixed" factors among competing activities. To arrive at the value added function total cost of purchased inputs was subtracted from the total value of farm output. Thus given value of farm output for farm j as, Yj = J2i YijPij = PijQ(Xj, Zj, Edj), where Y{j is the amount of output for product i from farm j and PtJ- is the price of the same product for farm j. Edj is the education level in school standards completed by farm operator for farm j. And given the total expenditure for farm j on purchased inputs as, where X{j is the quantity of input X purchased at price Pxij by farm j, then the value added can be expressed as, VAj = JZtj YtjPij - XijP^j = PijQiXj, Zj} Edj) - XtjP^j. Chapter 5. RESULTS 74 where Zj is the quantity of farm j supplied inputs. This value added was then ex-pressed as a function of hired labour wage rate, land rental rate, farm supplied inputs and farm operators educative factors. The farm supplied inputs include family labour, land, livestock, buildings, machinery and equipment, and tree crops. The capital component data is biased downwards due to the fact that some farmers did not report the value of their buildings. The value of buildings, machinery and equipment used in this study are those which were acquired between 1975 and 1982. Livestock value is as reported by the farmers, by their own valuation of their livestock based on local prices. A capital service flow component was computed by aggregating the level of service flow of livestock, land, machinery and equipment. The idea was to use this aggregated capital in the value added equation and hold it fixed so as to capture the effect of education in allocation of the capital components among competing activities. The estimated schooling coefficient remained positive and significant at 5% level two tail test in the value added and profit function. The size of the coefficient (equation 2 table (5.8)) increased by about 17% compared to that estimated in the farm revenue function (equation 2 table (5.6)). This increase in coefficient size is a measure of the allocative and input selection effect in maximizing farm profits. When the level of farm "fixed" supplied capital was aggregated into one component (equation (2) table (5.8) ) the estimated schooling coefficient increased in size by about 5% showing the positive allocative effect of schooling on farm capital. The estimated schooling coefficient remained stable and significant after aggregation of the farm capital. The estimated coefficient for cost of hired labour was positive and significant at 5%, possibly reflecting that wage rate is based on quality of labour. In equations (1) and (2) the estimated coefficient for cost of labour is greater than unit,, showing this factor is underutilized, thus farmers can increase their farm profits by hiring more labour. Ideally in a profit maximizing situation we expect a shilling spent on labour to earn a shilling in Chapter 5. RESULTS 75 Table 5.8: OLS Regression Results For Various Specifications of Value added Equations independent dependent variable variable (1)VA (2)VA (3)PRT (4)VA (5)VA constant 886.28 2506.3 664.49 7.499 7.404 (1.234) (4.002) (0.687) (26.259) (26.253) edl 259.14 272.80 281.68 0.024' 0.053 (2.028) (2.243) (2.081) (1.492) (3.735) Fml 1.579 2.160 1.639 0.025 0.026 (2.120) (2.780) (2.154) (0.547) (0.579) cost of 1.994 2.412 0.0001 0.0001 labour(CL) (2.563) (2.626) (4.837) (5.036) wage rate(w) 42.088 (0.637) k (lv+mc+ldv) 0.005 (2.762) land (Id) 252.90 (2.542) 316.28 (1.974) LV 0.210 1.926 0.00003 0.00003 (2.542) (1.989) (4.401) (4.373) mach,eqp(mc) 0.076 0.087 & bldgs (0.932) (1.065) lm(ld+mc) 0.091 (3.185) 0.091 (3.205) mt 0.633 (1.429) 0.720 (1.355) 0.604 (1.340) barz -415.65 502.82 -729.54 -0.461 -0.396 (0.446) (0.436) (0.811) (3.532) (3.156) ext -1466.4 -1552.5 -1623.1 0.090 0.066 (1.758) (2.043) (2.019) (0.975) (0.723) demo -1625.3 -1552.2 -1364.8 0.115 0.105 (2.240) (1.902) (2.144) (0.749) (0.690) R2 0.19 0.14 0.12 0.10 0.13 R2 0.20 0.15 0.13 0.11 0.12 f 18.41 16.37 10.93 11.19 13.35 n 748 748 748 711 711 f/; this coefficient is estimated using education of the farm operator spouse (equation number 4). /;equation 3 is estimated with farm revenues less cost of purchased inputs less cost of hired labour Chapter 5. RESULTS 76 return, but if it earns more as shown in equations (1) and (2) then farmers can increase their profits by hiring more labour. This result is consistent with earlier results. Family labour remained positive and significant at 5% level two tail t-test considering the whole sample. This variable is fitted as the number of mandays of family labour spent on family farm and the estimated coefficient reflects the shadow price of family labour. The size of this coefficient is not surprising as most family members have no alternative off-farm employment and thus their opportunity cost is very low. Number of tree crops variable (mt) had a positive and significant coefficient at 20% t-test (see equation (1) through (3)), but positive and insignificant when schooling farm operator is fitted in dummy form (equation (8) table 5.9). In equation (3) table (5.8) cost of hired labour is subtracted from value added, so that we can pick up the returns to education from allocating hired labour. The result is an increase in the schooling estimated coefficient by 8.7% (compare equation (1) and (2) table (5.8)) reflecting the return to education from allocating hired labour. The schooling coefficient remained significant at 5% level two tail t-test. All the information (extension) variables had negative coefficients. The extension contact coefficient was negative and significant in most of the estimated equations. In equation (2) and (3) for example the estimated extension contact coefficient was negative and significant at 5% level t-test. The estimated demonstration coefficient behaved much the same way as the extension contact variable. The baraza variable had mostly a negative and insignificant coefficient, with the exception of the equation where schooling of the farm operator was fitted in dummy form (equation 8 table 5.9) or the schooling of the spouse was used (equation 4 table 5.8), in which case the estimated coefficient was negative and significant at 5% in both cases. In equation (4) schooling of the farm operator's spouse is used and the results indicate that it has a positive effect on farm profits, but only significant at 20%. This equation Chapter 5. RESULTS 77 is estimated for those farms which used family labour and had greater than zero value added. This equation is expressed in loglinear form with value added, family labour and farm machinery equipment, building and land variables fitted in log form while the other variables are entered exponentially. The estimated schooling coefficient indicates that an increase of the spouses schooling by one standard is associeted with a 2.5% increase in farm profits. Equation (5) is estimated with the same specification as equation (4), but with schooling of the farm operator used instead of that of the spouse. The estimated schooling coefficient indicates that an increase in the level of farm operator schooling is associated with a 5.3% increase in farm profits. Comparing the schooling coefficients in the two equations we find that the return to schooling of the farm operator is more than double that of the spouse. This explains the fact that farm profits depend more on the schooling of the farm operator than that of the spouse. The signs on the coefficients of these extension variables seem to indicate that either the way the variables were measured is wrong or extension service in Kenya is far off the mark of helping farmers to increase their productivity. It may be that these meetings which are surpposedly meant to educate farmers do not measure up to their expectation or they emphasize other issues other than farming concerns. The other possible explanation for the signs on these coefficients is that maybe farmers only seek information when they encounter problems, implying that only those farmers who have problems seek this kind of information. When the sample size was limited to those farms which employed labour (equation 6 table 5.9) the schooling coefficient increased in size to 661.04, capturing the return to education from managing an outside labour. This is an increase of about 155% (compare equation (1) and (6)). The family labour coefficient also increased for the same reasons. Similar results were obtained by Barichello (1979) using Canadian data. The estimated schooling coefficient for this equation is positive and significant at 5% level two tail t-test. Chapter 5. RESULTS 78 Table 5.9: OLS Regression Results For Various Specifications of Value Added functions independent dependent variable variable (6)VA (7)VA (8)VA (9)VA constant 582.29 2386.00 1012.0 1460.0 (0.329) (4.402) (0.837) (1.614) edl 661.04 -1.213 339.84 (2.351) (0.007) (1.810) Fml 2.350 0.775 1.648 3.151 (1.422) (0.859) (2.112) (1.573) wage rate(w) 203.45 133.06 1.650° (1.544) (1.778) (1.269) land (Id) 370.61 (1.884) 170.89 (1.540) LV 0.405 0.257 0.175 0.274 (2.949) (3.607) (1.612) (2.961) mach,eqp(mc) -.081 0.168 0.106 0.115 & bldgs (1.974) (2.178) (1.374) (1.127) barz -1420.6 1136.8 -1830.4 3758.0 (1.309) (0'.428) (1.985) (1.200) ext -1114.5 -2061.7 -1756.7 -4297.5 (0.681) (2.880) (2.122) (3.373) demo -1819.5 -696.01 -1125.3 -2312.6 (1.544) (1.030) (1.791) (1.742) mt -0.262 2.106 0.558 0.611 (1.373) (3.339) (1.246) (1.310) rent -8.304 (1.160) 5.664 (1.123) dl 548.37 (0.389) d2 818.53 (0.956) d3 -657.92 (0.988) R2 0.11 0.19 0.07 0.19 R2 0.14 0.17 0.09 0.16 f 4.24 10.70 6.46 7.28 n 274 474 748 320 a:- this variable is the cost of hired labour Chapter 5. RESULTS 79 Equation (7) is estimated for those farms which did not use hired farm labour. The estimated schooling coefficient for this subsample shows that schooling has no effect on farm profits. The family labour coefficient is also insignificant for this sample showing that family labour has no effect on farm profits for this subsample. This could imply that these farms have excess family labour and therefore it's marginal • contribution to value added is negligible. This could be the case particularly if the reason for these farms not hiring labour is because they already have excess family labour. Equation (8) is estimated with schooling variable defined in dummy form for different education categories. The results indicate that four and above standards of education have positive though insignificant effect on farm profits. One to three standards of schooling had negative but insignificant effect on farm profits. The other educational variables remain positive and insignificant. Equation (9) is fitted using Central province sample. The reason for choosing this province is based on the fact that its sales function showed approximate optimal use of purchased inputs and so meets most of the conditions for profit function. The results from this equation are as predicted and they show a marginal profit return to schooling of Kshs.335.00. All the other variables had signs similar to those estimated in the sales function. In conclusion the schooling of farm operator has a strong positive effect on farm revenues and profits, while at the same time having strong positive effect on off-farm income. The logical conclusion from these results is that educated farmers have better incomes than less educated farmers whether we are concerned about farm incomes or non-farm incomes. On the contrast the information variables; baraza, demostration and extension contact all had negative and mostly significant effects on farm revenues and profits, casting doubts on the effectiveness of Kenyan extension service in the early 1980's on increasing farmer productivity. Possibly this result should serve as a bench mark for Chapter 5. RESULTS 80 re-examination of the way these services are rendered. Ideally extension services are meant to enhance farmer productivity, but their effectiviness depends very much on their quality in terms of the message they convey and the way they are administered. In this case only three forms of extension services were reported and all had negative effects on farmer performance. However, we are not able to judge whether it is the way these services are rendered or the quality of the information they convey that leads to poor performance. Chapter 6 Summary, Recommendations and Limitations This study was undertaken to evaluate the effect of education (defined as number of school standards completed by the farm operator) on small farm revenues and profits in Kenya. It was hypothesized that education has a positive effect on farm revenues and profits and that education increases farm productivity. All the above hypotheses were confirmed for the small scale farmer in Kenya. The effect of other educational factors like extension contact, demonstration and baraza attendance on farm production and profits were also investigated. A simple procedure was followed to investigate whether educated farmers were more productive compared to less educated farmers. First correlation coefficients between education and other farm productive factors, and farm revenues were calculated. The results from the correlation analysis show that education is positively correlated to the levels of purchased farm variable inputs including hired labour, farm revenues, farm variable profits and off-farm income. Schooling of the farm operator was negatively correlated to baraza attendance, farm family labour supply, value of livestock (including poultry) and land rental rates. When productivity data was computed, the results showed that the more educated the farmers were the more productive they were taking into consideration all sectors. These results are reported in table (5.5) chapter (5). This analysis was systematically done in the following order; first farm revenue data per acre for each education classification was done, and it became evident that the more educated the farmers were, the more 81 Chapter 6. Summary, Recommendations and Limitations 82 productive they were. This was true even when the data was disaggregated for the crop and livestock sectors. Further analysis was done by computing the value added per acre for different education levels, the results showed that the more educated the farmers were the higher the value added they had per acre. In terms of value of farm output, the educated farmers had a higher gross farm revenue from a relatively smaller area of land compared to their non- educated counterparts. Further data analysis showed that the more educated the farmers were the more they spend on purchased inputs per acre. This suggests that education is positively related to adoption of new technology. For example those farmers with eight or more standards of education used crop related purchased inputs valued at Kshs.421.56 per acre, while those with no education spent only Kshs.178.19 per acre. Similar results were obtained when the overall farm expenditures on purchased inputs per acre were computed. This indicates that education has a positive effect on innovation and adoption. This could imply that educated farmers are overusing farm purchased inputs but if this were so then the education coefficient in the profit function should have a negative sign. To test if educated farmers were overusing purchased inputs a new equation with the ratio of cost of purchased inputs to total farm revenue as the dependent variable was estimated and the schooling coefficient had a negative sign showing educated farmers were not overusing these inputs. There are a number of reasons why educated farmers may be better positioned to spend more on purchased inputs, first the fact that they are educated enables them to have more access to off-farm income (a fact already supported by data analysis in this study) which they can use to purchase farm inputs. These farmers are also in a better position to understand the use of purchased inputs and for this reason are more likely to use them as they understand their benefits. Exposure to education leads to exposure to many things like capital markets and so such farmers are better positioned to acquire Chapter 6. Summary, Recommendations and Limitations 83 borrowed capital. Livestock and crop specific functions are estimated and reported in the appendix. The results indicated that education had higher returns in the crop sector. This is not surprising because the crop sector has more options compared to the livestock sector, it is also a more dynamic sector, and hence a better beneficiary of decision making compared to the livestock sector. Difficulties encountered in estimating farm specific enterprises are reported in chapter (V). When the two sectors were aggregated, which was the next step of testing allocative effect the schooling coefficient was evidently bigger compared to the case where the two sectors were estimated separately. The estimated schooling coefficient remained significant at 5% level two tail t-test. When the level of variable input expenditures was aggregated into one variable the schooling coefficient increased in size and was also more significant showing that education influenced selection and allocative efficiency. A further step was taken in which the off-farm income was aggregated with the farm output revenue to generate total farm income. When a new regression equation was estimated with total farm income as the dependent variable the estimated education coefficient was bigger and more significant compared to the previous regressions. The estimated education coefficient allowing for education extension interaction showed the mariginal return to schooling of the farm operator to be Kshs.778.89 ceteris paribus and was significant at 5% level two tail t-test (see table 5.7 equation (2)). In the variable profit functions table (5.8), the schooling coefficient remained positive and significant showing that schooling increases farm profitability. To test the allocative effect of education on farm supplied inputs, these inputs were aggregated into one capital component and a new equation estimated, the schooling coefficient increased by about 5.27% indicating schooling has a positive allocative effect in farming (compare equation (1) and (2) table (5.8)). This increase in the marginal product of schooling is a measure of allocative effect of schooling Chapter 6. Summary, Recommendations and Limitations 84 on farm capital. The effect of schooling in allocating hired labour was also explored by subtracting the cost of hired labour from the value added and observing the behavior of the schooling coefficient. As expected the schooling coefficient increased in size by about 8.7% reflecting the return to education from allocating hired labour (compare equation (1) and (3) table 5.8). An equation using a subsample of farmers who employed labour was estimated and the schooling coefficient increased in size showing the return to education from managing an outside labour. The extension variables did not appeal much as policy options to increase farm rev-enues and profits. This is indicated by the fact that they all had negative and mostly significant coefficient showing that they lead to a decline in farm revenues and profits. Extension contact appears to lead to a decline in farm revenues and profits as evi-denced in the regression analysis. This is based on the fact that extension service had negative and significant coefficient at 5% levels in most of the regressions. Baraza atten-dance was only observed in Central province and when fitted in the regression equations it had a negative effect on farm profits, but positive and insignificant effect on farm rev-enues. It is surprising that none of the Nyanza province respondents reported to have attended a baraza considering that baraza is widely used in Kenya as a means of educat-ing the masses on different government programs and that Nyanza subsample was bigger than Central province subsample. The negative coefficients on these educational vari-ables remains suspicious since policy-wise they are meant to enhance farm productivity, but of course their success depends on how they are communicated and the quality of the extension service or demonstration. It also depends on how farmers respond to these services but as it is, based on the present data set these extension services fall short of expectations. Similar results were obtained by Hopcraft (1974) for small farm sector in Kenya using a different data set. This shows there are problems in Kenya concerning the Chapter 6. Summary, Recommendations and Limitations 85 way extension service is administered and the whole system needs re-examination to find why it is not serving its intended purpose. Even mere productivity data analysis as reported in table (B.12) (see appendix) indicates that those farmers who did not attend any of these meetings faired better in their farm operation compared to those who attended, except for those who attended field-day, but they were only six out of the whole sample size of 748. Hired labour had a higher shadow price than family labour showing a higher marginal productivity compared to family labour. This is not surprising as hired labour is used during peak period when marginal returns to labour are high. Hired labour also has higher opportunity cost compared to family labour. This is evidenced by size of the hired labour coefficient on the farm revenue functions (see table 5.6 and 5.7). This coefficient which is a shadow price of the cost of labour is higher than the average hired labour farm wage rate for whole sample which is about Kshs. 10. This could be explained on the basis of lack of credit at the beginning of the farming activity to hire labour. This difference also could result from the fact that farmers face other costs associated with hired labour which are not reflected in the wage rate, like cost of housing and feeding the labour. There are also risks involved with hired labour eg. possible non availability of this labour when needed. In the value added equations cost of hired labour is used as an independent variable and the estimated coefficient indicates that this factor is underutilised. The estimated coefficient is significantly greater than unit implying farmers can increase their profit margins by increasing use of hired labour. The number of mature tree crops owned by the farm operator had positive effect on farm profits. This is not surprising as coffee and tea are two of the most important revenue generators in the rural areas in Kenya. This variable had a coefficient which was signficant at 20% level. Perhaps the low size of significance of this variable should be associated with the way it was specified. For example if this variable could be specified Chapter 6. Summary, Recommendations and Limitations 86 in a way that reflects quality of the trees, then a more significant coefficient could be obtained. The present data set does not allow us to do this estimation. Results based on this study cast doubts on the current trend of use of profit functions to study agricultural production developing countries (Hopcraft 1974, Pudasaini 1983, Jamison et al. 1980). As shown in the sales function and in the crop function reported in the appendix, a number of profit maximizing conditions are violated even though the profit function still produced agreeable results. Thus estimating straight profit functions may give a false impression about the underlying agricultural production conditions in the developing countries. Results based on this study indicate that formal schooling though not specifically undertaken for agricultural purposes has a considerable impact on farmer performance. It was shown using the present data set that adoption of new agricultural techniques is positively related to level of the farmers' formal schooling. Thus with changing technolo-gies and as the population pressure limits the size of available land those farmers who have invested in education will come out as better performers. Thus the government should intensify its campaign for literacy and education for all. Such a campaign will not only benefit- the farmers by increasing their productivity, but also lead to increased rural incomes and of course more food. For example the marginal return tb education on farm profits was estimated at Kshs.281 (see equation (3) table 5.8), not to mention the benefits of having an educated population. This study does not prescribe schooling as the panacea for all small farm productivity problems in Kenya. As the study shows other factors like expenditure on purchased inputs like fertilizer, feed, pesticides had strong positive effect on farm revenues. This suggests that the government should in addition to increasing education levels among the farmers advance credit facilities to the farmers to purchase farm inputs and hire labour. The estimated coefficient for purchased farm inputs variable indicates that farmers are Chapter 6. Summary, Recommendations and Limitations 87 not using optimal levels of purchased inputs and they can increase their revenues/profits by increasing expenditures on purchased inputs, especially on feed and fertilizer. The government should design policies aimed at encouraging farmers to use more purchased inputs. Wolgin (1975) has noted that there is a considerable lack of credit facilities for small farmers in Kenya to purchase much needed farm inputs, while Collier (1989) associates the discrepancy between marginal returns to small farm hired labour in Kenya and the wage rate to lack of credit facilities and risk as labour is paid for before the produce is harvested from the farm. Results based on this data set indicate extension service has negative effects on farm revenues and profits. This calls for a re-examination of the way these services are rendered and the kind of information they convey to the farmers. A possible solution is to give more emphasis to training and visit method as it has been found to be more effective in other parts of Africa. The advantage with this method is that extension officers will ensure farmers utilize the training they get appropriately. It also gives farmers an opportunity to get clarifications on things which become unclear when they actually try the new methods of production learned from extension. Expenditure on purchased inputs like fertilizer, feed, pesticides had strong positive ef-fect on farm revenues. This suggests that the government should together with increasing education levels among the farmers advance credit facilities to the farmers to purchase farm inputs and hire labour. The estimated coefficient for purchased farm inputs vari-able indicates that farmers are not using optimal levels of purchased inputs and they can increase their revenues/profits by increasing expenditures on purchased inputs, espe-cially on feed and fertilizer. The government should design policies aimed at encouraging farmers to use more purchased inputs. There are three reasons why farmers may not be optimally utilizing these inputs, 1) lack of credid to purchase these inputs at the beginning of planting season, this calls for Chapter 6. Summary, Recommendations and Limitations 88 more resources to be directed to agricultural credit facility. 2) Farmers being risk averse, thus spending amounts which will at least ensure positive returns as their operations are affected by other factors outside their control like weather conditions, 3) These inputs are bought at the beginning of the season and farmers have no way of knowing how much they will get in return after using them. The current data set is limited in making conclusive remarks concerning the negative coefficients in the extension variables as there are a number of reasons that might lead to negative coefficient as discussed. Also concerning the use of purchased inputs it is not clear whether farmers are underutilising these inputs due to lack of credit facilities or they are just being risk averse. Another possible reason why farmers may not optimally use these inputs is due to lack of these inputs in appropriate packages. Similar reasons could be advanced for lack of optimal use of hired labour. These are issues that arose in the course of this study and should be studied further for any conclusive recommendations to, be made. Nevertheless education of the farm operator though not undertaken specifically for the purpose of gaining useful knowledge to work on the farm high marginal returns to farm production. Bibliography [1] Anker, R. and Knowles, J.C. "Population Growth, Employment and Economic-Demographic Interactions in Kenya: Bachue-Kenya," (ILO/Gower, St. Martin's Press: New York) (1983) [2] Barichello, R. "The Schooling of Farm Youth in Canada." Unpublished Ph.D. dis-sertation, University of Chicago, (December 1979). [3] ."The Productive Value of Schooling in Canadian Agriculture." Working paper in Microeconomics Workshop in Labour and Population. Yale University and University of British Columbia. February, 1984. [4] Collier, P. "Contractual Constraints on Labour Exchange in Rural Kenya." Inter-national Labour Review (ILO/Geneva) Vol. 128 No.6 (1989) 745-768. [5] Cunningham-Dunlop, Catherine. "The Effect of Schooling on Farm Productivity and profits in Canada." M.Sc. Thesis, University of British Columbia (June 1986) [6] Eicher, C.K.; Baker, D.C. "Research on Agricultural Development in Sub-Saharan Africa: A critical survey." Department of Agricultural Economics, Michigan State University.(1982): 47 [7] De Boer, J. and Chandra, S. "A Model of Crop Selection in Semi- Subsistence Agriculture in Fiji." American Journal of Agricultural Economics. Vol. 60 No. 3 (August 1978) 436-444 [8] Europa Publications Limited: Africa South of Sahara. England (1988) p. 572-593. [9] Fane, C.G. "The Productive Value of Education in The U.S. Corn Belt". Ph.D. Dissertation Harvard University, Cambridge, Massachusetts. May 1972. [10] Gintis, Herbert. "Educational Production Relationships, Education, Technology, and The Characteristics of Worker Productivity." American Economic Review, Pa-pers and Proceedings Vol. 61, No. 2, (May 1971): 266-279. [11] Gisser, M. "Schooling and the Farm problem." Econometrica. Vol. 33, No. 3, (July 1965): 582-592. [12] Griliches, Z. "Specification Bias in Estimates of Production Functions." Journal of Farm Economics. Vol. 39, No.l (February 1957): 8-20. 89 Bibliography 90 ."Estimating the Returns to Schooling: Some Econometric Problems." Econometrica. Vol. 45, No.l (January 1977):l-22. -."Estimates of Aggregate Agricultural Production Function from Cross-Sectional Data." Journal of Farm Economics. Vol. 45, No. 2 (May 1963) Hazlewood Arthur. "The Economy of Kenya: The Kenyatta Era." Oxford University Press. (1979) Hopcraft, Peter. N. "Human Resources and Technical skills in Agricultural Devel-opment: An Economic Evaluation of Educative Investments in Kenya's Small-Farm Sector". Ph.D Dissertation, Stanford University, 1974. Huffman, W. "Decision Making: The Role of Education." American Journal of Agricultural Economics. Vol. 56 (February 1974): 84-97. . "Allocative Efficiency: The Role of Human Capital." Quarterly Journal of Economics. Vol. 91 (February 1977): 59-79. ."Farm and Off-Farm Work Decisions: The Role of Human Capital." Review of Economics and Statistics. Vol. 62, No.l (February 1980): 14-23. Jamison, D.T. and Lau, L.J. "Farmer Education and farm Efficiency: A Survey." Economic Development and Cultural Change. Vol. 29, No. 1, (Oct. 1980). 35-74 "Farmer education and Farm Efficiency. Baltimore: John Hopkins Uni-versity. Johnson,C.A.Jr;Johnson,M.B.; and Buse,R.C..Econometrics:Basic and Applied MacMillan Publishing Company, New York (1987). Kenya: Growth and Structural Development. Vol. II The World Bank, Washington, D.C. (1983) .Sessional Paper No.l on Economic Management for Renewed Growth. Govern-ment Printer Nairobi (1986). -.Statistical Abstracts Various Issues. Central Bureau of Statistics, Ministry of Planning and National Development. Khaldi, N. "Education and Allocative Efficiency in U.S. Agriculture." American Journal of Agricultural Economics. Vol. 57, (Nov 1975): 650-657 "The Productive Value of Education in The U.S. Agriculture" Ph.D Disser-tation, Southern Methodist University, Dallas, Texas January 1973. Bibliography 91 Kenya Long Range Planning Unit: Technical working papers, Various Issues. Min-istry of Planning and National Development, Nairobi. 1985-1988. Maitha,J.K. Coffee in The Kenyan Economy, An Econometric Analysis. East African Literature Bureau 1974. Moock, P.R. "Education and Technical Efficiency in Small-farm Production." Eco-nomic Development and Cultural change. Vol. 29, No. 4, (July 1981) 723-739. "Managerial Ability in Small Farm Production: An Analysis of Maize Yields in the Vihiga Division of Kenya." Ph.D. dissertation, Columbia University, 1973. Petzel, T.E. "Education and The Dynamics of Supply". Ph.D. dissertation, Univer-sity of Chicago, 1976. Pudasaini, S.P. "The Effects of Education in Agriculture: Evidence From Nepal." American Journal of Agricultural Economics. Vol. 65, (August 1983): 509-515. Shultz, T.W. "The value of Ability to Deal with Disequilibria." Journal of Economic Literature. Vol. 13, (September 1975): 827-846. Stefanon, S.E and Saxena, S. "Education, Experience, and allocative Efficiency: A Dual Approach." American Journal of agricultural Economics. Vol.70 No.2, (May 1988):338-345. Welch, F. "Measurement of the Quality of Schooling." American Economic Review, Papers and Proceedings. Vol.56, (May 1966): 379-392. ."Education in Production." Journal of Political Economy. Vol.78, No. 1, (Jan-uary/February 1970): 35-59. ."The Determinants of The return to schooling in Rural Farm Areas". Ph.D. Dissertation, University of Chicago, Chicago, Illinois. June 1966. Wolgin, J. M. "Resource Allocation and Risk: A case study of Smallholder Agricul-ture in Kenya." American Journal of Agricultural Economics. Vol. 57, No.4, (Novem-ber 1975): 624-630. Yotopoulos, P.A. and Nuget, J.B. Economics of Development: Empirical fnvestiga-tions. New York: Harper and Row, 1976. Appendix A A . l The Livestock and Crop Output Functions The difficult with estimating separate equations for crop and livestock enterprises arose because it was not possible to disaggregate some of the variables for each enterprise. Such variables like family farm labour supply and hired labour were only available for the entire farm activity and so estimating individual farm activities with these variables compromised the quality of the estimated coefficients. But it was quite obvious that most of these variables were more applicable to the crop sector more than the livestock sector. This is because after carefull data analysis it turned that livestock farming was not as an important activity as the crop farming. Furthermore, some of these variables which would obviously apply to both activities are more likely to be used in the crop sector. Family labour for example is more usefull in the more demanding crop farming than in the less labour demanding livestock sector. After all livestock in this region involves raising one or two cows for production of home consumed milk. The definatipn of livestock used here also involves certain forms of livestock like poultry which do not really need much labour. The other variable that equally belonged to the two sectors was the value of livestock. The other variables like purchased inputs, land and extension variables were easy to separate. However, as education entered the model as a decision variable there was no need to separate it, even if there was it would have been impossible to do so. Despite the above mentioned difficulties, these enterprise separate equations were estimated, particularly because there was need to pick up the separate effects of 92 Appendix A. 93 the farm purchased inputs, and also to pick up the effect of education on different farm enterprises. However, estimated coefficients for some of the variables like family labour should be interpreted with caution. These results favour the crop sector more than the livestock sector. Livestock Sector The livestock function was estimated using the aggregate value of livestock product sales as the dependent variable. The products sold include milk, hides and skins, eggs, and sales of live animals. The independent variables used include schooling of the farm operator, expenditure on livestock related variable inputs (purchased feed, expenditure on veternary services, dipping and other livestock costs), land and family labour time in mandays per year. It should be borne in mind that this function has a lot of loopholes due to inadequate data for the livestock activity. There was no price or quantity data for the particular products sold, neither were there data for the home consumed products. This means that we could not account for value of home consumed products which on average could be around 60% of the total output value. This percentage depends on the total value of output for each particular farm, for example those who produce small amounts of milk will consume a higher percentage of their milk output compared to those who produce large amounts of milk. This makes it difficult to adjust this sales figure for the home consumed value of output, consequently only the observed sales data was used. The results for the livestock sector are not very good, but this is not surprising given that the data available for the sales was not adequate. The results show that value of livestock and value of purchased inputs had positive and significant effect on livestock product sales. Education of the farm operator, family labour and extension service had insignificant effects on the value of livestock output. These results are not surprising because livestock is not the most important farm activity in the small scale farm sector in either Central Appendix A. 94 or Nyanza province. Most farmers are preoccupied with growing of crops for domestic consumption and commercial purposes. On the other hand what amounts to livestock is one or two heads of cattle usually tethered within the farmily farm or a few goats, sheep, pigs or a combination of any depending on the family. Poultry is widely kept but rarely for commercial purposes. As such it is not surprising that family labour, land acreage and even the educational variables are not important factors. For example raising of poultry or pigs requires very little land. Also having a large amount of land in this region is likely to induce farmers to grow more crops than keep extra cattle. The insignificant coefficient for family labour coefficient is not surprising as only a few hours by one individual are necessary for feeding livestock, thus increasing the number of family hours spent on raising livestock will have no effect on livestock products. Further more this variable is measured for the whole farm activity and not just the livestock sector. The coefficient for expenditure on purchased inputs was positive and significant at the 5This is not surprising because these inputs have a direct positive effect on the quality of livestock which affects the productivity of the livestock. The estimated coefficient for the expenditure on purchased inputs variable shows that farmers are maximizing the use of these inputs. This coefficient indicates that a shilling spend on purchased inputs has a marginal return of approximately a shilling. Crop Output Equations Total value of all the crops grown by the farmer was computed by summing the value of all crops produced in the family farm. There were several cases in which the price for particular crops for some farmers were not reported. To overcome this problem an average district price for the same product was used as a proxy for the price of that product. This was done with a clear understanding that there may be a bias due to transport costs. However, given that the regions where the sample data was obtained have relatively small districts, very little price variation was expected within the same Appendix A. 95 district for the same crop. This was also evident from the sample data. The survey sought information for twenty seven crop classifications and ofcourse no particular farmer grew' all the crops classified, infact most of the farmers reported about three to four different kinds of crops with very small outputs for each. The total value of crop production for each family farm was calculated as follows, Yi = Hi QaPij where Yj is the value of aggregate crop output for farm j, Qij is the total produce of crop i of farm j, and PtJ- is the price per unit of the same crop for family j. This value was regressed against various conventional and educational variables. The choice of the variables was based on the stated hypothesis, thus education of the farm operator and various forms of extension service were fitted in the equation along with, a variable which was a measure of expenditure on crop related purchased inputs (fertilizer, seed, pesticides and chemicals), other variables included are hired labour in mandays, family labour in mandaj's, land used for crops expressed as number of acres. Number of mature trees and value of farm machinery, equipment and buildings were also tried but the esti-mated coefficient for these variables were insignificant, these equations are not reported. Droping these two variables from the model increased the size of the adjusted coefficient of determination. The estimated schooling coefficient was positive and significant at a 5% two tail t-test, showing schooling increases farm productivity. In equation (1) where the level of farm purchased inputs are fitted separately the marginal return to schooling in the crop enterprise is estimated at Kshs.150. By aggregating the level of these farm purchased inputs the marginal return to schooling increases by about Kshs.46 capturing the return to education from allocation and selection of these purchased inputs. The estimated coefficient for this variable allowing extension interaction shows a marginal return to schooling of Kshs.222. The increase in crop output value arising from one extra year of Appendix A. 96 schooling. Both family and hired labour variables had positive and significant coefficients. How-ever, as expected hired labour coefficient was larger than the family labour coefficient. This shows hired labour has a higher benefit and opportunity cost compared to family labour. When the farm operator labour was specified separately from the family labour it had a larger coefficient than the family labour coefficient. As with hired labour, this indicates a higher opportunity cost for the farm operator time compared to the family aggregate labour. This coefficient was however still smaller than the hired labour coef-ficient (this equation is not reported). These results are not surprising as, apart from the family head and maybe the spouse in some cases, other family members usually have no alternative employment and their shadow price is very small compared to that of the family head or employed labour. The negative and significant coefficient for the extension variable indicates that there is a problem with extension services in the the small farm sector. This result could mean that extension officers only go to those farms which have problems or either the extension service per see confuses farmers or the way it is administered is not effective and so is a waste of time for farmers. In an earlier study by Moock (1973) it was found that client initiated contact had negative but insignificant effect on maize yields. This makes sense as in most cases only those farmers who experience problems will try to seek help, thus farms belonging to such farmers are already in problems. On the other hand if it is an extension officer initiated contact then we can expect a more reliable measure of the effect of extension on the farm output as long as he does not exercise bias when visiting farms, such as visiting only well to do farms or vise versa. In the present case the data does not explain the initiator of the contact. The demonstration coefficient was negative and significant at 20% level, showing that demonstration does not enhance crop production. This result was also found by Hopcraft Appendix A. 97 Table A.10: OLS Linear Regression Results For Crop(VCO) and Livestock (VLS) Sectors independent variable (l)(VCO) (2)(VCO) (3)(VCO) (4)(VLS) (5)(VLS) constant 1178.8 1020.7 943.33 199.74 206.32 (1.818) (1.558) (1.378) (1.561) (1.850) edl 150.04 195.93 222.34 29.840 27.416 (1.546) (1.986) (1.905) (1.109) (0.790) fml 2.034 1.831 1.836 -0.195 -0.195 (2.961) (2.645) (2.657) (0.992) (0.992) hrl 14.632 (2.718) 15.409 (2.763) 15.480 (2.775) Id 253.94 257.62 257.54 -8.563° -8.591° (3.723) (3.651) (3.669) (0.679) (0.685) piel 1.060 (3.907) 1.057 (3.848) piec 2.766 2.775 (fert+psd+ps) (5.679) (5.673) fert 4.293 (7.743) psd 1.135 (2.726) ps 2.942 (2.210) ext -1822.6 -1821.2 -1669.0 -277.80 -291.03 (2:647) (2.713) (2.421) (2.379) (3.068) demo -1180.8 -1118.8 -968.53 -156.47 -170.20 (2.010) (1.918) (1.515) (0.995) (1.123) barz 267.20 312.78 438.75 -152.47 -162.87 (0.268) (0.313) (0.456) (0.969) (1.180) edext -26.445 2.248 (Edl*Ext) R2 (-0.505) (0.222) 0.34 0.31 0.31 0.12 0.12 R2 0.34 0.32 0.32 0.13 0.13 F 38.73 42.93 38.16 12.88 11.44 N 748 748 748 673 673 a:- This denotes that this coefficient is estimated with land variable defined as grazing land Appendix A. 98 (1974) for tea production in Kenya, but it contradicts his finding for the same variable at the farm aggregate level (tea, maize and livestock output). Moock (1973) found crop demonstration had no effect on maize yields in Vihiga division of Kenya in some of equations while in one equation it was positive and effective at 10% level on two tail t-test. This result was not expected as demonstration is meant to teach farmers ways of increasing their farm yields and therefore should be reflected by a positive and significant coefficient. But the problem could be associated with the data, because it is not clear when farmers attended these demontrations and therefore we cannot say that attending these demonstrations led to poor performance. For example if a farmer attended the demostration when his crop was almost ready for harvest, this will not have significant effect on his output. It may also be possible that farmers attended the demonstrations because they were already having problems with their crops. The estimated coefnecient for the purchased inputs was positive and significant at 5% level two tail t-test in all the equations and it shows that, these inputs have positive contribution to crop output. This result shows that for every shilling spent on purchased inputs value of crop output increases by about Kshs.2.77. These results unlike those for livestock function show evidence of underutilization of purchased inputs. Farmers can increase their revenues substantially by increasing use of these purchased inputs. Looking at equation (1), it appears that fertilizer, pesticides and chemicals are mostly underutilized. There appears to be near optimal use of purchased seeds (psd). This ar-gument is validated by looking at the coefficients for these inputs. For example if farmers increased the level of fertilizer use by one shilling, they will increase the value of their crop output by approximately Kshs.4.30. This obviously indicates underutilization of this input. There are reasons however, as to why farmers may not optimize the use of these inputs, these are, risks involved in the sense that the farmer can only know the return to an input after he has harvested his product, but he pays for this input before Appendix A. 99 realizing the output. Thus, there is risk involved, and farmers being risk averse will be-have as predicted by these regression results. New technology which farmers may not be too familiar with is also embodied in some purchased inputs. As per Shultz hypothesis educated farmers will under such circumstances perform better than less educated farm-ers. Credit facility availability is also important in this case, thus lack of it may limit the farmers ability to acquire these inputs, hence leading to underutilization of the inputs. Also considering the fact that transport cost and costs of searching for information con-cerning these inputs is not included in the value of the inputs the estimated coefficients could be justified. In those equations where extension service was allowed to interact with education the coefficient for the interaction variable was negative, but the schooling coefficient was positive and bigger than in the other cases (see equation (2) and (3)). For example by introducing the schooling and extension interaction variable in equation (3) increased the schooling coefficient by 13.48%. Drawing upon this result and the positive correlation between EDI and Ext (see table 5.3), it appears that more educated farmers are some-what more likely to accept visits by extension officers, that their advice reduces output, and when this effcet is separeted out, the derived return to farmer education is increased, here. There is also a cost of renting land like missuse or lack of control of how the land is used by the person who rents it since there are no legal contracts stating how the land can be used once it has been rented. Appendix B B. l Land Use by Level of Education In the table (B.ll) farm land use by level of education of the farm operator is reported. There appears to be a uniform distribution of land ownership by level education, even though farmers with no education have the highest average land acreage the differences are not significant. On average farmers use most of their land for crop enterprise (approx-imately 52%) and only a small percent for animal grazing. Farmers with 8+ standards of education (atleast secondary education) devote the highest and smallest percentage of their land to crop and livestock grazing respectively. In terms of rent paid for rented land secondary educated farmers on average pay the least rent per acre. B.2 Yield levels by Type of Extension Service and Crop Adoption by Level of Education In table B.12 yield data by type of extension service attended is presented, this data shows that only those farmers who attended field day had higher yields than those who did not attend it. Farmers who attended baraza had lower yields than those who did not. It appears that barazas' are not an effective way of communicating extension service to farmers. Those who had extension contact as well as those who attended demonstrations produced less per acre than those who did not attend either of the two. These statistics tie well with results from regression analysis reported in chapter (5). Statistical tests show that these statistics are differen at 1% level two tail t-test. In table B.13 I present 100 Appendix B. 101 Table B.ll: Land Use by Level of Education no Educ. 1-3 4-7 8+ stds stds stds Average land acres (all land) 6.06 5.77 5.57 5.87 Average crop land 3.20 2.89o1 2.87cl 3.26a3 % of under crop) 52.8 50 51.3 55.6 Average grazing land 1.38 1.87a2 1.39c2 1.04a4 % under grazing 22.8 31.72 25 17.72 Land rent paid/acre 339 412.5 250 210 The following statistics are significantly different at 20% :(al and a2; a3 and a4; cl and c2 statistical analysis of the relationship between crop adoption and education of farm op-erator. These statistics show that a higher proportion of farmers with more years of schooling growing more hybrid maize than traditional maize compared to less educated farmers. However, when it comes to combination of coffee and local maize, and coffee and hybrid maize there does not seem to be any trend in combination of these crops on the basis of farmer education. The reasoning could be that coffee growing areas are some of the most developed areas in the country and even though some farmers may not have formal schooling, by virtue of their living in agricultural}' commercialized areas they have learned the benefits of doing commercial farming, and understand the returns from high yielding crops. For this reason they may be able to adopt high yield crops even though they have less or no education. Appendix B. 102 Table B.12 Value of Fara output/acre by Type of Extension Service Extension Baraza Demonstration Fieldday Contact attendance Attentance Value of crop 1761.OB 1335.58 1552.06 2082.49 output/acre (1914.081 (1894.07)(1886.781 (1861.26) Value ot livestock 47.05 71.00 67.04 148.06 product, sales/acre (82.15! (72.25) (72.51) (71.35) Total value of 1118.89 668.66 957.2 1233.62 Fare ouput/acre (1023.08) (1076.42!(1056.63) (1048.34) (n«226» !n=44) (n=571 (n=6) ** in parenthesis are y ie ld levels for those who did net receive the specified extension service : - a l l the above are s ign i f ican t ly different at IX level t - tes t . Table S.-3 DISTRIBUTION OF CROP AD0PTI0K BY LEVEL DF EDUCATION (coffee, local and hybrid a3ize) Total no. c f people f of stds. A i l threeCcffee k Coffee & Hdsaize iCoffee local hybrid in th is caaplsted. Above Loctaize Hdmaizs Locsaize saize aaizs category 8+ stds. 2 4 22 0 0 20 30 78 ll of this 2.6 5.1 28.2 0 0 25.6 38.5 edu. catg.) 4 - 7 stds. 2 8 34 8 4 44 72 172 (I of this 1.2 4.7 19.3 4.7 2.3 25.6 41.9 edu.catg.) 1-3 stds. 0 1 2 ! 1 0 26 35 (1 of this 0 1.2 25 1.2 0 31 37.2 84 edu. catg.) no schooling 5 15 70 9 8 134' !15 356 '1 of this 1.4 4:2 1-9.7 2.5 2.2 37.6 32.3 edu. catg.) total 9 28 147 18 12 224 ' 252 690 (J of total) 1.3 4.1 21.3 2.6 1.7 32.5 36.5 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0098394/manifest

Comment

Related Items