0096-B1

Socio-Economic Factors Affecting Farmers' Decisions to Adopt Farm Forestry: An Application of Multivariate Logistic Analysis in Busia District, Kenya

M.K. Lwayo^[1] and H.K. Maritim

ABSTRACT

The high population in Busia has led to an increased demand for tree products in the district resulting in high deforestation. The district does not have government maintained forests. Farmers therefore depend on trees on their farms and/or on their neighbours' farms for fuelwood and poles. Timber is imported from the neighbouring districts of Kakamega and Bungoma. The art of planting and harvesting trees for home consumption has not been adopted in the district to a level that could make households self-reliant. The situation is aggravated by the fact that most farmers in Busia engage in small-scale farming and the majority in subsistence production. As a result, farm forestry is given little attention (Lwayo, 1999).

This study aims to identify various socio-economic variables that affect the farmers' decisions to adopt farm forestry in Busia. It considers the policy framework that needs to be put in place to stimulate the adoption of farm forestry technologies. The study area was divided into five clusters and using households as sampling units, 40 households were selected per cluster using the simple random sampling technique. A total of 200 respondents were interviewed using structured questionnaires. A logit model was employed to identify the main factors influencing farm forestry adoption by households on farms surveyed in the district. The SAS computer package was used to derive the maximum likelihood estimates of the adoption process and to calculate the chi-square. The results of the logit analysis showed that education, age and land size were significant at 0.05 level of significance. Formal education is vital and farmers should be educated about the need to promote farm forestry. Attention should be focused on farmers over 55 years old who are mainly the decision-makers in most households. The role of extension in promoting farm forestry should not be played down especially as concerns providing technical advice. Small-scale farmers should be encouraged to commercialise farm forestry so as to diversify their source of income.

INTRODUCTION

Public concern over the environment has brought forests into the public eye as an environment-friendly alternative land use in agriculture (Lloyd et al., 1995). Obstacles to investment in farm forestry include the long payback horizon, fear of trespass, cost of conversion, vandalism and psychological inertia (we're farmers not foresters). However in the current climate of agriculture reform, public and political concern over the environment, environment-friendly alternatives to intensive agriculture are being sought. Forestry is one such alternative and commands support from many of the parties involved with the notable exception of farmers (Watkins et al., 1995). Survey evidence suggests that farmers are not wholly and irretrievably opposed to forestry. Dis-inclination to tree planting is created by a variety of factors. There is evidence to suggest that permanent loss of flexibility of the farm through felling restrictions is an important component of the problems (Saxena, 1992). Traditional rural societies in Kenya always depended on locally available trees for their domestic uses such as fuelwood, poles, shade and medicine (Foley and Benard 1984). As the country developed and the population increased, people's demand for these tree uses increased rapidly. The standing stock of trees in Kenya is declining fast as tree exploitation is estimated to be 45 percent higher than sustainable yields (Hankins, 1987)

Trees are an important asset for any farmer. They can be used as boundaries whilst also supplying fodder for animals and fuelwood for the homes. Leaves from the trees can be applied as mulch. Leguminous tree species can be intercropped with crops to improve soil fertility. Others are used for medicinal and cultural purposes. The high population in Busia district has led to increased demand of tree products in the district, which in turn has led to the high deforestation. The district does not have government maintained forests. Farmers therefore have to depend on the trees on their farm and/or neighbour farms for fuelwood and poles. Timber is imported from the neighbouring districts of kakamega and Bungoma. Inspite of this shortage, a few farmers have adopted farm forestry to a level that they can be self-sufficient. The few adopters have regenerated several species on their farms and maintained those that regenerate naturally such as Markhamia Lutea (Maina,1994).

The paper presents results of an on-farm survey conducted in 1997/98 in the six divisions of Busia. The survey focused on the socio-economic variables of age, sex, education, land size, disposable income and off- farm income and how they affect the farmer's decision to adopt farm forestry in the district. The aim of the study was to develop and safeguard a natural resource base. The fundamental issues discussed in the paper are; the threshold disposable income that can trigger the adoption of farm forestry in Busia district, the implication of age in the decision making mechanism with regard to adoption of farm forestry as a farming technique, the influence of the farmers sex on the farmer's decision to adopt, at what level is the size of the land crucial in growing trees, the willingness of the farmer to allocate part of his land resource for farm forestry technologies and the impact of non-farm income on adoption of farm forestry as a farming technique, the policy framework that needs to be put in place to stimulate the adoption of farm forestry technologies. These and other related issues form the basis of this study.

METHODOLOGY

Secondary and primary data was collected and used in the analysis. Secondary data was collected from existing and relevant literature and publications. Vital information was also collected from the Ministry of Agriculture, Ministry of Environment and Natural Resources and Kenya Woodfuel Afforestation Programme (KWAP). The study area was divided into 5 clusters. Using households as sampling units, 40 households were selected per cluster using the simple random sampling technique. This was to give each household an equal likely chance of being selected and avoid any bias that may arise otherwise. Both adopters (farmers currently managing farm forestry) and non- adopters (farmers who have interacted with adopters but decided not to adopt farm forestry, dropped it after trying or never tried at all) were interviewed, using a structured questionnaire that had lead and open ended questions. Two analytical approaches were used. The chi-square and logit model.

Chi-square Analysis

The status of the respondents level of adoption (adopters or non-adopters) of farm forestry was classified in groups and with respect to each socio-economic variable, a contingency table was drawn up. The chi-square statistic was used to analyse the contingency table data. The formula is given below;

Where;
c² = Chi-square
f_e = Expected frequency
f_o = Observed frequency

The use of chi-square helps to decide whether two variables independent or dependent are related in a population. The test also determines if a conspicuous discrepancy exists between the observed and expected counts. It was employed in the analysis to test whether the explanatory variables were related among the adopters and non-adopters.

Logit Model

For this study in Busia, the logit model was used because it reflected the empirically observed status of farm forestry on any particular farm. Such observations reflect a dichotomous variable, adoption or non-adoption. This 'adoption behavioural model' with dichotomous (or binary) dependent variables can be used as a conceptual framework to examine variables associated with the adoption of technology. Although least square estimates can be computed binary models, the error terms are likely to be heteroscedastic leading to inefficient parameter estimates; thus classical hypothesis tests, such as the t-ratios are inappropriate (Pindyck and Rubinfeld, 1981). The application of the conventional OLS techniques result in bias by over estimation and inconsistency (Maddala, 1983,p.2) and it has been shown both theoretically and empirically that a logit or tobit analysis is more appropriate in such cases (Maddala, 1983, p149-194). The use of logit, which gives the maximum likelihood estimates, overcome most of the problems associated with linear probability models and provides estimators that are asymptotically consistent, efficient and gaussian so that the analogue of the regression t-test can be applied. The logit model based on the cumulative logistic probability function, is computationally easier to use than the probit and tobit models and was used in this study (Pindyck and Rubinfeld, 1981, p.311 and p.287).

Conceptually, the following is the general adoption behavioural model used to examine the factors influencing the farmer's decision to adopt farm forestry;

Where:

P_i = The probability that an individual will adopt a given resource base
(the binary variable, P_i = 1 for an adopter and P_i = 0 for an non-adopter)

Z_i = Estimated variable or index for the i^th observation

F = The functional relationship between P_i and Z_i

i = 1,2 ...,m are observations on variables for the adoption model.
They are defined in Table 1 for this analysis, m being the sample size 200

X_ji = The j^th explanatory variable for the i^th observation, j = 1,2 ... n

bj = A parameter, j = 0,1 ......n

j = 0,1......,n where n is the total number of explanatory variables.

The logit model assumes the underlying index, Zi is a random variable that predicts the probability of the farmer's decision to adopt farm forestry;

Therefore:

This is the logit model (Engelman, 1981 and Gujarati, 1988)

Model Specification

For this study, the stimulus index Zi was determined as a linear function of the explanatory variable summarized in the Table1.

Table 1: Explanatory and Corresponding Binary Variables for Adoption of Farm Forestry in Busia District 1997/98

MODEL ESTIMATION

Social Factors that influence the Decision to Adopt Farm Forestry in Busia District

Table 2: Analysis of Maximum Likelihood Estimates for Adoption

* Significant at 0.05 level of significance

The age of the member of the household who manages the farm indicates their capacity to work. It also affects ones ability to adopt innovations and changes. The maximum likelihood analysis results (Table 2) showed a positive relationship between age and the decision to adopt farm forestry. This indicates that age influences the farmers' decision to adopt. The age of the farmer affected the farmer's knowledge and the awareness of the activities in the surrounding environment among other farmers. Analysis of the data using chi- square showed that c² = 17.410, which was statistically significant at 0.05 level of significance.

The results of the maximum likelihood analysis in table 2 showed that there was a non-significant positive relationship between sex and the decision to adopt farm forestry showing that males are not necessarily better adopters than females. Analysis using chi-square gave c²= 0.028, which was statistically non-significant at 0.05 level of significance. Sex is thus not a critical issue in a farmer's decision to adopt farm forestry. A significant difference was found between the level of literacy among adopters and non-adopters at 0.05 level of significance. The logit model indicated a positive significant relationship between adoption of farm forestry and education. This accords with Oram (1988) who showed formal education as a vital aspect in the farmer's decision to adopt farm forestry and the fact that literate farmers would be adopters. Formal education would therefore be a critical factor in influencing the effectiveness of the farmer's participation in farm forestry. Chi-square c²= 6.05 indicating education is statistically significant at 0.05 level of significance. An educated farmer can readily access information on the value of farm forestry and how it can be effectively implemented.

Economic Factors that influence the Decision to Adopt Farm Forestry in Busia District.

Table 3: Analysis of Maximum Likelihood Estimates for Adoption.

* Significant at 0.05 level of significance

The logit model in table 3 showed a significant relationship between land size and the farmer's decision to adopt farm forestry. Chi-square c²= 25.192. Land size is an indicator of the available economic resources and the willingness to adopt a new technology. It revolves around factors such as the risk, preference, capital constraints, labour requirement and the tenurial arrangements (Arnold, 1990). In agricultural zones, tree crops compete with cash crops with the latter being preferred. Farmers is high potential areas are therefore unwilling to divert land available for food and cash crops to trees which do not generate an equally lucrative product.

Non farm income incorporates income earned by the household from different sources other than the farm. It was apparent that non-farm income sources varied greatly. This included trade, employment, casual work, credit, relatives, friends and miscellaneous sources. The logit model showed that non-farm income was non-significant at 0.05 level of significance. Chi-square c² = 0.44. Thus the off farm income earned by he household did not affect the farmers ability to adopt farm forestry. This is because its investment is low cost.

Disposable income is the income that is left to the household to spend after taxation. It encompasses money accrued from different sources and used as expenditure for the household and savings. Judging from the logit coefficient, the household's level of income is a pre-disposable factor. It is not critical in the decision-making framework. Statistically, the decision to adopt is not based on the income level. This is attributed to the fact that tree seedlings are cheap and in other instances the farmers are given the seedlings free by organizations trying to promote farm forestry in the area.

Since majority (70%) of the households in Busia fall in the low income category of less than Kshs 60000 per annum then, for farm forestry to be adopted, this is the income group that needs to be targeted. This is also the threshold disposable income that can trigger the adoption of farm forestry in the district. This is arrived at by the fact that farm forestry is a low cost investment. The low price of seedlings and readily available family labour makes the low income group (less than Kshs 600000 per annum) the target group. This low-income group comprises the impoverished lot and their meager earnings cannot support expensive technologies. They are resource poor farmers most of them having less than three acres of land thus justifying the income level of less than Kshs 600000 per annum the threshold income.

Of the respondents interviewed, 77.5% of them were adopters and 22.5% were not. The government working hand in hand with interested Non Governmental Organizations should put in place a clear policy that emphasizes on the need to promote farm forestry within a view of alleviating general poverty. Each of the socio-economic variables studied should be addressed at levels in which it affects the farmer's decision to adopt farm forestry. The policy implementation should be concentrated at the district level to bring it closer to the people. Promotion of farm forestry will help to reduce the imbalance in the market of timber and poles and make the marketing of the product efficient. Formal education is vital in promoting farm forestry in the area through educating farmers on its importance and the risk of deforestation. Attention should be focussed on farmers over 55 Years who are mainly the decision-makers in most households and conservatives in technology adoption. Small-scale farmers should be encouraged to grow more trees and to commercialise this investment so as to diversify on their source of income.

ACKNOWLEDGEMENTS

I would like to express my deep appreciation and gratitude to Moi University for awarding me the partial scholarship that enabled me carry out this research and compile a thesis, from which this paper was extracted. I am greatly indebted to Prof. H.K. Maritim, the co-author for his intellectual contribution, guidance and commitment to this paper. I acknowledge his collaborative effort in its write up.

REFERENCES

Arnold, J.E.,1991. Tree Components in farming Systems. 41:35. FAO Paper. International Journal of Forestry and Forestry Industries. IUCN. Unasylva. Gland Switzerland.

Engelman, L.,1981. PLR: Stepwise Logistic Regression. BMDP Statistical Software 1981. (Ed. Dixon, W.J.), Berkely: University of California Press, pp.330-346.

Foley.G and Benard,G.,1984. Farm and Community Forestry. Technical Report No.3 Earthscan. 236pp.

Gujarati, D.N.,1988. Basic Econometrics. McGraw-Hill Book Co.-Singapore (2nd Edition.).

Hankins, M.,1987. Renewable Energy in Kenya. A Paper Presented at the American Association for the Advancement of Science Annual Meeting, Washington D.C 26pp.

LIoyd, T., Watkins,C and Williams,D.,1995. Turnings Farmers into Foresters Via market Liberalization; Journal of Agricultural Economics (46)(3) 361-370.

Lwayo, M.K.,1999. The Effect of Socio-economic Factors on the Adoption of Farm Forestry as A resource. A Case Study of Busia District, Kenya. An MPhil Thesis, Moi University, Faculty of Agriculture.

Maddala,G.S.,1983. Limited-Dependent and Qualitative Variables in Econometrics, Cambridge University Press, New York.

Maina, R.,1994. A Socio-cultural; Survey of Agroforestry in Tesoland, Busia, Kenya; A Thesis Submitted in Partial Fulfillment of the Award of a Diploma in Environment Studies, Kenya Polytechnic, Nairobi.

Oram, P.A.,1988. Moving Towards Sustainability; Building the Agroecological Framework: In Environment (30)(9)

Pindyck, R. S and Rubinfeld, D. L.,1981. Econometric Models and Economic Forecasts. 2^nd Edition; London: McGraw Hill.

Saxena, N. C.,1992. Adoption of Long Gestation Crop: Eucalyptus Growers in North West India; Journal of Agricultural Economics (43)(2) 257-267.