The economy in western Kenya, like most of the other regions in Kenya is agriculture based with smallholder farm households forming the bulk of the population. While all smallholder households engage in agricultural production to meet their food and cash needs, income earned outside the farm forms a significant component of household income. For these households, labour is the main input in both farm and off-farm activities. This study was motivated by three reasons: Firstly, there are contradicting perceptions regarding availability of labour in smallholder farm households of western Kenya. One view point suggests abundant labour due to the high population density and few employment opportunities. On the other hand, poor uptake of labour intensive agricultural technologies is frequently attributed to lack of labour and/or poor returns thereby suggesting scarcity of labour in the households. Secondly, previous studies mainly focus on the pros and cons of diversification in terms of employment, income effects, consumption effect or lost labour effect. Few however, have studied the efficiency or outcomes of household behaviour given the wide portfolio of economic activities they engage in. Thirdly, there are significant differences in outcomes between livelihood strategies and significant gains to be made by shifting between strategies. Labour has desirable characteristics, such as its versatility, divisibility and mobility, which make it attractive for technical and policy interventions targeting poor households. However, labour as an entry point for change may become an ineffective where its mobility is constrained. The aim of this thesis was to identify barriers that hinder farm households from benefiting from interventions targeting them. Efficiency studies provide an indication of how well households use their resources and the mobility of factors of production in the household. Where households are efficient, it means that factors of production easily move to activities with highest returns. The aim of this study was therefore achieved by establishing whether farm households are efficient in allocation of labour and by identifying the factors responsible for deviations from expected behaviour. The study contributes to the AfricaNUANCES project “Exploring tradeoffs around farming livelihoods and the environment” by illuminating the dynamics which lead to varied perceptions about the opportunity cost of labour and its availability in smallholder farm households. Most studies on allocative efficiency are mainly carried out at the household level with the assumption that allocative efficiency prevails within the household. Such an assumption although convenient ignores the effects on labour use of; seasonality, heterogeneity within farms and the uncertainty in uncontrolled environments. Where these situations prevail, within household efficiency cannot be assumed. For example, farm households normally sub-divide their land into independent units either due to the heterogeneity in biophysical conditions and/or the wide range of crops grown so as to meet food and cash needs. When agricultural production is simultaneously carried out in such semi-autonomous units, allocative efficiency may be at stake. The second situation where within household allocative efficiency may be questioned is where household resources are allocated on the basis of gender. The third situation where within household allocative efficiency cannot be assumed is where there is strong seasonality in production activities or where there is uncertainty because agricultural production occurs in uncontrolled environments. In such cases where labour demand varies over the season, efficiency in labour allocation cannot be assumed. The assumption of most behavioural studies is that households are faced with a single price in decision making which under perfect markets is the market wage rate. There are situations however where households may experience more than one wage rate for their labour. For example, where households cannot sell as much labour as they would wish because of rationing in the labour market or where financial or insurance markets have failed. In this case, labour allocation between farm and off-farm activities is influenced by the prevailing conditions on the farm as well as off the farm. A holistic approach to allocative efficiency was adopted by addressing three research questions namely: whether farm households are efficient within their farms i.e. whether marginal reallocation of labour would lead to better outcomes on the farm, whether farm households are efficient over the cropping season and lastly whether farm households are efficient in labour allocation between farm and off-farm activities i.e. whether their involvement in labour markets affects the marginal efficiency of labour use. The study combined one and two-step production functions, and labour supply and demand functions in addressing these questions. In Chapter 2, the data used and the characteristics of the study area are described. In addition the diversity of household strategies in terms of labour use and incomes sources is described and comparisons made to highlight the relative importance. The dataset used comprises of household and plot level data collected from a random sample of farm households in Kakamega, and Vihiga districts of western Kenya. Like other parts of Kenya, the economy in western Kenya is mainly driven by an agricultural sector which mainly comprises of smallholder farmers. Much of the land in the two districts (Kakamega, and Vihiga) falls in the high to medium potential areas where the rainfall ranges from 1400mm to 2000mm. Farms are on average small. The average farm size is 0.97 ha and 0.58 ha for Kakamega and Shinyalu respectively and fifty percent (50%) of the households own not more than 0.81 ha in Kakamega and 0.5 ha in Vihiga. It is a common practice for smallholder farmers to subdivide their farms into distinct plots. The plots are normal small with a mean of 0.24 hectares in Kakamega and 0.16 hectares in Vihiga. In spite of the small farm sizes, farming is the main economic activity for households in the study area and the family is the main source of labour. Nutrient mining, soil erosion and little use of fertilisers characterise the farming landscape in the two districts. Consequently, majority of the households experience food deficits and earn little from their farms. Household income is normally supplemented with cash from sale of livestock products and live animals, off-farm employment and remittances. Majority of the households have at least one member working off-farm. The poverty incidence in western Kenya is however high with 57% falling below the poverty line. In Chapter 3, a horizontal view point of labour use was adopted whereby within farm allocative efficiency was examined. The total number of distinct plots for each household range from one to six with most households having two to four plots. The outcomes of household behaviour in labour allocation to various crops was examined while controlling for variability in bio-physical characteristics of the farm. An index indicating the extent of inefficiency within the farm was created from model parameters and the factors influencing this inefficiency determined. The study reveals that there is room for improvement in efficiency of labour allocation within the farm. It shows that farm households tend to allocate more labour to food crops like maize and beans compared with other crops that are more market oriented. The labour applied to maize relative to other crops increases with the household’s labour capacity and with remoteness but reduces when the liquidity position of a household improves. These results may be explained by the different perception of output prices by the household when there are imperfections in output markets. A vertical perspective of labour use was adopted in Chapter 4. Seasonality and uncertainties in agricultural production result in a varying labour constraint over the season. A model which captures the sequential nature of farm household decisions over the cropping season was adopted in deriving optimal levels of stage-specific labour input and output. The resultant econometric model of two labour functions and a production function was applied to maize production and household’s allocative efficiency of labour use over the season evaluated. Efficiency was evaluated by comparing returns to labour applied at the beginning of the season and at weeding. For over 90% of the farm households the marginal product of labour applied at the beginning of the season was larger than the marginal product of labour applied at weeding. This suggests that households in western Kenya are most labour constrained at the beginning of the season. There are a few households that are more labour constrained at weeding or maintain constant returns to labour across the season. The results are suggestive of imperfections in the labour market. An index of inefficiency was created and related to various farm and household characteristics. Simulations using model coefficients show how the model results can be interpreted for farm households with different characteristics. The change in the relative value of labour (due to changing farm and household characterstics) was expressed in monetary terms to demonstrate the application of model results to real life situations. In Chapter 5 we examine household efficiency in labour allocation between farm and off-farm activities and the determinants of labour supply and demand. Efficiency was evaluated by comparing information generated in chapter 3 regarding returns to labour within the farm with information generated in chapter 5 regarding returns to labour off the farm. In determination of factors influencing household behaviour, we did not choose a particular wage but allowed the household to demonstrate how the payments to labour on the farm, off-farm and village wage rates influence its decisions. On average differences in returns to labour between farm and off-farm employment are large and significant implying that efficiency in use of family labour can be improved. However, households which participate in labour markets as sellers or buyers are more productive on the farm and efficient in use of family labour on the farm. There is unavailability or rationing in off-farm jobs for some categories of skilled labour. Distance to market is an important determinant in the decision to sell labour off-farm. We simulated the expected farm household response to changes of 10% in the key factors influencing supply and demand namely; farm size, family size, income and education level. Increasing household income results in increased consumption of leisure as households reduce farm labour supply by 6.8%. This reduction in farm labour supply is however compensated with an increase (6.3%) in demand for hired labour. Households with a higher labour capacity supply more labour on-farm (3.7%) and off-farm (5.8%). Such households generally demand less hired labour (2.7%) with the exception of those that sell labour off-farm where a 2.5% increase in demand for hired labour was observed. The largest increase in supply of labour to the farm may be achieved through an increase in farm size which results in 3.7% increase in supply of labour and a 2.4% increase in demand for hired labour. Moving from complete illiteracy to a level where the head has some basic skills resulted in reduced off-farm labour supply probably because there are few opportunities for persons with basic skills compared to opportunities available for illiterate persons. Off-farm employment opportunities are least available for persons with secondary and above secondary education because moving from just primary education to secondary or above results in a big drop in labour supplied off-farm. The simulations also showed that when off-farm opportunities are unavailable, more labour is employed on the farm. In Chapter 6 we conclude the study by revisiting the key issues addressed and discussing the key findings. Household inefficiency in labour allocation has been attributed to market imperfections particularly output and labour markets. These imperfections hinder labour mobility within the farm, across the season and between farm and off-farm activities. Policy measures that would steer farm households towards more efficient use of labour include reducing marketing costs for farm output, increasing participation in labour market and improving the functioning of other rural markets. This study quantified the degree of efficiency of the allocation of labour in the household. It did so for the allocation over crops, over time, and between on and off-farm employment. In this way the study contributes new measures of the effects that the varying degrees of integration in the markets have on the allocation of the household's most important production factor, their labour.