5.5.2 Body Mass Index <strong>of</strong> mothersThe Body Mass Index (BMI) is a measurement <strong>of</strong> the nutritional status that is based on heightand weight (BMI = weight/height 2 ). It is used to compare and determ<strong>in</strong>e the health effects <strong>of</strong>body weight on human be<strong>in</strong>gs. A BMI score between 22 and 24 is considered normal. Belowthe lower limit, the <strong>in</strong>dividual is underweight; and above the upper limit, the <strong>in</strong>dividual isoverweight or obese.For western <strong>Kenya</strong>, the average BMI score <strong>in</strong>dicates a normal situation for both Nyanza and<strong>Western</strong> prov<strong>in</strong>ces (Table 5.10). However, a breakdown <strong>in</strong>to BMI classes <strong>in</strong>dicates that only46% <strong>of</strong> mothers were <strong>in</strong> a normal category <strong>of</strong> BMI, 36% were underweight and about 18%were overweight <strong>in</strong> Nyanza. In <strong>Western</strong> prov<strong>in</strong>ce, 61% <strong>of</strong> mothers were <strong>in</strong> a normal category<strong>of</strong> BMI, 27% were underweight and about 12% were overweight. The results show that moremothers were undernourished <strong>in</strong> Nyanza than <strong>in</strong> <strong>Western</strong> prov<strong>in</strong>ce.Table 5.10. Body Mass Index and BMI group<strong>in</strong>gs <strong>of</strong> mothersNyanza<strong>Western</strong>N 364 461BMI (average) 21.54 21.9Normal (%) 45.9 60.7Underweight (%) 36.3 27.3Overweight (%) 17.9 11.9N = Number <strong>of</strong> mothers40
Chapter 6Empirical Analysis <strong>of</strong> <strong>Maize</strong> Production Efficiency <strong>in</strong> <strong>Western</strong> <strong>Kenya</strong>6.1 Stochastic production frontier estimatesThe maximum likelihood estimates <strong>of</strong> the parameters <strong>of</strong> the translog stochastic frontier and<strong>in</strong>efficiency model were obta<strong>in</strong>ed <strong>us<strong>in</strong>g</strong> FRONTIER 4.1 (Coelli, 1996) (Table 6.1). Thechoice <strong>of</strong> the empirical frontier production function was made based on the generalisedlikelihood ratio test follow<strong>in</strong>g Coelli and Battese (1996), without hav<strong>in</strong>g to impose anyfunctional form. The functional form <strong>of</strong> the stochastic production frontier and <strong>in</strong>efficiencymodel was thus determ<strong>in</strong>ed by test<strong>in</strong>g the adequacy <strong>of</strong> the Cobb-Douglas model aga<strong>in</strong>st themore flexible translog model. The null hypothesis that the Cobb-Douglas model is anappropriate representation <strong>of</strong> the data, given the specifications <strong>of</strong> the translog, was highlyrejected (Table 6.2), <strong>in</strong>dicat<strong>in</strong>g that the Cobb-Douglas model was actually not appropriate.Therefore, the translog stochastic frontier and <strong>in</strong>efficiency model was used as the preferredmodel for the analysis <strong>of</strong> maize production efficiency <strong>in</strong> western <strong>Kenya</strong>.The results <strong>in</strong>dicate that maize output <strong>in</strong>creases with land, labour, fertiliser and oxen power.The sizes <strong>of</strong> the elasticities <strong>of</strong> frontier output with respect to the <strong>in</strong>puts show the relativeimportance <strong>of</strong> the various factors <strong>of</strong> maize production. In this regard, maize output is mostresponsive to land and least responsive to fertiliser. The low production elasticity <strong>of</strong> fertiliserconfirms the observation that only 15% <strong>of</strong> the sample farmers used fertiliser. Moreover, most<strong>of</strong> these farmers (70%) applied fertiliser below recommended rates. Low adoption and<strong>in</strong>tensity <strong>of</strong> use <strong>of</strong> fertiliser could be associated with the <strong>in</strong>creas<strong>in</strong>g prices <strong>of</strong> fertiliser relativeto maize. On the other hand, the high production elasticity <strong>of</strong> land is consistent with the factthat land is a critical factor <strong>of</strong> production for smallholders, especially <strong>in</strong> western <strong>Kenya</strong> wherethere is a grow<strong>in</strong>g population pressure on land.The coefficients <strong>of</strong> the variables represent<strong>in</strong>g adoption <strong>of</strong> hybrid maize and <strong>Striga</strong> controltechnologies are both positive and significant, <strong>in</strong>dicat<strong>in</strong>g that adoption <strong>of</strong> these technologies<strong>in</strong>creased frontier maize output. That is, adoption <strong>of</strong> hybrid maize and <strong>Striga</strong> control practices<strong>in</strong>creased maize production among the efficient farmers. The positive effect <strong>of</strong> hybrid maizeadoption is consistent with the survey result that adopters obta<strong>in</strong>ed average yields <strong>of</strong> about1.3t/ha compared with non-adopters who obta<strong>in</strong>ed only 0.8t/ha. Similarly, the positive<strong>in</strong>fluence <strong>of</strong> <strong>Striga</strong> control on frontier maize output is consistent with the result that maizefarmers who used <strong>in</strong>tegrated <strong>Striga</strong> control practices obta<strong>in</strong>ed average maize yields over 1t/hacompared with the average 0.8t/ha for farmers who did not adopt this practice. Indeed, <strong>in</strong>western <strong>Kenya</strong> where <strong>Striga</strong> damage on maize yield could be as high as 75–100%, themarg<strong>in</strong>al benefit <strong>of</strong> a control practice is high. The coefficient <strong>of</strong> the household headshipvariable is positive and significant at 10% level, <strong>in</strong>dicat<strong>in</strong>g that male headed households hadhigher frontier maize output than female headed households. A separate model, constructedby <strong>in</strong>teract<strong>in</strong>g gender with all other variables, was estimated to test whether the sametechnology was available to male and female heads <strong>of</strong> households. None <strong>of</strong> the <strong>in</strong>teractionterms was significant, and a generalised likelihood ratio test could not reject the nullhypothesis that the <strong>in</strong>teraction terms are jo<strong>in</strong>tly equal to zero. This demonstrates that male andfemale headed households actually have a homogenous production technology, imply<strong>in</strong>g thatif their respective frontiers were to be estimated separately, they would have the same S-curveshape usually found <strong>in</strong> adoption studies <strong>of</strong> new crop varieities. This is consistent, for <strong>in</strong>stance,with the observation that both female headed (12%) and male headed (20%) householdsadopted hybrid maize. Higher frontier output among male headed households implies an<strong>in</strong>tercept shift <strong>in</strong> the frontier and hence a change <strong>in</strong> its placement. Significant genderdifferentials <strong>in</strong> maize output were found from a separate conventional production function41