Income and Gender Pathways

The positive effects of commercialization on the consumption of calories and micronutrients from purchased foods suggest that the cash income pathway plays an important role. This is now analyzed more explicitly in table 3.7. The first column in table 3.7 reveals a significantly positive association between the level of commercialization and household income.

Controlling for other factors, a 10 percentage point rise in the level of commercialization is associated with almost 25 thousand Ksh higher income (27% of mean household income of the least commercialized households). The other columns in table 3.7 confirm that gains in household income are significantly associated with higher calorie and micronutrient consumption. Only for vitamin A, the association is not statistically significant.

62

Table 3.7. Commercialization, Household Income, and Calorie and Nutrient Consumption

Variables Household

Male household head (dummy) 33.086^** -91.857 93.278 -3.401^*** 0.433

(14.531) (106.832) (116.864) (0.830) (1.109) Education of household head (years) 6.421^*** 34.434^*** 11.898 0.173^** 0.434^***

(1.830) (10.832) (13.681) (0.080) (0.156)

Poor agroecology (dummy) 16.435 86.027 74.530 0.650 -0.740

(18.644) (111.290) (168.858) (0.953) (1.540)

Constant -227.919^*** 3780.066^*** 1030.677^** 26.198^*** 24.435^***

(38.844) (204.770) (352.596) (1.506) (3.170)

Sub-county dummies Yes Yes Yes Yes Yes

Observations 805 805 805 805 805

R-squared 0.384 0.297 0.040 0.183 0.139

Note: Coefficient estimates are shown with robust standard errors clustered at farmer group level in parentheses. All models were estimated with ordinary least squares. AE, male adult equivalent; RE, retinol equivalent; Ksh, Kenyan shillings. ^*, ^**, and ^*** significant at 10%, 5%, and 1% level, respectively.

To evaluate possible effects of commercialization on gender roles, we look at who within the household controls the revenues from farm output sales. Most households sell different crops, for which the control of revenues can vary. Revenues from cash crops are often controlled by men, whereas for food crops the situation is more diverse (Fischer and Qaim 2012a). Hence, calculation of a single variable that captures gendered revenue control across households and crops is not straightforward. For this part of the analysis, we decided to focus on two of the most important food crops in the study region, namely maize and beans. Most of the sample households grow these crops primarily for home consumption, 25-30% of the households also sell some maize and beans to generate cash income. We only focus on the subsample of households that sold some of their maize and beans during the 12-months period covered by the survey.

63

For both crops, the question “who controls the revenues?” was asked with three possible and mutually exclusive answers, namely “male control”, “female control”, or “joint control”. In all cases, the answers were clear and straightforward, regardless of whether the respondent was male or female. Based on these data, we constructed separate dummy variables for both crops that take a value of one if a male household member controls the revenues alone, and zero if a female member controls the revenues either alone or jointly with a male member.

This “male control” dummy takes a value of one in 23% of the households for maize, and in 17% of the households for beans.

Table 3.8 presents estimation results of models with this “male control” dummy as dependent variable for the case of maize (for beans, the models are presented in table A3.15 in the appendix). Two specifications are shown, a linear probability and a probit model. Both specifications lead to similar results. The level of commercialization is positively and significantly associated with male control of revenues in those households that sell at least some of their maize (beans). This is consistent with earlier research showing that commercialization can be associated with women losing control of how to use crop harvest and income (von Braun and Kennedy 1994; Chege et al. 2015).

Table 3.8. Association between Maize Commercialization and Male Control of Maize Revenue Linear probability model Probit model

Male controls maize revenue

Variables Male controls maize revenue

Coefficients Coefficients Marginal effects

Maize commercialization (0-1) 0.326^*** 1.312^*** 0.365^***

(0.112) (0.451) (0.122)

Age of household head (years) -0.002 -0.009 -0.002

(0.002) (0.009) (0.003)

Male household head (dummy) 0.720^*** 6.861^*** 0.684^***

(0.143) (0.554) (0.041)

Education of household head (years) -0.012 -0.066 -0.018

(0.012) (0.046) (0.013)

Household head married (dummy) -0.679^*** -6.587^*** -0.988^***

(0.111) (0.391) (0.009)

Note: Robust standard errors clustered at farmer group level are shown in parentheses. Only maize-selling households were included. ^*, ^**, and ^*** significant at 10%, 5%, and 1% level, respectively.

For the same subsample of maize-selling households, table 3.9 shows that male control of revenues is associated with lower consumption of calories, vitamin A, and zinc from purchased foods (for bean-selling households, only the effect for vitamin A is statistically

64

significant, table A3.16). In other words, women spend more on food and dietary quality than men, which seems especially relevant for vitamin A. As the models control for total household income, this negative gender pathway is a partial effect, which does not imply that the total effect of commercialization on nutrition is negative. But the analysis suggests that the total nutrition effects of commercialization could have been more positive if the loss of female control of revenues was prevented.

Table 3.9. Household Income, Gender Roles, and Consumption of Purchased Calories and Nutrients

Variables Calories

Male control of maize revenue (dummy) -314.030^* -233.409^** -1.850^* -0.271

(160.900) (113.886) (1.069) (2.026)

Age of household head (years) 1.634 -6.659 -0.064 -0.079

(8.635) (8.269) (0.060) (0.113)

Male household head (dummy) -90.309 247.160^** -1.432 -1.255

(162.464) (103.838) (1.161) (1.779)

Education of household head (years) -15.785 -21.426 -0.338^** -0.162

(18.603) (17.628) (0.147) (0.227)

Farm productive assets (1,000 Ksh) -1.526 -3.450^* -0.007 -0.003

(2.719) (1.795) (0.012) (0.023)

Access to credit (dummy) -157.467 148.089 -1.058 1.049

(192.238) (88.888) (1.133) (2.036)

Distance to closest market (km) 18.947 -5.054 0.098 -0.085

(18.854) (10.045) (0.146) (0.160)

Group official (dummy) 180.032 29.078 0.950 0.289

(150.986) (110.044) (1.083) (1.734)

Surveyed in December (dummy) 246.329 70.259 1.746 0.905

(229.295) (228.718) (1.897) (3.036)

Poor agroecology(dummy) 296.658^** -60.207 1.954^* 1.340

(131.591) (138.604) (0.989) (1.947)

Constant 2359.941^*** 571.188 17.336^*** 19.730^*

(519.677) (551.268) (3.705) (8.698)

Sub-county dummies Yes Yes Yes Yes

Observations 191 191 191 191

R-squared 0.272 0.150 0.216 0.119

Note: Coefficient estimates are shown with robust standard errors clustered at farmer group level in parentheses. All models were estimated with ordinary least squares. Only maize-selling households were included. AE, male adult equivalent; RE, retinol equivalent. ^*, ^**, and ^*** significant at 10%, 5%, and 1% level, respectively.