**4. Findings**

**Table 1** summarizes the data used in estimating the regression models. The results are presented for farmers that participated in the project (and hence were sensitized about the benefits


**Table 1.** Summary statistics and t-test of differences in means by project participation.

of OFSP) and those that did not. Growers of orange-fleshed sweetpotato were participants of the project. As results indicate, project participants differed from the nonparticipating ones in terms of age, many of the asset endowment variables, agroecology, and varietal traits, as shown by the very low *p*-values of tests of differences in means. Specifically, participating farmers differed in terms of income and membership to farmer groups. There is also weak evidence that farmers significantly differed with regard to access to land on valley bottoms.

group (i.e., participants) and the last group constitutes the nonintervention group (i.e., nonparticipants). The list of farmers in each category (i.e., intervened and nonintervention) was then compiled at the village levels in each of the project wards and districts. A random sample of farmers was selected from each category of farmers for personal interviews. In total 481 project participants and 251 nonparticipants were interviewed. The sample contained 221 and 511 male and female farmers, respectively. The high number of female farmers reflects the fact that women mostly grow sweetpotato and that the project also targeted female household members. Data collected included farmer and farm characteristics, asset endowments, insti-

**Table 1** summarizes the data used in estimating the regression models. The results are presented for farmers that participated in the project (and hence were sensitized about the benefits

**Mean Std Dev Mean Std Dev t-stat** *p***-value**

**Variable Participant (***N***=481) Nonparticipant (***N***=251) Test of diff. in means**

gender 0.31 0.03 0.02 0.47 −0.81 0.4182 lneduc 0.58 3.13 0.27 3.41 −1.23 0.2175 lnage 3.79 0.27 3.84 0.28 1.97 0.0246

valley 0.68 0.47 0.64 0.48 −1.32 0.0939 lndistmkt 1.26 1.61 1.24 1.58 −0.14 0.8878 lnfmsize 1.08 0.81 1.02 0.98 −0.89 0.1966

lncropinc 6.84 8.59 5.63 8.6 −1.79 0.0363 group 0.64 0.48 0.39 0.49 −6.59 0 assetindex 3.56 2.15 3.4 2.02 −1.01 0.1549

moreyield 0.93 0.25 0.9 0.29 −1.41 0.0808 sugary 0.75 0.43 0.78 0.41 0.91 0.819

aezP8 0.34 0.47 0.5 0.5 4.09 0 aezP5N10 0.17 0.38 0.02 0.15 −5.97 0 aezP4W3 0.48 0.5 0.47 0.5 0.15 0.4373

**Table 1.** Summary statistics and t-test of differences in means by project participation.

tutional characteristics, and varietal traits.

**4. Findings**

26 International Development

*Farmer/household specific variables*

*Farm specific variables*

*Capital endowment factors*

*Varietal trait variables*

*Agroecological factors*

**Figure 2** presents the proportion of farmers growing different varieties of orange-fleshed sweetpotato as well as the cleaned-up local white-fleshed variety, New Polista. New Polista and Kabode were planted by nearly one-half of the farmers interviewed while only about one-quarter of the farmers planted Ejumula and Jewel. The high percentage of farmers planting New Polista is due to its popular traits such as high dry matter content and relatively high sugar content making it tastier than orange-fleshed varieties. New Polista is also more tolerant to the sweetpotato pests, especially the sweetpotato weevil, and to sweetpotato virus disease (SPVD). Among the orange fleshed varieties, Kabode is most widely grown by the farmers owing to its resistance to SPVD, good taste, and the fact that it is rich in vitamin A. However, compared to New Polista, Kabode has a lower dry matter content and is less tolerant to moisture stress caused by droughts.

The low adoption of Ejumula and Jewel is mainly attributed to their susceptibility to SPVD, hence requiring higher disease management, and low dry matter content (for the case of Jewel), despite the latter being more tolerant to moisture stress than Kabode.

This study estimated a multivariate regression model to understand the factors that drive the choice of variety of sweetpotato to grow, and to test the null hypothesis that choice of sweetpotato variety planted is not affected by awareness of nutritional benefits of OSFP. The sweetpotato varieties included in the estimated regression model were Jewel, Ejumula, Kabode (for orange-fleshed varieties), and New Polista (for improved local varieties, also promoted by the project but quite popular among the consumers).

**Table 2** presents the tests of interdependence/correlation in the decision to plant different varieties promoted by the Marando Bora project, i.e., the atrho and rho. As expected, the

**Figure 2.** Proportion of farmers growing different varieties of sweetpotato: % (*N* = 732).


*Note*: Likelihood ratio test of rhoejumula-kabode = rhojewel-kabode = rhonewpolista-kabode = rhojewel-ejumula = rhonewpolista-ejumula = rhonewpolista-jewel: Chi2 (21) = 118.42; *p*-value = 0.000.

**Table 2.** Tests of correlations in decision to use different sweetpotato varieties.

results indicate that there is statistically significant and positive interdependence/correlation in the decision to choose among the orange-fleshed varieties and also between the orangefleshed varieties and the white-fleshed New Polista. This finding indicates that estimating separate Probit or Logit regression models to assess the determinants of the decision to plant OFSP varieties would result in biased estimates and justifies the use of multivariate regression technique.

The results of the estimated multivariate probit model are presented in **Table 3**, by variety. We discuss the results by variety below.

#### **4.1. Factors affecting the decision to grow Kabode**

As hypothesized, results show that awareness of the nutritional benefits of OFSP and household food security status significantly affect farmer's decision to plant Kabode, an OFSP variety. The *p*-values for the two variables are both 0.0000 indicating that there is very strong evidence from the data to suggest that participation in the Marando Bora project (hence being aware of benefits of eating OFSP) and being food insecure affects the decision to plant Kabode.

Results also indicate that age and agroecology of the area affect the choice of variety planted. Specifically, age is negative and statistically significant indicating that older farmers are less likely to choose to grow Kabode. This finding is probably because older farmers are more used to the local varieties, thus find it difficult to switch to the orange-fleshed varieties. At the same time, aezP8, one of the variables included to capture the effect of agroecology on the choice of variety to plant, is both positive and strongly significant. Specifically, results indicate that farmers in areas falling within zone P8, which is less dry, are more likely to plant Kabode than those falling in the regions covered by zone W3 and P4 (i.e., aezP4W3, the base).


results indicate that there is statistically significant and positive interdependence/correlation in the decision to choose among the orange-fleshed varieties and also between the orangefleshed varieties and the white-fleshed New Polista. This finding indicates that estimating separate Probit or Logit regression models to assess the determinants of the decision to plant OFSP varieties would result in biased estimates and justifies the use of multivariate regression

*Note*: Likelihood ratio test of rhoejumula-kabode = rhojewel-kabode = rhonewpolista-kabode = rhojewel-ejumula =

**Coeff.** *p***-value Coeff.** *p***-value**

0.313 0.000 rho jewel-kabode 0.303 0.000

kabode

ejumula

jewel-ejumula

newpolista-jewel

(21) = 118.42; *p*-value = 0.000.

ejumula-kabode

0.417 0.000

0.271 0.000

0.537 0.000

0.264 0.000

0.181 0.011

0.444 0.000 rho

0.599 0.000 rho

0.183 0.013 rho

**Table 2.** Tests of correlations in decision to use different sweetpotato varieties.

0.278 0.000 rho newpolista-

0.270 0.001 rho newpolista-

The results of the estimated multivariate probit model are presented in **Table 3**, by variety. We

As hypothesized, results show that awareness of the nutritional benefits of OFSP and household food security status significantly affect farmer's decision to plant Kabode, an OFSP variety. The *p*-values for the two variables are both 0.0000 indicating that there is very strong evidence from the data to suggest that participation in the Marando Bora project (hence being aware of benefits of eating OFSP) and being food insecure affects the decision to plant

Results also indicate that age and agroecology of the area affect the choice of variety planted. Specifically, age is negative and statistically significant indicating that older farmers are less likely to choose to grow Kabode. This finding is probably because older farmers are more used to the local varieties, thus find it difficult to switch to the orange-fleshed varieties. At the same time, aezP8, one of the variables included to capture the effect of agroecology on the choice of variety to plant, is both positive and strongly significant. Specifically, results indicate that farmers in areas falling within zone P8, which is less dry, are more likely to plant Kabode than those falling in the regions covered by zone W3 and P4 (i.e., aezP4W3, the base).

technique.

atrho ejumula-kabode

atrho jewel-kabode

atrho jewel-ejumula

atrho

atrho newpolistakabode

28 International Development

atrho newpolistaejumula

newpolista-jewel

Kabode.

discuss the results by variety below.

rhonewpolista-ejumula = rhonewpolista-jewel: Chi2

**4.1. Factors affecting the decision to grow Kabode**

**Table3.** Factors influencing the use of sweet potato varieties: multivariate probit regression.

A Study of Household Food security and Adoption of Biofortified Crop Varieties in Tanzania: The Case... http://dx.doi.org/10.5772/67677 29

#### **4.2. Factors affecting the decision to grow Jewel**

The factors driving the decision to plant Jewel are presented in columns 4 and 5 of **Table 3**. As in the case of Kabode, we find, as hypothesized, a strong evidence that both participation in the project (a proxy for awareness of benefits of OFSP) and household food security status affect the decision to plant Jewel. Both variables have *p*-values of 0.0000 indicating that there is firm evidence from the data against the null hypothesis that awareness of the nutritional benefits of OFSP and food insecurity in the household have no effect on the decision to plant Jewel. The results also indicate that education increases the likelihood that a farmer adopts Jewel. This finding is in line with our priori expectations: more educated farmers are more likely to adopt an OFSP variety because education is associated with the demand for quality attributes (in this case the vitamin A content) of food [23, 24].

Results further show that farm-specific, asset endowment, and agroecological variables also affect the decision to plant Jewel. In particular, distance to the market (a proxy for transaction costs) and farm size both reduce the likelihood that a farmer chooses to plant Jewel. The finding that high transaction costs reduce adoption of OFSP variety is in line with the adoption literature which suggests that high transaction costs reduce the price farmers earn and hence dampen the incentives to invest in agricultural technology [25]. At the same time, the results indicate that being in areas under zone aezP5N10 and being a member of a farmer group increase the likelihood that a farmer will decide to plant Jewel. These results are as expected: moving from the less dry zone (i.e., aezP4W3) to the less dry (i.e., aezP5N10) provides more of the moisture needed to grow the less drought tolerant OFSP varieties such as Jewel. The finding that membership to a farmer group increases the likelihood of planting Jewel may be because such groups provide technical and social support to members as well as some valueaddition activities [18, 26–28].

#### **4.3. Factors affecting the decision to plant Ejumula**

The factors affecting the decision to plant Ejumula, the final OFSP variety we examined in this study, are presented in columns 6 and 7 of **Table 3**. As in the two previous cases, the results indicate that benefit awareness and household food insecurity increase the likelihood that a farmer will grow Ejumula. The null hypothesis that knowledge of the benefits of OFSP and the household food insecurity have no effect on farmer's decision to plant Ejumula is rejected. Results also show that age affects the likelihood that a farmer plants Ejumula. Specifically, as in the case Kabode, the results indicate that older farmers are less likely to adopt Ejumula than younger ones.

Among the farm-specific variables, distance to market and size of farm are both negative and statistically significant suggesting that farmers that face higher transaction costs in the marketing of sweetpotato and those with large farms are less likely to adopt Ejumula. The finding relating the distance to the market (a proxy for transaction) supports our argument above that high transaction costs dampen incentives to grow OFSP. Asset endowment variables, income from crop sales and membership in a farmer group, also affect the decision to plant Ejumula. An increase in crop income and being a member of a group both increase the likelihood of planting Ejumula. The effect on crop income on the adoption of Ejumula is likely to be related to the financial requirements for labor use in the production of sweetpotato. Hired labor is one of the most expensive external inputs in the commercial production of vegetatively propagated crops [29]. Results also show that, among the agroecology variables, farmers in the less dry zone aezP5N10 are more likely adopt Ejumula compared to their counterparts in the drier zone aezP4W3 indicating that agroecology is important in the decision to adopt Ejumula.

#### **4.4. Drivers of decision to adopt New Polista**

**4.2. Factors affecting the decision to grow Jewel**

30 International Development

attributes (in this case the vitamin A content) of food [23, 24].

addition activities [18, 26–28].

than younger ones.

**4.3. Factors affecting the decision to plant Ejumula**

The factors driving the decision to plant Jewel are presented in columns 4 and 5 of **Table 3**. As in the case of Kabode, we find, as hypothesized, a strong evidence that both participation in the project (a proxy for awareness of benefits of OFSP) and household food security status affect the decision to plant Jewel. Both variables have *p*-values of 0.0000 indicating that there is firm evidence from the data against the null hypothesis that awareness of the nutritional benefits of OFSP and food insecurity in the household have no effect on the decision to plant Jewel. The results also indicate that education increases the likelihood that a farmer adopts Jewel. This finding is in line with our priori expectations: more educated farmers are more likely to adopt an OFSP variety because education is associated with the demand for quality

Results further show that farm-specific, asset endowment, and agroecological variables also affect the decision to plant Jewel. In particular, distance to the market (a proxy for transaction costs) and farm size both reduce the likelihood that a farmer chooses to plant Jewel. The finding that high transaction costs reduce adoption of OFSP variety is in line with the adoption literature which suggests that high transaction costs reduce the price farmers earn and hence dampen the incentives to invest in agricultural technology [25]. At the same time, the results indicate that being in areas under zone aezP5N10 and being a member of a farmer group increase the likelihood that a farmer will decide to plant Jewel. These results are as expected: moving from the less dry zone (i.e., aezP4W3) to the less dry (i.e., aezP5N10) provides more of the moisture needed to grow the less drought tolerant OFSP varieties such as Jewel. The finding that membership to a farmer group increases the likelihood of planting Jewel may be because such groups provide technical and social support to members as well as some value-

The factors affecting the decision to plant Ejumula, the final OFSP variety we examined in this study, are presented in columns 6 and 7 of **Table 3**. As in the two previous cases, the results indicate that benefit awareness and household food insecurity increase the likelihood that a farmer will grow Ejumula. The null hypothesis that knowledge of the benefits of OFSP and the household food insecurity have no effect on farmer's decision to plant Ejumula is rejected. Results also show that age affects the likelihood that a farmer plants Ejumula. Specifically, as in the case Kabode, the results indicate that older farmers are less likely to adopt Ejumula

Among the farm-specific variables, distance to market and size of farm are both negative and statistically significant suggesting that farmers that face higher transaction costs in the marketing of sweetpotato and those with large farms are less likely to adopt Ejumula. The finding relating the distance to the market (a proxy for transaction) supports our argument above that high transaction costs dampen incentives to grow OFSP. Asset endowment variables, income from crop sales and membership in a farmer group, also affect the decision to plant Ejumula. An increase in crop income and being a member of a group both increase the likelihood of The results of the factors affecting the decision to use the white-fleshed New Polista variety are presented in the last two columns of **Table 3**. Among, the farmer-specific variables, participation in the project (i.e., intervene) is the only variable that affects the decision to grow New Polista. Specifically, the results indicate that participation in the project increases the likelihood of planting New Polista. While this finding may appear to be contrary to expectations, it actually captures the way the project was designed. While the Marando Bora promoted awareness of the benefits of OFSP, it also promoted the growing of other popular local varieties such as New Polista for the food security purposes and as a source of income. Indeed, this is the reason why the project cleaned up the Polista variety, to remove viruses and to increase its yield potential. The results however do not find food insecurity a significant factor in the decision to grow New Polista.

Two farm-specific variables affect the decision to grow New Polista, namely having access to a valley bottom and size of land. The former increases while the latter decreases the likelihood of planting New Polista. The finding that access to valley bottom influences the decision to grow New Polista likely relates to the fact that access to land in such areas enables farmers to conserve planting materials (i.e., vines) during the dry periods. Results also show that crop income, taste (i.e., sugary), and being in less dry agroecological zone (aezP8) all increase the likelihood of growing New Polista. Notably, the results indicate that farmers who perceive New Polista to be sweeter (i.e., sugary) are more likely to grow it.

The results above have shown that agroecology and asset endowment play a major role in the decision to grow the various improved and cleaned up sweetpotato varieties, including the OFSP. To examine whether these variables jointly affect the decision to plant the four sweetpotato variables analyzed in this study, we conducted appropriate hypothesis tests for each. Following [18, 19], we also examined if the varietal traits affect the likelihood that a farmer will plant these improved varieties. Specifically, a Wald joint-exclusion test of the agroecology variables (proxied aezP5N10, aezP8) and access to the valley bottom, all of which are associated with moisture availability, yielded a Chi-square and *p*-value of 60.88 and 0.0000, respectively. This result indicates these variables jointly affect farmer's decision about the choice of sweetpotato varieties to plant. Thus, the null hypothesis that agroecology of the area has no effect on the choice of sweetpotato variety planted by the farmer is therefore rejected. On the other hand, a Wald test of nonsignificance of asset endowment variables (represented by asset-index, crop income, and size of land) yielded a Chi-square and *p*-value of 24.81 and 0.0158, respectively. This finding indicates that there is firm evidence from the data to suggest that farmers' endowment with physical assets (proxied by asset-index and farm size) and financial capital (proxied by income from previous crop) affect the decision about the sweetpotato variety grown. A Wald test of joint exclusion of the varietal attributes (proxied by moreyield and sugary) however finds no evidence that these variables jointly affect the choice of variety planted. The test yields a Chi-square and *p*-value of 10.68 and 0.2203.
