Does Food Insecurity exist among Farm Households? Evidence from Ghana

Household food security exists when households have physical, social and economic access to sufficient, safe and nutritious food at all times that meets their dietary needs and food preferences for an active and healthy life. Food security remains a serious challenge for many households in Ghana and the situation is even more prevalent among smallholder farmers. Using data collected from 2,603 farm households across Ghana and employing an ordered probit model the determinants of food security among farm households were assessed. The food security indicator-Food Consumption Score (FCS) which combines diet diversity, frequency of consumption and relative nutritional importance of different food groups was used for the analysis. Results indicated that farm households (76%) across Ghana were within the acceptable household food consumption groups. Nonetheless, 19% and 6% of farm households respectively were within the borderline and poor food consumption groups respectively. Further analysis revealed the determinants of food security to include experience, gender, improved variety adoption, access to credit and location. The suggestion is that government and private institutions should create an enabling environment to enhancing production capacities, economic and social resilience to improve on food security and nutrition.


Introduction
defines Food security as "a situation in which all people at all times have physical and Measuring farm household food security 140 Food security is multidimensional and thus presents a variety of measurements. Several 141 indicators have been developed as proxies for food security. A comprehensive all-encompassing 142 measure of food security would be that measure that is valid and reliable, comparable over time 143 and space, and which captures different elements of food security [18]. Nevertheless, the 144 complexity of food security, as a crosscutting discipline, has engrossed the challenge to finding a 145 summative (or 'gold standard') measure of household food insecurity [19].
[20] includes thirty-146 three indicators in its recommended list of indicators for measuring food insecurity access 147 alone. [21] asserted that the need to finding a simple and realistic measure of household food 148 insecurity that can be labelled as "golden rule" combining rigour and statistical efficiency to 149 conclude food insecurity from the household level upwards is important. Some measures of food 150 security from various literature [19,22] are as enumerated in Table 1. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made 153 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Respondents were asked about frequency of consumption of specific food items (in days) over a 173 recall period of the past 7 days. Food items were grouped into 8 standard food groups with a 174 maximum value of 7 days/week [24]. The consumption frequency of each food group was 175 multiplied by an assigned weight ( Table 2) that is based on its nutrient content. Those values 176 were then summed up obtaining the Food Consumption Score (FCS). FCS was then classified . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made granting that the outcome is discrete, the multinomial logit or probit models would fail to 187 account for the ordinal nature of the dependent variable. Whereas the logit assumes a logistic 188 distribution of the error term, the probit assumes a normal distribution [29]. According to [28] 189 the logistic and normal distributions generally give similar results usually. Moreover, [30] point . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 4, 2021. ; https://doi.org/10.1101/2021.02.04.429712 doi: bioRxiv preprint out that the ordered probit is the most used model for ordered response data in applied 191 econometric work. The ordered probit is used in this study as a result. Following [29], an ordered The probability of observing a particular outcome, for 1 ≤ ≤ is given by:

211
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made where F is the cumulative distribution function for , 0 = -1 = -∞, and = ∞.The 215 presence of F leads to a maximum likelihood estimation framework written as: This log-likelihood function is maximized with respect to , and the cut points 1 < 2 . In the 219 case of two discrete outcomes, the log-likelihood function in(4) simplifies to the binary choice 220 model with one cut point which is normally set to be 0 to achieve identification of the intercept 221 term [29].
Consequently, the empirical model of this study is specified as: where FCS is food consumption score used as food security proxy; j represents a household, i (i 232 = 0, 2, 3) represents the three categories of alternative dependent ordered variables indicating (i) . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 4, 2021. ; https://doi.org/10.1101/2021.02.04.429712 doi: bioRxiv preprint whether a household falls within poor food consumption group category, (ii) whether a 234 household falls within borderline food consumption category, and (iii) whether a household is 235 within the acceptable food consumption category group. W, X and Z are, respectively, 236 socioeconomic, food production, institutional and location characteristics hypothesized to 237 influence food security (Table 3); , , , are parameters to be estimated.  (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Access to institutional factors is very important to increasing production of farm produce and 263 thus increasing food access. Credit may increase the probability of a household's ability to 264 procuring production inputs as seeds, chemicals, and hiring of labour [38], which could improve 265 production and thus the household food situation. Access to credit by households was therefore 266 predicted to positively correlate with household food security status.  . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made     The implication is that some farmers from that part of Ghana may not have the luxury to produce 317 diverse crops that contribute to food security.

267
[27] stated that households that cultivate at least 318 three different types of crops have better food consumption score than those that only cultivate 319 one type in the Northern part of Ghana in their study of food security situation in the region. The Western region is known for the production of cash crops at the expense of food crops and thus 321 the result was not surprising. We also discovered that there were some farm households across   We found that male headed farm households were more likely to be poorly food secured (poor 345 food group) and moderately food secured (borderline food group). Nonetheless, female headed 346 farm households were more likely to be highly food secured (acceptable food group). Female 347 headed households mostly produced food crops that are mainly eaten by the household. They 348 would only sell the surplus from production. The reverse is true for the male headed households 349 who would sell most of their produce. Our results discovered that growing of improved varieties decreased the probability of being 352 severely food insecure and moderately food insecure but increases the probability of being 353 highly food secure. Improved varieties are high yielding and those that produce them are likely 354 to have enough for the household and also for sale to be able to purchase other food items. [43] . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 4, 2021. The odds of being food secure by farm households were determined by factors that include, 412 gender, experience, education level, and access to credit, improved variety cultivation and 413 location. Our result showed that females were more food sure than males. The suggestion is that 414 equitable access to inputs (credit access, labour acquisition, planting material, etc) and 415 infrastructure be made available to all farmers to increase production and contribute to food The general understanding is that education is important in uptake of improved technologies.

422
Education determining food security suggests that educated members of farm households are 423 able to diversify their livelihood portfolios, which could improve supply of diverse food items.  (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made    . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made   (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made