Factors Affecting Food Security among Households in Nigeria: The Role of Crop Diversity

: Agriculture is central in order to achieve nutrition goals through the provision of food, energy and essential micronutrients for the physical and mental development of humans. Dietary diversity is a good indicator of human food security status. Using a dataset obtained from the World Bank’s Living Standards Measurement Study—Integrated Surveys on Agriculture (LSMS-ISA), this paper examines the linkages between crop diversity and food security (measured as dietary diversity) among farming households in Nigeria using ordinary least squares, Poisson regression, and instrumental variables (IV) Poisson regression to estimate the relationship. In addition, we investigate the determinants that lead to the consumption of each food groups considered for generating dietary diversity using a logit regression model. The results show that food expenditure, asset ownership and location of households are the key factors driving the types of food consumed by the households. Furthermore, increased food expenditure and access to credit were found to positively inﬂuence food security. The result established a positive and signiﬁcant relationship between crop diversity and dietary diversity. Our ﬁndings call for more attention to diet diversity, as well as the need to harmonize the roles of rural income improvement, especially through non-farm livelihood diversiﬁcation in tackling multiple nutritional deﬁciencies in Nigeria.


Introduction
It is well known that increased malnutrition can be caused by poor diversification of diets while the consumption of various nutritious food contributes to good health, and the consumption of diets from animal sources, vitamin-rich fruits and vegetables as well as nutrient-rich legumes can help substantially to reduce malnutrition [1,2]. According to the FAO, IFAD and WFP, about 700 million people go to bed hungry each night, and of equal significance, more than two billion of world's population suffer from hidden hunger due to lack of essential micronutrients, such as vitamin A, iron and zinc in diets [3][4][5]. Furthermore, nutritional deficiencies result in impaired physical and mental development of humans, loss of productivity, and susceptibility to various diseases among others [6,7].
Nutritional deficiencies are caused not only by low quantities of food consumed but also by poor dietary diversity since dietary diversity is a good indicator of broader nutritional status [8]. More diverse diets are also associated with lower rates of nutritional problems in many parts of the world [9,10]. Therefore, in order to improve nutrition, it is important to increase the diversity of healthy diets consumed. Dietary diversification (through increased availability and accessibility to nutritionally diverse foods) and bio-fortification (by increasing nutrients in existing staple grains) are suggested solutions for mitigating total agricultural output of the country on less than 2 hectares of land. Seventy-five percent of total land area in the country is potential agricultural land with only 48.5% of this area being under cultivation, while only 7% of the total irrigable land is being put to use [53][54][55]. In recent years, the Federal Ministry of Agriculture and Rural Development (FMARD) has inaugurated an Inter-Ministerial Agriculture/Nutrition Working Group, which is saddled with the responsibility of designing strategies for the promotion of healthy living, addressing malnutrition and increasing hygienic food production, thus placing particular emphasis on issues related to food and nutrition security in the country. This adds further to the policy relevance of this study which seeks to examine the linkages between crop diversity and dietary diversity in Nigeria.
Against this background, this paper contributes to the growing literature on the agriculture-nutrition-health nexus by using household survey data, the Living Standards Measurement Study-Integrated Surveys on Agriculture (LSMS-ISA) General Household Survey Panel Data conducted between 2012 and 2013 in the case of Nigeria, in order to specifically examine the relationship between crop diversity and dietary diversity in the country. In addition, we examine the factors influencing the consumption of each group items. These objectives will be useful to understand what drives the consumption of food group items and how the number of crops cultivated by the households affects the number of food groups consumed in the household accordingly. This study explores the two mechanisms by which agricultural pathways can increase nutrition, i.e., one of the pathways supports the fact that income generated from crop production can be used to purchase the non-cultivated food items to enhance their dietary diversity, while the second mechanism is through own-production to reflect cultivated crop and dietary diversity.

Review of the Relevant Literature
Previous studies have established linkages between land cropping patterns and the dietary diversity of households [56][57][58]. Some other studies have focused on the determinants of crop diversity or dietary diversity, but most of them have not examined the causal linkages between these two variables [59,60]. The relationship between farm diversity and dietary diversity among households in Tanzania and Kenya was examined by Herforth and Ahmed [44] and Jones et al. [61]. For instance, Jones et al. [61] found that the association between farm diversity and dietary diversity was positive among the households (p < 0.05). Additionally, Ochieng et al. [62], Rajendran et al. [63] and Habtemariam et al. [64] examined crop diversity and dietary diversity by specifically investigating whether the inclusion of vegetables in maize-based farming systems significantly improved the dietary diversity of the Tanzanian farming households involved in the study. Habtemariam et al. [64] found that production diversity is significantly and positively associated with dietary diversity.
The findings also revealed that the educational status of female heads of households and opportunities for women to earn income should be prioritized in order to extract further gains in dietary diversity [65]. For example, agricultural intensification programs in Northern Ghana, which targeted the low-income and female-headed households, significantly increased the production diversity of participating farmers [66][67][68]. Furthermore, for enhanced dietary quality among vulnerable households, attention should be paid to variables such as the cost of access to infrastructure, transport and food storage facilities.
Higher production diversity has also been found to improve dietary diversity [69], and it should therefore be encouraged to achieve enhanced nutritional outcomes for rural households, especially those in remote areas with difficulty in accessing markets [28,70,71]. Other studies examined the determinants of crop diversification with specific focus on micronutrients; the Household Dietary Diversity Score and Household Micronutrient Access Indicators were used to estimate the effect of crop diversification on household access to zinc, iron, vitamin A and folate. The findings emerging from these studies revealed that crop production diversification can be used to achieve nutrition-sensitive agriculture [72][73][74]. Zezza and Tasciotti [75] examined the relationship between urban agriculture and household food security, dietary diversity and calorie intake in 15 developing/transition countries. The study showed that there is an association between urban agriculture, dietary diversity and calorie intake, and it is closely related to food security. Two-thirds of the countries analyzed showed that there is a correlation between the active participation of urban households in agricultural activities and greater dietary diversity after controlling for economic welfare and a set of household characteristics. African countries included in the above study include Ghana, Malawi, Madagascar and Nigeria.
Turning to specific studies in Nigeria, Dillon et al. [76] examined the effect of exogenous variation in planting season decisions as a result of climate variability on household dietary diversity. The results showed that improving dietary diversity may not be achievable by agricultural income growth or increased crop diversity, a finding also in line with the study of Mulmi et al. [77]. Furthermore, when agricultural revenues increase, the diet composition of respondents remained unchanged. However, when considering the effect of agricultural income on the share of calories by food groups, diet composition changes a great deal [78][79][80]. At the same time, in Nigeria, there have been studies on food security, nutrition and agriculture mostly using primary data in different states of the country. Babatunde et al. [81] examined income and calorie intake among farm households in rural areas of Kwara State, Nigeria. The study showed a positive relationship between income and calorie intake as calorie intake does not increase substantially with income.
Other studies from Nigeria also show that households with large members have a negative relationship with food and nutrition security [82] and that education beyond the primary level, in particular, is effective in reducing malnutrition in young children as it increases nutritional and health knowledge for mothers to make better nutrition and health choices for the family [82,83]. Few empirical studies, however, have been carried out to examine the causal linkage between crop diversity and dietary diversity in sub-Sahara Africa and specifically in the case of Nigeria.

Data, Descriptive Statistics and Methodology
The source of the data is the household survey data of the LSMS-ISA General Household Survey (GHS) Panel Data. The survey is a nationally representative survey of approximately 5000 households conducted between 2012 and 2013. It is a long-term project that is aimed at the collection of household-level panel information, with data on household characteristics, welfare and agricultural activities. The GHS sample is comprised of 60 Primary Sampling Units (PSUs) or Enumeration Areas (EAs) chosen from each of the 37 states in Nigeria, thus making a total of 2220 EAs nationally. Each EA contributes 10 households to the GHS sample, resulting in a sample size of 22,200 households. Out of these 22,000 households, 5000 households from 500 EAs were selected for the panel component and 4916 households completed their interviews.
Given the panel nature of the survey, some households had moved from their location by the time of the Wave 2 visit, resulting in a slightly smaller sample as compared to Wave 1 (conducted in 2010-2011), with 4716 households in total for Wave 2 (conducted in 2012-2013) (NBS/FMARD, 2014). Data used in the analysis were from 2640 households in the dataset; this is because they were the households with the complete information needed to examine the objectives of the present study. Table 1 shows the food consumed by households across the country using 12 food groups which include: cereals, spices, eggs, fish, fruits, meat, milk, oil and fat, pulses, roots and tubers, sweets and vegetables. The food groups consumed were profiled by gender of the households (male, female) and location of residence (rural, urban) of the households. From the aggregated data, most households consumed cereals (96.85%), vegetables (95.79%), oil and fat (93.37%) among others, while the less-consumed food groups were eggs, fruits and milk, representing 11.89%, 28.48% and 31.97% of the households consumed, respectively (Table 1). On the disaggregated data, we found that more male-headed households consumed cereals (97.20%) than female-headed households (93.97%), while for vegetables, the reverse was the case (97.18% versus 95.59%). More female-headed households also consumed eggs, fish, fruits, milk and pulses (12.06%, 87.23%, 41.13%, 38.30 % and 86.87%, respectively) than male-headed households (11.87%, 57.55%, 26.97%, 31.21% and 81.21%, respectively). This implies that female-headed households are more conscious of the benefits derived from the consumption of these food groups, which are major sources of proteins. Eggs, fruits and milk were the least-consumed food groups in rural areas (10.89%, 26.67% and 29.62%, respectively). Additionally, few urban households consumed eggs, fruits and milk (19.54%, 42.35% and 49.84%). However, more households consumed these items in the urban areas than the rural areas. It was expected that rural households consume more eggs and fruits as these could be readily available in the location. Thus, this finding could be due to the fact that most of the rural households with access to these food items sell them off without considering the importance of consuming them for their health benefits. This is also consistent with the findings of Rajendran et al. [63].

Methodological Framework
The specific variables of interest in this study include dietary diversity and crop diversity, which we discuss below in detail, including the methodology used in the study.

Measures of Dietary Diversity and Crop Diversity
Dietary diversity: This is closely related to calorie adequacy and anthropometric outcomes [27,84], and as such, it can be used as a proxy food security in nutrition surveys. It can be defined as the number of different foods or food groups consumed over a given reference period [8]. There are two measures of dietary diversity: Food Consumption Score and the Household Dietary Diversity Score. This study used the Household Dietary Diversity Score. While the Dietary diversity score is an identified key indicator for surveillance of actions that aim to tackle various nutrition-related problems and food insecurity [1], it is also an effective indicator of food utilization [8,10].
Dietary diversity "measures the degree to which the variety of food consumed by households differs in terms of nutrient intakes over a given period of time" [85]. The Household Dietary Diversity Score (HDDS) is normally constructed using data on dietary intake in the previous 24 h, but in the absence of such data, a modified HDDS can be used which is based on food consumed in the household over the previous 7 days [86,87]. There were 12 categories of food groups used, with each food group having a score if anyone in the household consumed any food item from the specific group in the previous 7 days. The food groups used to calculate the score was similar to that of Snapp and Fisher [28], which included: cereals, roots and tubers, vegetables, fruits, meat and poultry, eggs, fish and seafood, pulses, legumes and nuts, milk and milk products, oils and fats, sugar and honey, and miscellaneous (such as condiments, coffee and tea).
Crop diversity: This is measured using crop count and agricultural revenue according to [76,88,89]. Crop count is the total number of different crop species cultivated by the household in a cropping year. Count variables have been previously used in the relevant literature to assess biodiversity and farm genetic diversity [61,[90][91][92].

Econometric Approach
Firstly, the logistic regression method was used to estimate the factors affecting the choice of food groups consumed by the households. Logit estimations are used when the outcome variable takes two possible states, hence the name binary models [93]. This will identify the variables that have significant influence on probability of consuming a food group in rural households. In this analysis, food group consumption (Z) is the dependent variable which takes the value of 1, if consumed among the farming households, and 0 otherwise, i.e., Z = 1, if a particular food group is consumed in the household, 0 otherwise. The logistic model postulates the probability (P i ) that food group consumption is a function of an index (Z i ), where: The probability of food group consumption is given by where e represents the base of natural logarithms (2.718).
The probability that a food group is consumed is calculated from Z i value: where: X 1 -X n are the independent variables; Z i = food group consumption in a household (1 if yes, 0 otherwise); b 0 = constant; b 1 = is the coefficient of the x's variables; Z i = food group consumption (1 if yes, 0 otherwise). Exogenous variables (X) used in the model include: household characteristics such as age, gender, years of education of the household's head, household size, monthly per capita expenditure on food items, monthly per capita expenditure on non-food items, ownership of assets (such as television, radio), farm size, and access to credit including regional dummies.
To examine the linkage between crop and dietary diversity, multiple linear regression was used following [61]; the Poisson model was also used following [28]. The Poisson estimator is commonly used for count data models [94]; furthermore, an auxiliary regression test (such as heteroskedasticity) [95] was used to tackle the problem of over dispersion or under dispersion that may occur in Poisson estimations. In short, unlike linear regressions, Poisson regressions allow for dependent variables to be censored at zero, and allow for noncontinuous dependent variables, but at the same time, they make the strong assumption that the variance is equal to the mean. In addition, we employed the Instrumental Variable (IV) Poisson regression to account for possible simultaneity between the two variables [96]. The equation is as follows: Exogenous variables (X) used in the model were based on previous studies [28,96], and include: household characteristics such as age, gender, regional dummies, agricultural revenue, ownership of assets (such as television, radio), household size, extension services, access to credit, monthly per capita expenditure on food items, farm size, monthly per capita expenditure on non-food items and marital status, while = random error term. The dependent variable is dietary diversity (HDDS) and the key independent variable is crop diversity-crop count and agricultural revenue. For the IV-Poisson, crop diversification variables were instrumented with one variable: access to extension services (such as seed services, market services as well as fertilizer distribution); the interaction with farmers' groups (and extension) is expected to influence farmers' knowledge about different crops and access to agricultural inputs [96]. We assumed (and tested) that this variable is correlated with the crop diversification variables, but not with the dietary diversity variable.

Results and Discussion
In this section, we discuss the findings regarding the factors influencing the consumption of each of the food groups-cereals, miscellaneous (such as condiments, coffee and tea), eggs, fish and seafood, fruits, meat and poultry, milk and milk products, oils and fats, pulses, roots and tubers, sweets (sugar and honey), and vegetables-used in assessing the household dietary diversity score (HDDS). Following that, we estimate the effect of crop diversification on dietary diversity.

Determinants of Food Groups Consumed by Households in Nigeria
A logistic regression was carried out to examine the socio-economic characteristics of households that determine their consumption of each of the 12 food groups. The results reported in Table 2 revealed the following: Notes: Standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

Cereals:
It was found that, compared to the southern region, households from the northern region of Nigeria (North Central and North West) have the higher likelihood to consume cereals. This is rather expected because most cereal production in Nigeria emanates from the northern region relative to southwest. Additionally, asset ownership, food expenditure and land size all significantly increased the likelihood of consuming cereals (Table 2). Moreover, the results implied that as households increase their food expenditure, the consumption of cereals increased. This is consistent with the findings of Musyoka et al. [97] and Torres [98]. The increase in land size could lead to having more land being used to produce cereals, which could make households increase their consumption of cereals. This finding is also in line with Yawson et al. [99], that land size plays a significant role in expanding production cereals.
Spices: It was observed that food expenditure and asset ownership increase the likelihood to consume spices. In Nigeria, the southern region is known to be high consumer of spices compared to the northern region. In corroboration, we found that living in the northern Nigeria (North Central and North East) reduces the probability of consuming spices. On the other hand, increased food expenditure and ownership of assets increased the likelihood of consuming more spices. Increased income vis-à-vis assets strengthens the purchasing power of households to expend more on food.
Eggs: The households of the northern region are known as higher consumers of eggs compared to the southern region. Contrary to this belief, we found a negative association between egg consumption and the northern region. The probable reason for this might be a result of conflict and extreme terrorism that have afflicted the region in recent years. This is consistent with the findings of [100], asserting that conflict, social crises and terrorism are responsible for increased food poverty in the region, among other things. Additionally, the age of the household's head significantly reduced the likelihood of consuming eggs. This observation is in accordance with the belief that eggs are consumed by younger individuals and not for older individuals. Conversely, food expenditure and agricultural revenue significantly increased the consumption of eggs as a food group, implying that households with more proceeds from agricultural production are able to consume more eggs with the revenue accrued from the sales of their produce.
Fish and sea foods: Increased food expenditure significantly increased the consumption of fish and sea foods. This could be due to the fact that the higher the amount of money allocated for food consumption, the higher the probability of dietary diversification. For those in South East and South South zones, a significant increase was observed in the consumption of fish and sea foods relative to those in the South West. This could be due to the fact that the South South is a riverine area with abundant fish and other sea foods. Additionally, marital status, increased age of the household's head, household non-food expenditure and household size increased the consumption of this food group. This supports the belief that older people should consume less meat and more fish and sea foods. In the same vein, being married increased the consumption of fish and sea foods. This is in line with the result on household size-this finding could be due to the fact that married people are expected to have large households, and as such, are expected to purchase more fish and sea foods as a good source of protein, especially for the children in the households.
Fruits: The sex of the household's head (male), asset index, and food expenditure positively and significantly determined the consumption of the fruit food group in the study area. Being a male-headed household, increasing food expenditure of the household and asset index increased the consumption of the fruit food group. However, residing in the region of the country (North Central, North East and North West) where there is conflict negatively determined households consuming fruits. In addition, household size negatively determined the consumption of fruits as a food group. This implies that an increased number of household members decreased the consumption of fruits.
Meat: Meat is a very good source of protein and its consumption is very important for improving nutrition outcomes, especially among children. We found that being a male household head, asset index, household food expenditure, land size and non-food expenditure positively and significantly determined a households' consumption of meat. This implies that male-headed households are more likely to consume meat than femaleheaded ones, as the male head allocates more expenditure on food, influencing the decision of a household caregiver to purchase more meat for the household's consumption. This is supported by Abubakar [101], who observed that female-headed households are less likely to report meat consumption than the males. On the other hand, being married, the age of the household's head, residing in North Central, North West and South East relative to the South West showed a negative and significant consumption of meat. The finding that being married reduces the consumption of meat supports earlier findings that married households consume more fish and sea food. Moreover, even though households in the northern region practice animal husbandry, they eat less meat than those in South West. The reason is that the northern herdsmen rear meat-producing animals for generating income rather than for consumption.
Milk: We found that regarding asset ownership, households having access to extension services and food expenditure positively and significantly determined households' consumption of milk in the study area (see Table 2). An increased asset ownership implies that there is an increase in income; hence, households are rich enough to purchase milk. Households can acquire knowledge on the consumption of different food groups (in this case, milk) through nutrition information they received from extension workers. In the same vein, households who can afford to increase their non-food expenditure can also increase their consumption of milk. However, the age of the household's head, household size and food expenditure negatively determined the consumption of milk as a food group. Larger household sizes consuming less milk could be due to the cost implication of its purchase. In addition, there is an increase in milk consumption as a result of households' increased food expenditure. The consumption of milk as a food group was negatively and significantly determined by residing in the North Central and North East geopolitical zones relative to the South West. As previously observed, residents in the northern zones as compared with the southwest consume less animal-sourced protein-rich foods.
Oil and Fat: Being married and household food expenditure positively and significantly determined the consumption of oil and fat (Table 2). Consistently, dwelling in the South East and South South positively and significantly determined the consumption of oil and fat. It is believed that these regions in Nigeria consume a lot of oil and fat compared to other regions, such as the northern parts. Additionally, residing in the North Central and North East zones relative to the South West negatively determined the consumption of oil and fat.
Pulses: Food and non-food expenditure positively and significantly determined the likelihood of households to consume pulses ( Table 2). In the same vein, residing in the South East relative to being in the South West positively and significantly determined the likelihood of consuming pulses. However, a negative significance was observed when North Central, North East and North West zones were considered relative to the South West. This implies that pulses are less consumed by households in the northern region, while they are more consumed by people in the South West. Land size and access to extension services were also negatively related to the consumption of pulses.
Root and tuber: From Table 2, we can see that socio-economic characteristics negatively and significantly determine the consumption of roots and tubers, including: age of the household's head, access to extension services, access to credit from cooperatives and food expenditure. The same was observed with households residing in the North Central, North East and North West zones relative to the South West. This is because roots and tubers are not a key component of diets in northern Nigeria. On the other hand, asset index and non-food expenditure of the household positively and significantly determined the consumption of roots and tubers.
Sweets: Furthermore, the results revealed that asset index, access to extension services, food and non-food expenditure all positively and significantly determine the consumption of sweets ( Table 2). The same was observed with households residing in the North Central, North East, North West, and geopolitical zones relative to the South West zone.
Vegetables: Agricultural revenue, household food and non-food expenditure positively and significantly determined households' consumption of vegetables. Thus, increasing the agricultural revenue of households as well as expenditure on food and non-food increases their consumption of vegetables because the households have enough money to purchase these vegetables. On the contrary, residing in the North Central, North East and North West relative to South West negatively and significantly determined the consumption of vegetables. The results imply that households in northern Nigeria consume fewer vegetables as compared to those in the South West zone.

Summary Statistics of Crop Diversity and Dietary Diversity
In order to measure crop diversity, the number of crops grown was counted ( Table 3). The maximum number of crops grown in Nigeria was nine. On average, farmers grow 1.846 ± 1.001 (2 crops). This depicts that most crop farmers in Nigeria carry out mixed cropping (75.92% of the farmers cultivate more than one crop). Based on the sector of residence, the average shows that there is somewhat a difference in the number of crops grown in rural households (1.87) and urban households (1.65). The number of food groups consumed was used to calculate the Household Dietary Diversity Score-used as a dependent variable ( Table 3). The maximum number of food groups consumed in the entire country was 12, with an average of 7.8 ± 2.00 (approximately 8 food groups); households in the urban areas consume more (approximately 8.85 ± 1.83) food groups (9) on average than those in rural areas 7.70 ± 1.98 (approximately 8 food groups). The result seems to suggest that the average dietary diversity of the households is rather fair.

Crop Diversity and Nutrition Diversity
The three regression models (OLS, Poisson and IV-Poisson) used in the analysis showed the relationship between crop diversity and nutrition diversity. Two indicators were used to measure crop diversity: Firstly, crop diversity considered as crop count (that is, the number of crops grown by the households) was used. Secondly, we proxied crop diversification with agricultural revenue from the different crops grown in the households. The rationale of the link emerges from the assumption that even though smallholder farmers are involved in mixed cropping, they cannot produce all the food items that would be consumed in the households. Hence, income generated from crop diversification can be used to purchase non-cultivated food items to enhance their dietary diversity. The analysis was carried out for the entire country using pooled data (Table 4); the disaggregated results by sector of residence (urban and rural) are presented in Tables 5 and 6. Although three regression results (OLS, Poisson, and IV-Poisson) are presented, our main focus is on the IV-Poisson estimation strategy results. The OLS results were found to be biased as they failed to fulfill the assumption of count variables. According to Greene [102], a common approach for count data models is to use a Poisson estimator. The Poisson estimator assumes equi-dispersion; that is, the mean and variance of the dependent variable are assumed to be equal. This assumption is often violated and can lead to incorrect standard errors. Overdispersion, where the conditional variance exceeds the conditional mean, is a common phenomenon. However, due to possible endogeneity between crop diversity and dietary diversity [27,61,96], we used the IV-Poisson estimation technique to explore the relationship between crop diversification and dietary diversification. Standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.
Our main discussion of the reported findings below will focus on the results emerging from the instrumental variable approach. We found that in some cases, the age of the household's head had a negative and significant relationship with nutrition diversity (HDDS). This implies that an increase in the age of the household's head decreases the likelihood of having dietary diversification. This could be due to the fact that the older the household's head becomes, the less likely she or he is to obtain additional income from off-farm activities which can be used to purchase diverse food groups in a bid to increase dietary diversity. These findings are also consistent with those of Babatunde et al. [81].
Ownership of assets has a positive significant relationship with dietary diversity in the model (column 3) using the crop count, but not in the agricultural revenue models (column 6). This implies that the ownership of assets leads to the households consuming more diversified diets; this is in line with the findings of Pellegrini and Tasciotti [27]. The findings from the study of Mannaf and Uddin [103] also suggest that the ownership of assets (such as livestock) is expected to reduce food insecurity and increase the likelihood of diversifying diets.
Household size was found to have a negative association with dietary diversity (both in the number of crops grown and agricultural revenue models). By implication, the larger the household becomes, the lower the food groups available to members, especially in low-income households where there is weak purchasing power for many food groups. This finding confirms Babatunde et al. [81] and Adebayo et al. [100], who observed that an increase in household size increases the probability of food poverty. In addition, access to credit has a positive and significant relationship with dietary diversity. Access to financial services has been associated with the adoption of productivity-enhancing technologies [69]. Therefore, if income increases through increased productivity, this could increase the purchasing power of households so they can purchase diverse food groups for consumption.
Food expenditure is positively influencing the likelihood to diversifying nutrients in Nigeria. Various studies have revealed that increased food expenditure increases the amount allocated to each food group, which will most likely increase the possibility of diversifying the food consumed by the household [97,98,104]. Non-food expenditure has a positive significant relationship with HDDS for all the models estimated. This implies that when households spend an additional budget on non-food items, they also increase expenditure on food items (see models 2 and 3), and it is expected that their dietary diversity increases.
On the relationship between crop diversity and dietary diversity using a number of crops grown by the household, we found that crops grown positively and significantly influences the diversification of diet in the household. This is possible since as the number of food crops increases the household tends to have more option to upgrade and diversify from the consumption of unbalanced diets that possibly influence malnutrition, stunting and wasting especially among children. Thus, the higher the crop count, the higher the likelihood to consume divers' kinds of food groups. This agrees with several findings in the literature, which revealed that an increase in farm production diversity is responsible for the number of food groups consumed in the household [63,97,98,[104][105][106][107][108][109].
The use of agricultural revenue as a proxy for crop diversity provided a consistent result with the number of crops grown by the households. This finding suggests that crop diversification is a key determinant of dietary diversity in Nigeria, as revealed also in other countries such as Malawi [61], Ethiopia [108] and India [63]. Additionally, the plot of polynomial associations between crop diversity and dietary diversity suggests that the higher the crop diversification, the higher the dietary diversity is in Nigeria (Figure 1).
Finally, the geopolitical zones in Nigeria were also modelled as a factor influencing dietary diversity. The estimation in all models revealed a negative and significant relationship between the region of residence and the nutritional diversity of households across Nigeria, especially for the northern region (North Central, North East and North West). Northern Nigeria have been plagued with security issues in recent years, which have contributed to a downward trend in development and growth indicators in the region, despite several interventions to reduce hunger, malnutrition and poverty. In order to obtain a better idea regarding the situation in rural and urban households, the data were disaggregated to investigate the relationship of crop diversity with dietary diversity. Interestingly, some similar findings were found from the relationship between the cofounding factors used along with crop counts and agricultural revenue as a function of dietary diversity. For instance, variables such as the age of the household's head, asset ownership, food and non-food expenditure as well as being resident in the northern region of the country, particularly the North East, all had a significant relationship.
However, the sex of the household's head was found to be positively related to nutrition diversity in the urban areas, while the reverse was the case in the rural areas. This implies that a male-headed household reduces nutrition diversity in rural Nigeria, while it increases it in urban Nigeria. The probable reason for this might be the empowerment, social and human capital level of male-headed households in the urban areas compared to rural households. Household size, access to credit and use of firewood significantly determine dietary diversity in rural Nigeria. Specifically, an additional member of the household reduces the nutritional diversity of such households. This could be due to the fact that an additional household member will put a strain on the consumption of the household. Access to credit, on the other hand, increases the dietary diversity of households as they will have enough funds for their expenditures.
In relation to crop diversity and nutrition diversity, we also found that crop diversification and agricultural revenue have a positive and significant relationship with dietary diversity in rural households, while an inverse relationship was recorded in the urban households. The findings suggest that low crop diversification has the likelihood of reducing dietary diversification. Although farming is a rural phenomenon compared to urban, these findings do not necessarily suggest that the rural households were able to diversify diets more than the urban households. Specifically, linking crop diversification with dietary diversification, rural households seem to be in a better position than urban households in Nigeria. Our findings suggest that crop diversification is a key factor in reducing malnutrition in Nigeria. This is further supported by the non-parametric analysis expressed in Figures 2 and 3. This finding, which is in line with the results reported by other studies, suggests that crop diversification increases the diversification of diets, especially in rural households [63,97,98,[105][106][107][108][109].

Conclusions and Policy Options
The main results emanating from this study illustrate two alternative approaches (crops grown and agricultural revenue) to measuring the effect of agricultural on-farm production on household nutrition using dietary diversity. Usually, agricultural pathways to increase nutrition are likely to occur through two mechanisms, either through income effects or via the increased consumption of own produced foods. The findings reveal that on-farm production diversity is positively associated with dietary diversity in Nigeria in both approaches. Hence, on-farm crop diversification may help to improve household diets to some extent. However, the magnitude of the estimates suggests that the positive dietary effects of further diversifying on-farm production will be relatively small. With regard to rural and urban disparity, we found that crop diversification and agricultural revenue have a positive and significant relationship with dietary diversity in rural households, while an inverse relationship was recorded in the urban households. The findings further show that cereals are the most consumed food groups among the farming households compared to other food groups. In a consistent manner, food expenditure, asset ownership and regional dummies (precisely in the northern region) are the key factors among others in driving the types of food consumed by the households. Although these findings call for improved attention on diet diversity, it is equally important to emphasize the harmonizing roles of rural income improvement, especially through non-farm livelihood diversification (since the households under examination were predominantly rural) in tackling multiple nutritional deficiencies in Nigeria. In particular, relevant stakeholders should consider the nutritional status of households in the northern region, which has been plagued by serious security issues in recent years. Furthermore, since it was found that crop diversification is assumed to influence household diet diversification, efforts should be made on the policy front to prevent and cure severe malnutrition in Nigeria, so that the increase in agricultural intensification and biodiversity as a strategy to fight malnutrition in Nigeria should not take the place of malnutrition-reducing policies, but rather should be seen as complementary and supporting approaches, especially in northern and rural households. Having said that, one of the limitations of this study are the data used, which are slightly "outdated" and may not fully reflect the current realities due to time difference. In addition, we acknowledge that the use of agricultural revenue as a proxy for crop diversity may not be a best-fit variable. However, the findings emerging from this study could serve as useful background material for future research work, using more recent data and more appropriate variables.

Acknowledgments:
We would like to thank the Editor of the journal and the three anonymous reviewers for their very helpful comments and suggestions, which were of great benefit to the current version. We would like to appreciate the comments received from the participants of the 2019 HORTISON Conference. Finally, we would like to express our appreciation to our respective institutes for providing us with an excellent environment in order to complete this research. Needless to say, the usual disclaimer applies.

Conflicts of Interest:
The authors declare that there is no conflict of interest.