Examining the relationship between farm production diversity and diet diversity in Malawi Andrew Jones School of Public Health University of Michigan 5 th Annual LCIRAH Research Conference
INERTIA IN GLOBAL AGRICULTURE Dominant development paradigms suggest that low- and middle-income countries must: reduce the number of farmers overall; expand the size of farms, and the extent of monocropped, commercialized agriculture; increase yields by orders of magnitude; and integrate agriculture sectors into global systems of trade Source: Collier P, Dercon S. African Agriculture in 50 Years: Smallholders in a rapidly changing world? Oxford, UK: University of Oxford, 2013.
- biophysical limits - provisioning capacity of cities - nutrition and health
OBJECTIVES 1. Determine the association between farm production diversity and dietary diversity among a cross-sectional, nationally-representative sample of households in Malawi 2. Examine the association between crop diversity and dietary diversity using longitudinal data 3. Assess heterogeneity in sociodemographic, agricultural, and dietary characteristics of households across a gradient of farm production diversity
METHODS 2013 Malawi Integrated Household Panel Survey (IHPS) (n = 4,000 households; final sample: n = 3,010 households) Multi-topic, nationally representative survey carried out using stratified, two-stage sample design Detailed, plot-level data on amount of each crop harvested, and crop storage and sales Crop diversity measured using species count, and Simpson s Index, an adapted measure of biodiversity Image source: http://sweetpotatoknowledge.org
SIMPSON S INDEX Simpson s Index i = 1 - s 2 j where, s j is the share of crop j in the total area cultivated by the household i and is given by s j = a ij /A i, where a ij is the area cultivated on the jth crop by the ith household and A i is the total area cultivated under all crops
METHODS Household-level consumption of specific foods, and frequency of consumption of specific food groups in previous week collected evenly across the survey period and across all regions of the country Dietary diversity measured using adapted Minimum Dietary Diversity Women (USAID) indicator (i.e., consumption of 5 or more of 10 food groups); and Food Consumption Score (WFP) (i.e., consumption frequency of 8 food groups)
METHODS Multi-level regression models adjusting standard errors for intracluster correlations of the IHPS multistage sampling frame Generalized estimating equation (GEE) estimation with unstructured correlation structure for repeat observations to model longitudinal association between farm production diversity and household dietary diversity Survey rounds: 2010-2011 (i.e., Malawi Third Integrated Household Survey (IHS3)) and 2013 Malawi Integrated Household Panel Survey (IHPS) 3.8% attrition rate across survey rounds
CHARACTERISTICS OF 2013 SAMPLE n Mean (SD) or proportion Sociodemographic characteristics Household size 3,010 5.0 (2.2) Women with no education (%) 2,897 85 Dietary diversity Diet Diversity Score 3,010 6.3 (1.5) Food Consumption Score (FCS) 3,010 Poor (0-21) 1.3 Borderline (22-35) 14 Acceptable (>35) 85 Agricultural characteristics Available area for cultivation (ha) 3,010 0.69 (0.63) Crop diversity (annual crops) 3,010 2.7 (1.3) Proportion of harvest sold (%) 3,010 0.16 (0.23)
MULTIPLE LOGISTIC REGRESSION MODELS: DIETARY DIVERSITY ON CROP DIVERSITY Crop diversity count Simpson s Index Dietary diversity (dependent var.) MDD FCS MDD FCS Crop diversity 1.12** 1.13** 1.67* 1.51* Household size 1.06 0.99 1.06 0.99 Sex of household head (ref: male) Female 1.11 1.13 1.12 1.15 Age of household head (years) 0.99 0.99 0.99 0.99 Education of female 1.87** 1.77** 1.89** 1.79** Urban/rural residence (ref: urban) Rural 0.54 0.48* 0.58 0.53 Available area for cultivation (ha) 0.90 1.09 0.94 1.15 Proportion of harvest sold (%) 3.56** 1.61 3.32* 1.54 Odds ratios are shown. Models adjust for household wealth and intracluster correlations. *P<0.05, **P<0.01, ***P<0.001
GEE REGRESSION MODELS: DIETARY DIVERSITY ON CROP DIVERSITY Crop diversity count Simpson s Index Dietary diversity (dependent var.) DDS FCS DDS FCS Crop diversity 0.07*** 0.43** 0.25** 2.0* Household size 0.03** -0.18 0.03** -0.16 Sex of household head (ref: male) Female 0.06-0.55 0.06-0.53 Age of household head (years) -0.01*** -0.07*** -0.01*** -0.06*** Education of female 0.26*** 3.8*** 0.26*** 3.8*** Urban/rural residence (ref: urban) Rural -0.54*** -6.1*** -0.50*** -5.8*** Available area for cultivation (ha) 0.00 1.7** 0.03 1.8*** Proportion of harvest sold (%) 0.14 0.99 0.13 0.85 N = 2,530 households. Regression coefficients are shown. GEE models adjust for household wealth, and survey year, and use unstructured correlation structure. *P<0.05, **P<0.01, ***P<0.001
EXPLANATORY PATHWAYS Households with greater crop diversity are wealthier, and wealth drives greater dietary diversity produce a larger amount of crops for own consumption have greater income specifically from agriculture that may allow for more diverse food purchases have different food preferences or purchasing behaviors
HARVESTED AMOUNTS OF SPECIFIC CROPS BY QUANTILES OF CROP DIVERSITY Quantiles of crop diversity F-value 1 2 3 4 Cash crops (kg) Tobacco 79 (1598) 19 (164) 80 (651) 74 (407) 1.3 Cotton 1.1 (14) 27 (436) 6.7 (45) 17 (273) 1.0 Cash/subsistence crops (kg) Soybean 2.4 (24) 9.1 (56) 17 (78) 34 (110) 22*** Subsistence crops (kg) Groundnut 21 (195) 89 (295) 173 (541) 204 (543) 22*** Pigeonpea 0 (0) 8.7 (53) 17 (108) 45 (155) 26*** Maize 495 (872) 573 (1877) 561 (671) 692 (830) 3.8** Millet 1.8 (21) 6.5 (55) 9.1 (49) 18 (89) 7.7*** Nkhwani 0 (0) 0.77 (8.0) 4.9 (33) 14 (69) 19*** Permanent/tree crops (#) Mango 0 (0) 1.6 (2.8) 2.3 (4.1) 4.3 (8.5) 16*** Papaya 0 (0) 0.09 (0.41) 0.27 (1.1) 0.70 (1.9) 16***
AGRICULTURAL CHARACTERISTICS BY QUANTILES OF CROP DIVERSITY Quantiles of crop diversity F-value 1 2 3 4 Simpson s Index 0.09 (0.29) 0.36 (0.24) 0.49 (0.21) 0.62 (0.18) 645*** Cultivated area (ha) 0.40 (0.42) 0.55 (0.53) 0.69 (0.59) 0.88 (0.75) 79*** Proportion of harvest sold (%) 0.09 (0.24) 0.15 (0.24) 0.19 (0.24) 0.17 (0.21) 17*** Value of sold production ($US) 14 (64) 40 (175) 71 (198) 166 (951) 10*** Cattle ownership 0.24 (0.83) 0.71 (5.2) 0.81 (2.4) 0.52 (2.2) 2.9* Mean (SD) are shown.
FREQUENCY OF CONSUMPTION OF SPECIFIC FOOD GROUPS BY QUANTILES OF CROP DIVERSITY Food group-specific FCS Minimum Dietary Diversity (mean) 4 3.5 3 1 2.5 2 1.5 1 0.8 0.6 0.4 0.2 Pulses Dairy Flesh foods Fruits 0.5 0 0 1 2 3 4 Crop diversity quantiles 1 2 3 4 Quantiles of crop diversity All differences are statistically significant at P<0.05
FOOD EXPENDITURES (PREVIOUS WEEK) BY QUANTILES OF CROP DIVERSITY Expenditures ($US) 3 2.5 2 1.5 1 Grains Pulses Green leafys 0.5 0 1 2 3 4 Quantiles of crop diversity Least square means adjusted for household wealth and urban location; all differences are statistically significant at P<0.05
KEY RESULTS In cross-sectional analyses, independent of household wealth, maternal education, land area owned, and the market-orientation of farms, households with more diverse farm production had more diverse diets When the evenness of crop distribution was accounted for in addition to crop diversity, the association with diet diversity remained Longitudinal analyses confirm these findings, adding strength to the causal interpretation of the inference
KEY RESULTS Compared to households with low crop diversity, households with greater crop diversity produced more subsistence crops, especially pulses and vegetables, but not more cash crops cultivated more land, sold more of their production, and earned more income from agricultural output consumed pulses and fruits more frequently, and dairy foods less frequently spent less on key subsistence foods including grains, pulses, and green leafy vegetables
LIMITATIONS Food consumption data are not as rigorous as data collected using standard dietary intake data collection methods Metrics of dietary diversity are not very discriminating Findings are associative and may be biased Strongly subsistence and semi-subsistence systems; external validity of findings is unclear
CONCLUSIONS The diversity of farms likely plays an important role in influencing the diversity of diets in low-income, semi-subsistence, smallholder settings These linkages may operate through different pathways and are dependent in part on purchasing behaviors Quality of diet diversity and crop diversity is important in understanding these relationships Policies, programs and research need to understand and assess context to predict these linkages (e.g., stage of economic development, urbanization, nutrition transition, sociocultural values, biophysical determinants of crop choice) Determinants of nutrition are multi-factorial and diverse diets may not be sufficient in many cases to improve nutritional status
Thank you Andrew Jones, PhD Assistant Professor Department of Nutritional Sciences School of Public Health University of Michigan jonesand@umich.edu