Insights into the Current Tobacco Farming Landscape in January 2019 About Fraym Fraym was born out of frustration. Before Fraym, we faced one consistent and fundamental challenge across the African continent— the lack of hyper-local data and analytics to make informed decisions. Many important decisions are made with highly inadequate information. Often, analysis is limited to anecdotes, high-level or aggregated statistics, or gut instincts. Fraym offers a better way. FRAYM || METHODS

Acquire Data geo-tagged household surveys remote sensing data satellite imagery

On-Board Data compile clean harmonize geospatially-enable

Produce Data Layers machine learning proprietary algorithms artificial intelligence automation

Deliver to Customers data layer APIs front-end tools analytic services Executive Summary EXECUTIVE SUMMARY || TOBACCO-FARMING HOUSEHOLDS

The 5 percent of households engaged in tobacco farming share many challenges with other agricultural households. They also tend to be more diversified, connected, and exposed.

Shared challenges with other agricultural households Limited access to reliable markets Few rural areas have reliable access to agricultural markets. Rumphi Limited food security and diet diversity 1 in 4 tobacco-farming households has a diverse diet, 3 in 4 are food insecure, and 1 in 3 children in tobacco households is stunted. Key differences of tobacco-farming households Higher rate of male decision-makers Dowa Significantly more likely to have a male as the household head and crop decision-maker.

More connected to select assets and financial services Higher mobile phone, bicycle, and bank account ownership.

More agricultural diversification 65 percent work with livestock and average household grows 3 crops.

Stunting rate in areas with tobacco Greater exposure to market shocks 1 activity (% children under 5) 40 percent are affected by high input costs and 75 percent are affected by low output prices, a higher proportion than other 0 35+ agricultural households. Note 1: 1 km x 1 km grids show the proportion of children under five affected by stunting. Areas with a probability of tobacco farming less than 60 percent are shown in gray. Water bodies are shown in blue. Source: Fraym, Fourth Integrated Household Survey 6 EXECUTIVE SUMMARY || VULNERABLE TOBACCO-FARMING HOUSEHOLDS

Generally, tobacco-farming households with smaller farms and limited market access are more vulnerable. Other components of vulnerability vary spatially.

The northernmost tobacco- Chitipa growing areas have seen precipitation decrease more than other tobacco areas, suggesting that drought-resilience should be a key consideration for alternative livelihoods.

Precipitation decrease Mzuzu Soil quality and market access are also challenges in this region. 0 55 mm/decade

Mzimba

Kasungu

Total nitrogen (mg/kg) Dowa 700 2,000 + Vulnerable population

Around Mzimba and Mzuzu, soil quality and 0 1,000+ deforestation pose a potential challenge to the Mchinji viability of both tobacco and potential alternative crops.

Stunting rates above 35 percent are an issue for The central districts with tobacco activity have pockets of this region as well. the highest population density. This translates into high numbers of stunted children and people in vulnerable communities.

Note 1: Main map shows areas with the highest probability of tobacco activity in red, and areas with a probability of tobacco farming less than 60% in gray. Each inset map shows select indicators at a 1 km x 1 km resolution. Water bodies are shown in blue. Source: Fraym, Isric World Soil Information, Malawi Hazards and Vulnerability tool, Fourth Integrated Household Survey 7 EXECUTIVE SUMMARY || DIVERSIFICATION OPPORTUNITIES

Given tobacco’s relatively higher per hectare production costs and falling price, alternative crops like groundnut, soya and other sources of livelihood, including livestock and milk production, offer increasingly competitive income opportunities.

Livestock Nationally, only about 10 percent of tobacco - farming households own cows, but in Chitipa Groundnut District, this number jumps to 25 percent. In Mchinji, Kasungu, and Dowa, smallholder Stunting rates are also lower around Chitipa, tobacco farmers could produce an estimated suggesting that benefits may extend beyond 13,000 metric tons of additional groundnut using income into health and other areas. improved seed varieties and inputs.³ In these districts, per hectare sales could grow from 115,000 MWK to 265,000 MWK.

Soya In Mchinji, Kasungu, and Dowa, smallholder tobacco farmers could produce an estimated 11,000 metric tons of additional soyabean using improved seed varieties and inputs. Per hectare Milk production sales for tobacco farmers in these districts could Only 1 in 5 tobacco-farming households that grow from 125,000 MWK to 375,000 MWK. own cows report producing milk, despite smallholder milk profits being high. Milk profits are particularly high in , where per cow profits1 are over 150,000 MWK per year.

Note 1: Per cow profits are the difference between sale value of all milk produced by a household in the last year and all costs associated directly with milk production, excluding the cost of the cows producing milk. Note 2: Water bodies are shown in blue. Districts with no households reporting activity related to tobacco cultivation, and districts with fewer than ten responses related to tobacco activity are shown in gray. Note 3: Improved yields were estimated based on data from the Feed the Future Soyabean Innovation Lab and input from the Agricultural Transformation Initiative team. Estimated revenue is based on average per hectare soya and groundnut revenue for tobacco-farming households in the three selected districts. Source: Fraym, Fourth Integrated Household Survey 8 I. Overview of tobacco activity

II. Tobacco farmers in context

I. Comparison with non-agricultural households

II. Comparison with other agricultural households Outline III. Farmer segmentation

IV. Investigation of potential alternatives

V. Conclusion and recommendations

VI. Appendices and sources Overview of tobacco activity TOBACCO ACTIVITY || LOCATION

Tobacco activity is primarily concentrated in the northern and central , and about 1 million people live in households engaged in tobacco farming.

Nationally, only about 1 in 20 families cultivates tobacco in home plots. However, they make up 1 nearly one-third of all households in districts like Rumphi Rumphi, Dowa, and Kasungu.

Over 20 percent of the population of Malawi, or 2 about 4 million people, live in areas where Kasungu tobacco is likely cultivated.1 Dowa

Population pressures, weather patterns, and the availability of other commercial crops make 3 tobacco cultivation less desirable in many parts of southern Malawi.

Probability of tobacco activity2

60% 85% + Note 1: Areas where tobacco is likely cultivated are areas with a probability of tobacco activity greater than or equal to 60 percent. Note 2: 1 km x 1 km grids show the estimated probability of tobacco being farmed in the area. Areas with a probability of tobacco farming less than 60 percent are shown in gray. Water bodies are shown in blue. 11 Source: Fraym, Fourth Integrated Household Survey TOBACCO ACTIVITY || SOIL QUALITY

Ensuring that farmers have access to fertilizers is crucial to increasing crop yields in areas with poor soil quality, as measured by nitrogen levels and organic carbon content.1

Too little soil nitrogen limits plant growth. Parts of northern tobacco areas have Mzuzu low levels of total nitrogen in Mzimba soil. Legume crops, such as soybean and groundnut, may improve nitrogen levels. Kasungu

Mchinji

Lilongwe Organic carbon content aids in soil microbe health, leading to increased nutrient availability for crops. Parts of have low soil organic content for the soil type in that area. Total nitrogen (mg/kg)2 Organic carbon (g/kg)3

700 2,000+ 3 20 Note 1: It is important to note that the evaluation of soil quality is a complex science. Total nitrogen and organic carbon are just two of many soil quality metrics. Low nitrogen levels are not a direct indicator of farming viability, as it is also important to also consider carbon to nitrogen levels and soil type in a given area. See Appendix A for more information on soil type. Note 2: 1 km² grids show predictions of soil organic carbon in g/kg at 30-60 cm depth. Areas with probability of tobacco farming less than 60 percent are shown in gray. Note 3: 1 km² grids show predictions of soil total nitrogen content in mg/kg (ppm) at 0-30 cm depth. Sources: Fraym; Isric World Soil Information, Africa Soil Grids, 2015 12 TOBACCO ACTIVITY || DEFORESTATION

Deforestation in tobacco-growing areas is most pronounced in Mzimbadistrict and could be caused by cutting trees to clear land for tobacco farming or for curing tobacco.

Mzimba District Mzuzu The western area of Mzimba district has experienced pockets of deforestation since 2001, 1 but the scale is less dramatic than in certain non- tobacco areas.

Key areas of concern for deforestation are near 2 Mzuzu, along the Northern lake shore, and in pockets of the South.

Loss of tree cover, 2001 - 20141

Low tree cover High tree cover Loss in tree cover detected, 2001 - 2014

Note 1: 1 km x 1 km squares show the extent of tree cover in the year 2001 and whether or not the area has experienced a loss in tree cover, as detected by satellite imagery, in 2014. Tree cover is defined as vegetation taller than 5 meters in height. Water bodies are shown in blue. Areas with a probability of tobacco farming less than 60 percent are shown in gray. Source: Global Forest Watch, Fraym 13 TOBACCO ACTIVITY || ENVIRONMENTAL VARIABILITY

Although less pronounced than elsewhere in Malawi, temperature and precipitation trends in tobacco-growing areas are still concerning for the long-term viability of tobacco cultivation.

Tobacco-growing areas have generally experienced Rumphi slightly above-average Mzuzu increases in temperature compared to the rest of the country. Kasungu Kasungu

Mchinji

Lilongwe The tobacco-growing area around Rumphi city has experienced the most concerning decrease in rainfall over the past three decades.

Temperature increase (degree Celsius), Precipitation decrease (mm/decade), 2002-20141 1981-20142

1 3.5 0 55

Note 1: 1 km x 1 km squares show the level of increase in temperature from 2002 to 2014. Water bodies are shown in blue. Areas with a probability of tobacco farming less than 60% are shown in gray. The source of the temperature trend raster data is the Malawi Hazards and Vulnerability tool, Regional Center for Mapping of Resources for Development, http://tools.rcmrd.org/vulnerabilitytool/. Note 2: 5 km x 5 km squares show the level of decrease in precipitation from 1981 to 2014. Same source as above. 14 Source: Fraym TOBACCO ACTIVITY || VULNERABILITY

Most of the main tobacco-growing areas are vulnerable to climate shocks, but less so than southern Malawi.

Communities in southern Malawi, excluding Community vulnerability is 1 large urban centers such as Blantyre, are calculated using 27 Rumphi generally the most vulnerable. indicators that measure exposure, sensitivity, and adaptive capacity in relation to environmental shocks. Kasungu

In areas with high tobacco activity, communities 2 outside of Rumphi city are relatively less vulnerable than others. Mchinji

Community vulnerability index1

Less vulnerable More vulnerable

Blantyre

Note 1: 1 km x 1 km grids show normalized vulnerability index in that area. Fraym’s customized vulnerability index is constructed using principal component analysis. Water bodies are shown in blue. Areas with a probability of tobacco farming less than 60% are shown in gray. See slide 13 for index methodology. Source: Fraym, Fourth Integrated Household Survey 15 TOBACCO ACTIVITY || VULNERABILITY INDEX METHODOLOGY

The customized FSFW vulnerability index follows the IPCC concept of vulnerability to climate shocks, which is a function of exposure, sensitivity, and adaptive capacity. Fraym mapped available indicators in survey and remote sensing data to the conceptual framework from the accepted vulnerability literature to construct a community-level vulnerability index.

Component Type of indicator¹ Indicator used in vulnerability index

• Percent of community reporting a drought in the last year Hazard events • Percent of community reporting irregular rainfall in the last year Exposure • Percent of community reporting a flood in the last year Change in environmental or climate conditions • Change in average monthly rainfall between 1960-1990 and 2000-2017² • Percent of agricultural households with 2 hectares or less of cultivated land (smallholders) Agricultural practices • Average crop diversification index (1 divided by the number of crops) • Presence of irrigation scheme in community Sensitivity Community structure • Dependency ratio • Percent of households food insecure in last 12 months Food and water security • Percent of households relying on unimproved water source • Presence of a farm support organization in the community Social capital • Percent of agricultural households using extension services • Literacy rate for people ages 15 and older • Percent of households heads with at least primary education Human capital • Percent of female-headed households • Average age of household head • Percent of households that have taken out a loan in the last year for business or farming • Adaptive Average amount borrowed in the last year for business or farming purposes Financial capital • Average net cash farm income capacity • Average total farm size

• Percent of households with access to piped water • Average distance to nearest road • Average time to school Physical capital • Average distance to nearest agricultural market • Percent of households with electricity • Distance to health clinic • Percent of households with a mobile phone

Note 1: The vulnerability index was constructed using principal component analysis. The categories used for the type of indicator are based on the framework in Gbetibouo, Glwadys Aymone and Ringler, Claudia. 2009. Mapping South African Farming Sector Vulnerability to Climate Change and Variability. IFPRI discussion paper – available at: http://www.ifpri.org/publication/mapping-south-african-farming-sector-vulnerability-climate-change-and-variability. Note 2: Change in precipitation comes from remote sensing data, rather than survey data. 16 Source: Fraym, Fourth Integrated Household Survey TOBACCO ACTIVITY || MARKET ACCESS AND STUNTING

Facilitating improved access to reliable markets will be crucial to successful transformation. Diversification into livestock management, particularly milk production, and growing additional crops could also bolster nutrition in tobacco-growing areas.

Market access is highly concentrated in urban areas. In many rural areas with Mzuzu Mzuzu tobacco activity, few households have market access. Kasungu Kasungu

Mchinji

Lilongwe Stunting of children under five is most pronounced in the central region, while the northernmost tobacco- Blantyre farming area has relatively lower child malnutrition.

Market access (% of households)¹ Stunting rate (% of children under 5)²

0 75+ 0 35+ Note 1: 1 km x 1 km grids show the proportion of households with access to a market. Access to a market is defined as being less than 1km from a community market, or less than 10km from a market and having a road that is open year round. This map does not include ADMARC outlets, which are often considered unreliable. Note 2: 1 km x 1 km grids show the estimated stunting rate of children under 5 in the area. Stunting was calculated following WHO standards and is defined as a height- for-age less than two standard deviations from the WHO Child Growth Standards median. 17 Source: Fraym, Fourth Integrated Household Survey TOBACCO ACTIVITY || POPULATIONS AT RISK NATIONALLY Taking into consideration population density and absolute levels of need is vital in targeting programs. Areas with the highest proportion of the population at risk may not be areas with the largest populations at risk.

Nationally, over 11 million people are Nationally, there are vulnerable to about 750,000 climate shocks. stunted children Nationally, nearly 12 million people lack under the age of five. adequate access to markets.

Stunted children¹ Vulnerable population2 Lack of market access3

0 75+ 0 1,000+ 0 1,500+

Map Note: For each map, 1 km x 1 km squares show the population with the listed indicator. Water bodies shown in blue. Note 1: Stunted children refer to children under the age of five whose height for age is two standard deviations below the WHO Child Growth Standards median. Note 2: This map shows the population living in communities more vulnerable than the national average. Areas in white have insufficient data. Note 3: Market access is defined as being less than 1km from a community market, or less than 10km from a market and having a road that is open year round. 18 Source: Fraym, Fourth Integrated Household Survey, LandScan™ TOBACCO ACTIVITY || POPULATIONS AT RISK IN TOBACCO AREAS

The majority of at-risk people living in areas with the highest probability of tobacco activity are in the central region, where population density is relatively greater.

Over 1.5 million people are vulnerable to Over 130,000 climate shocks in Over 2 million children⁴ under five tobacco growing people lack are stunted in areas. adequate access to tobacco growing markets in tobacco areas. growing areas.

Stunted children¹ Vulnerable population2 Lack of market access3

0 75+ 0 1,000+ 0 1,500+

Map Note: For each map, 1 km x 1 km squares show the population with the listed indicator. Areas in gray are areas with a probability of tobacco activity less than 85%. Water bodies shown in blue. Note 1: Stunted children refer to children under the age of five whose height for age is two standard deviations below the WHO Child Growth Standards median. Note 2: This map shows the population living in communities more vulnerable than the national average. Note 3: Market access is defined as being less than 1km from a community market, or less than 10km from a market and having a road that is open year round. Note 4: Values represent only areas with a probability of tobacco activity of at least 85%. 19 Source: Fraym, Fourth Integrated Household Survey, LandScan™ TOBACCO ACTIVITY || TOBACCO PLOT SIZE

The average tobacco-farming household has less than half a hectare of land in their farm under tobacco under cultivation.

Districts with the largest average tobacco plots1 Mzimba Tobacco plot District District size (hectare) Districts with the largest average tobacco plot Kasungu 0.75 1 size are in central Malawi while the smallest tobacco plots tend to be in the far north and Mzimba 0.65 south. Kasungu District Lilongwe 0.55

Average tobacco plot size2 The districts of Phalombe and Zomba in the (hectare) south have the smallest average tobacco plot 2 sizes. These are districts with few households 0.1 0.6 engaged in tobacco activity.

Lilongwe District

Phalombe 3 District

Note 1: Analysis of plot sizes includes only those households that have plots on which they are cultivating tobacco. Note 2: Water bodies are shown in blue. Districts with no households reporting activity related to tobacco cultivation and districts with fewer than ten responses are shown in grey. Average tobacco plot size is calculated using only households that have tobacco plots. Source: Fraym, Fourth Integrated Household Survey 20 Tobacco farmers I. Comparison with non-agricultural households

in context II. Comparison with other agricultural households IN COMPARISON || KEY INDICATORS

Along many measures, tobacco farmers are better off than most other types of farmers, but are still worse off than non-farmer Malawians.

Key indicators

Diverse dietDiet Food insecurity and stunting are prevalent Stunting Raterate across household types. Non-agricultural 1 households have significantly more diverse FoodFood insecureInsecure diets¹ and less food insecurity.

Literate householdHousehold headHead

Bank accountAccount accessAccess Tobacco households generally have a literate, 2 male household head. OwnOwn mobileMobile phonePhone

ElectricityElectricity accessAccess

Female Household Head Female household head Asset ownership, bank account access, and electrification are highest for non-agricultural 0% 20% 40% 60% 80% 3 households. Tobacco households often have greater access to these assets than other Non-tobacco agricultural households agricultural households. Tobacco households NonNon-agricultural-agricultural householdshousehold

Note 1: Dietary diversity measures the diversity, frequency, and nutrient density of food groups consumed by a household. The indicator was developed using methodologies from the World Food Program and USAID. Source: Fraym, Fourth Integrated Household Survey 22 IN COMPARISON || KEY INDICATORS Tobacco Non-tobacco Non-agricultural Indicator households agricultural households households (n = 571) (n = 9,133) (n = 2,743) Household composition Household size 5.8 5.2* 4.9* Female household head 9% 32%* 23%* Total farm area (hectares) 1 1.4 0.6* - Women's empowerment Decision-maker regarding planting of crops is a woman 11% 38%* - Decision-maker regarding use of crop output is a woman 8% 31%* - Decision-maker regarding revenue from crops is a woman 9% 27%* - Assets Households with at least one mobile phone 60% 45%* 75%* Educational attainment and literacy Household head is literate 80% 66%* 85%* Public services provision Electricity access 3% 4% 33%* Bank account 26% 18%* 35%* Nutrition2 30% 24%* 59%* Household dietary diversity (n = 571) (n = 9,133) (n = 2,743) 33% 29% 26% Stunting rate, % of children under 5 years of age (n = 309) (n = 4,850) (n = 1,128) 7% 10% 9% Wasting rate, % of children under 5 years of age (n = 302) (n = 4,749) (n = 1,100) Food security Households food insecure in last 12 months 73% 79%* 49%* Households food insecure in last 7 days 62% 68%* 51%* * Indicates statistics are significantly different from tobacco households at the 95% confidence level.

Note 1: Total farm area refers to all farm land owned by the household including tobacco plots and plots of other crops. Note 2: Stunting and wasting are reported at the individual level, with sample sizes shown beneath statistics. Other indicators are at the household level, and so sample sizes are not equal. 23 Source: Fraym, Fourth Integrated Household Survey IN COMPARISON || DIETARY DIVERSITY Tobacco household diets are more diverse than other agricultural households, but not as diverse as non-agricultural households.

Frequency of food group consumption per week1 Non-tobacco Non- Tobacco agricultural agricultural households Tobacco households consume significantly less Food group households households 1 (n = 571) animal products than non-agricultural households. (n = 9,133) (n = 2,743) Main staples 7 7 7 Pulses 2.3 2.2 2.0* Vegetables 5.9 5.5* 5.9

On average, each household group consumes at Fruit 1.6 1.3 1.9 2 least one main staple per day, and similar amounts Animal protein2 2.2 2.1 3.4* of pulses, fruits, and vegetables. Dairy 0.7 0.5 2.0* Sugar 3.2 2.5* 5.3* Oil 3.7 3.3* 5.3* Food diversity scores Increasing household access to affordable animal Average unweighted products, both protein and dairy, would help 3 food consumption 26 24* 33* increase dietary diversity among tobacco and score3 agricultural households. Average weighted food 40* 54* 43 consumption score4 * Indicates significant difference from tobacco households at the 95% confidence level.

Note 1: A food group can be consumed more than once per day, but the maximum consumption frequency of each food group is capped at seven per week. Note 2: Animal protein includes eggs, fish, and meat, but does not include milk or other dairy products. Note 3: Unweighted food consumption score is a measure of total number of different food groups consumed in a week. The maximum score is 56 for a week. Note 4: Weighted food consumption score is a measure that gives higher weights to more nutrient dense foods. The maximum score is 112 for a week. See appendix for full methodology, including weights for each food group, and extended information on the dietary diversity indicator. Source: Fraym, Fourth Integrated Household Survey 24 IN COMPARISON || DIETARY DIVERSITY Over fifty percent of all food groups consumed in non-diverse diet households comes from main staples, sugar, and oil.

Frequency of food group consumption per week1 among all households Diverse diet Not diverse diet Food group On average, households with diverse diets (n = 8,250) (n = 4,196) 1 consume foods from each food group at least Main staples 7 6.9* twice per week. Pulses 3.7 1.8* Vegetables 6.5 5.7* Fruit 2.5 1.2* Animal protein2 3.9 1.6* In households with non-diverse diets, less than 2 ten-percent of consumed come from animal Dairy 2.1 0.1* protein and dairy. Sugar 5.4 2.3* Oil 5.3 3.0* Food diversity scores Unweighted food consumption 36.4 22.5* Increasing consumption of nutrient dense foods, score3 such as animal products and pulses, would help Weighted food consumption 63.4 35.3* 3 increase dietary diversity, as households with non- score4 diverse diets consume significantly less animal * Indicates statistics are significantly different from tobacco households at the 95% protein, dairy, and pulses. confidence level.

Note 1: A food group can be consumed more than once per day, but the maximum frequency of any food group is capped at seven per week. Note 2: Animal protein includes eggs, fish, and meat, but does not include milk or other dairy products. Note 3: Unweighted food consumption score is a measure of total number of different food groups consumed in a week. The maximum score is 56 for a week. Note 4: Weighted food consumption score is a composite measure that gives higher weights to more nutrient dense foods. The maximum score is 112 for a week. See appendix for full methodology and extended information on dietary diversity indicator. Source: Fraym, Fourth Integrated Household Survey 25 IN COMPARISON || LABOR AND PRODUCTION

Tobacco households have higher farm profits, costs, and sales per hectare and spend more time on agriculture and livestock activities than other farming households.

Hours worked¹ Annual costs, sales, and spending²

Tobacco household Despite higher per Wage members typically Household spending on food hectare profits, work fewer hours per tobacco-farming week but spend more households have Labor time on farming and similar outcomes in livestock activities. Farm costs per hectare areas like food There are periods of spending and Livestock intense agricultural nutrition. activity and periods of Farm profit per hectare Agriculture idleness.

Farm sales per hectare Total

0 5 10 15 20 - 50 100 150 200 250 Hours worked per worker per week MWK (thousands)

Non-tobacco agricultural households Tobacco households Non-tobacco agricultural households Tobacco households

Note 1: Data on labor comes from a seven day recall of the number of hours worked in the past week. Statistics above do not include individuals who reported working zero hours in the past seven days. See Appendix E for more information about the timing of data collection. Note 2: Data on farm sales and profits only include households that are engaged in agricultural activities and made sales. Cost data is only for households that had costs. Farm costs are defined as any cost associated with farming, and includes seeds, inputs, labor, and transportation. Sales and profit data do not include households identified as price outliers, defined as household that reported per-unit crop sales three standard deviations above or below that crops national mean. Source: Fraym, Fourth Integrated Household Survey 26 IN COMPARISON || CHILD LABOR

About 34 percent of tobacco-farming households rely on their children for some amount of labor, compared to about 26 percent of other agricultural households.

Non-tobacco Child hours worked¹ Tobacco Per child working in agriculture household household1 household (n = 433) Although children work (n = 6,170) Labor similar numbers of hours per Total hours worked 6.5 7.2 week across households, a Hours worked in 3.4 3.5 higher proportion of tobacco farming households rely on children Hours worked in 2.1 2.1 Livestock for some amount of labor livestock Hours worked in every week. 0.65 1.1 manual labor Non-tobacco Tobacco Per household agriculture Agriculture household household Percent of households 34% 26%* using child labor2 (n = 433) (n = 6,170) Total Average number of child workers among 1.5 1.4 households using child (n = 142) (n = 1,635) 0 1 2 3 4 5 6 7 8 labor Hours worked per child worker per week * Indicates statistics are significantly different from tobacco households at the 95% confidence level. Non-tobacco agricultural households Tobacco households

Note 1: Statistics are the average hours worked per working child the household. Data on labor comes from a seven day recall of the number of hours worked in the past week. Child labor is defined as any person aged 5-14 who reported working at least 0.5 hours in the last seven days, and does not include fetching water or firewood. Statistics only include households that have children. Age range of 5-14 years is in line with age ranges used by the U.S. Bureau of International Labor Affairs in their estimation of child labor in Malawi. Note 2: Statistics only include households that have children. Source: Fraym, Fourth Integrated Household Survey 27 IN COMPARISON || FARM STATISTICS

Farm statistics by household type Production, sales, and costs1 Tobacco household Non-tobacco agriculture household

Farm profit per hectare (2017 Malawi Kwacha) 170,000 70,000* (n = 560) (n = 3,710) 250,000 105,000* Farm sales per hectare (2017 MWK) (n = 533) (n = 3,710) 70,000 44,000* Farm costs per hectare (2017 MWK) (n = 568) (n = 7,415) 1,260 1,130 Kilograms of crops harvested per hectare (n = 571) (n = 9,048) 715 650 Kilograms of crops sold per hectare2 (n = 527) (n = 3,215) 1.4 0.6* Total plot area (hectares) (n = 571) (n = 9,057) 70,000 67,000 Fertilizer cost per hectare (2017 MWK)2 (n = 456) (n = 9,057) Tobacco household Non-tobacco agriculture household Crop usage and nutrition (n = 571) (n = 9,057) Number of crop types planted 3.3 2.2* Number of own crops consumed 3.3 2.2* Percent of households consuming own crops 90% 80%* Percent of households with diverse diets 30% 24%* Annual household spending on food (2017 MKW) 55,000 60,000 33% 29% Stunting rate, % of children under 5 years of age3 (n = 309) (n = 4,850) * Indicates statistics are significantly different from tobacco-farming households at the 95% confidence level.

Note on statistics and sample sizes: Discrepancies in sample sizes, in particular sales and cost data, are due to non-answers, or reposes of zero, on the part of respondents. Note 1: Data on farm sales and profits only include households that are engaged in agricultural activities and made sales. Cost data is only for households that had costs. Fertilizer costs only include households that purchased fertilizer. Farm costs are defined as any cost associated with farming, and includes seeds, inputs, labor, and transportation. Note 2: Fertilizer cost per hectare is higher than farm cost per hectare for as not all households that had farm costs purchased fertilizer, which tended to be the highest farm cost. Note 3: Stunting rate refers to the stunting rate of all children under five within each household type. A 33% stunting rate for tobacco households means that 33% of children in tobacco households are stunted. Source: Fraym, Fourth Integrated Household Survey 28 IN COMPARISON || SHOCKS

Tobacco households are more exposed to market fluctuations, which highlights the importance of improving market reliability and predictability for alternative crops.

Households experiencing negative shocks (percent)

Flood Around 40 percent of tobacco households were negatively affected by low output prices and Crop disease 1 about 75 percent were negatively affected by high input costs in the last year. Loss of other income

Low price of output Tobacco-farming households in central Malawi are the most likely to report challenges with 2 market fluctuations. Nearly 80 percent report Drought issues with high input costs, and nearly 45 percent report issues with low output prices. Irregular rains

Both types of agricultural households are highly High cost of input exposed to environmental shocks, of which 3 irregular rains are the most common issue. Any environmental shock Tobacco-farming households in northern Malawi are the least likely to experience drought. 0% 20% 40% 60% 80% 100% Non-tobacco agricultural households Tobacco households

Source: Fraym, Fourth Integrated Household Survey 29 IN COMPARISON || SHOCKS BY HOUSEHOLD TYPE

Percent of households negatively affected by each shock in last 12 months

Tobacco households¹ Non-tobacco agricultural households Shock type (n = 571) (n = 9,075)

Drought 50% 42%*

Flood 2% 5%*

Crop disease 11% 10%

Low price of output 39% 15%*

High cost of input 75% 58%*

Loss of other income 13% 13%

Irregular rains 70% 70%

Any environmental shock² 88% 85%

* Indicates statistics are significantly different from tobacco-farming households at the 95% confidence level.

Note 1: Tobacco households include only those households with a plot used for tobacco cultivation. Note 2: Environmental shocks include drought, floods, earthquakes, unusually high crop or livestock disease, irregular rains, and landslides. Source: Fraym, Fourth Integrated Household Survey 30 IN COMPARISON || SHOCKS BY REGION

Percent of tobacco-farming households¹ negatively affected by each shock per region

Central North South Shock type (n = 363) (n = 168) (n = 40)

Drought 55% 22%* 60%

Flood 1% 6%* 9%

Crop disease 11% 9% 14%

Low price of output 44% 26%* 25%*

High cost of input 79% 68%* 54%*

Loss of other income 14% 11% 8%

Irregular rains 68% 72% 76%

Any environmental shock² 88% 84% 95%

* Indicates statistics are significantly different from the central region at the 95% confidence level.

Note 1: Tobacco households include only those households with a plot used for tobacco cultivation. Note 2: Environmental shocks include drought, floods, earthquakes, unusually high crop or livestock disease, irregular rains, and landslides. Source: Fraym, Fourth Integrated Household Survey 31 IN COMPARISON || FARM SIZE

Tobacco households have an average of 1.4 hectares of farmland, which is more than other farmers. This underscores the high potential for producing alternative crops.

Distribution of total farm size Distribution of total farm size (tobacco farms) (non-tobacco farms) 60% 60% 50% 50% tobacco

40% - 40% 30% 30% 20% 20% On average, tobacco farmers have more land to

plant more crops. Tobacco farmers plant three 10% Percent of non agricultural households 10% 1 different crops, whereas non-tobacco farmers 0% 0% 0 0.5 1.0 1.5 2.0 + 0 0.5 1.0 1.5 2.0 +

plant two. Percent of tobacco households Hectares Hectares Small tobacco farm Midsize tobacco farm (farm size < 2 ha) (farm size > 2 ha) 85% 15% % of tobacco households Using a threshold of two hectares highlights (n = 486) (n = 85) significant differences in tobacco households. Number of different crops 3.2 3.7 2 Tobacco smallholders are notably more food Household head is female 10% 5% insecure than tobacco farmers with larger farms. Household head is literate 78% 90% Household food insecure in 77%* 52%* last 12 months Annual per capita household 155 223 spending (nominal USD) Decision-maker regarding 12%* 4%* crops is female * Indicates statistics are significantly different at the 95% confidence level. 32 Source: Fraym, Fourth Integrated Household Survey IN COMPARISON || DEFINING TOBACCO SMALLHOLDERS There is little variation in households characteristics for tobacco farmers with less than two hectares of farmland.

Household characteristics of tobacco smallholders1 Tobacco farm Tobacco farm While the threshold of one hectare to define less than 1 ha 1 – 2 ha smallholder farmers may be fitting for non- 1 43% 42% tobacco farms, it is less applicable to tobacco % of tobacco households farms as they tend to be larger. (n = 248) (n = 238) Total number of crops planted 3.0 3.3 Household head is female 11% 8% Tobacco farmers with less than one hectare of Household head is literate 77% 79% land and those with between one and two Household food insecure in last 12 2 80% 75% hectares of land have no statistically significant months differences in demographic and economic Annual per capita household characteristics. 162 150 spending (nominal USD) Decision-maker regarding crops is 14% 10% female

3

Note 1: No household characteristics are statistically different between tobacco farm size groups at the 95% confidence level. Source: Fraym, Fourth Integrated Household Survey 33 Farmer segmentation FARMER SEGMENTATION || CROP GROUPS

Fraym used a statistical technique¹ to categorize agricultural households into five different groups based on their crop mixture.

Tobacco households also growing maize and either beans, groundnut, or nkhwani: 4% of agricultural households

Other tobacco households, often growing maize and fewer than three crops: 2% of agricultural households

Non-maize households, primarily growing rice, sorghum, or cassava: 7% of agricultural households

Maize households growing nandolo or nkhwani: 45% of agricultural households

Most prevalent group by district Other maize households, growing less common crops Non-maize or no second crop Maize and nandolo or nkhwani 41% of agricultural households Other maize

Note 1: Latent class analysis is a statistical modeling method used to create mutually exclusive and exhaustive groups. Source: Fraym, Fourth Integrated Household Survey 35 FARMER SEGMENTATION || CROP GROUPS COMPOSITION

Percent of households in group growing major crops Proportion of Household crop agricultural group households Sweet Tobacco Maize Groundnut Rice Beans Nandolo Nkhwani Soya Sorghum Cassava potato

Tobacco household with 0.04 maize and select 100% 100% 36% 0% 32% 1% 74% 2% 22% 1% 1% (n = 357) legumes or nkhwani

Other tobacco 0.02 100% 88% 1% 0% 0% 9% 4% 2% 11% 2% 1% household (n = 214)

Non-maize 0.07 0% 0% 7% 27% 1% 1% 1% 3% 5% 25% 27% household (n = 902)

Maize household 0.45 with nandolo or 0% 100% 13% 4% 11% 51% 68% 2% 8% 10% 7% (n = 4,625) nkhwani

Other maize 0.41 0% 100% 18% 2% 16% 0% 0% 2% 11% 5% 4% household (n = 3,764)

Source: Fraym, Fourth Integrated Household Survey 36 FARMER SEGMENTATION || CROP GROUPS MAPPING

Only a handful of districts have more than 10 percent of agricultural households in either type of tobacco group.

Proportion of agricultural households in each crop group by district

Tobacco, maize and groundnut/beans/nkhwani

0 0.1+

Proportion of agricultural households in each crop group by district

Other tobacco

0 0.1+

Source: Fraym, Fourth Integrated Household Survey 37 FARMER SEGMENTATION || CROP GROUPS MAPPING

Relatively few non-maize households are located in tobacco-growing areas, whereas both maize groups have higher concentrations in tobacco areas.

Proportion of agricultural households in each crop group by district Non-maize Maize and nandolo/nkhwani Other maize

0 0.5+ 0.05 0.75+ 0.05 0.75+

Source: Fraym, Fourth Integrated Household Survey 38 FARMER SEGMENTATION || GROUP 1: TOBACCO, MAIZE, AND SELECT CROPS

Tobacco, maize and select crop households have the largest farms, at 1.4 hectares on average, with the most crop diversity across the five groups. Despite high transport and per hectare costs, this group has high cash sales per hectare and consumes the largest variety of own-produced crops. Large, young homes Around five household members, with more children than other groups Most diversification through livestock Over 70 percent work with livestock, more than any other crop group Led by literate men Literate male household head, with men in charge of crop decisions Farthest from agricultural market Despite being, on average, more than 33 Higher low-end asset ownership kilometers from an agricultural market, more likely Mobile phone and bicycle ownership around 60 to sell crops at market than other households percent, higher than non-tobacco groups

Most reliably connected communities Most exposed to market shocks 55 percent live in a community with a paved road, 50 percent were affected by the low price of more than any other crop group outputs and over 80 percent were affected by high input costs in the last year

More financially included Lack of farm inputs is also a top cause of food At around 30 percent, most likely to have a bank insecurity account

Note: Highlighted differences are statistically significant at the 95% confidence level. 39 Source: Fraym, Fourth Integrated Household Survey, Icons by Noun Project FARMER SEGMENTATION || GROUP 2: OTHER TOBACCO

Other tobacco households also have relatively large farms, 1.2 hectares on average. Despite high transportation and per hectare costs, this group has high per hectare cash sales.

Large homes Diversification through livestock Around 5 household members, larger than non- Around 50 percent also work in livestock, more tobacco groups than non-tobacco households

Led by literate men Far from agricultural market Literate male household head, with men in charge of An average of 31 kilometers from the nearest crop decisions agricultural market

More exposed to market shocks Higher low-end asset ownership Low output prices affected 25 percent and high Over 50 percent own a mobile phone and around input costs affected over 60 percent, more than 70 percent own a bicycle, higher than other groups non-tobacco households

Most likely to be food secure Most access to tobacco club Around 67 percent experienced food insecurity in Over 60 percent live in a community with a the last year, lower than any other crop group tobacco club, more than any other group

Note: Highlighted differences are statistically significant at the 95% confidence level. 40 Source: Fraym, Fourth Integrated Household Survey, Icons by Noun Project FARMER SEGMENTATION || GROUP 3: NON-MAIZE

Non-maize households have the smallest harvests per hectare and cultivate the fewest number of crops. Nonetheless, they have have high cash sales per hectare and handle the highest number of transactions to sell crops.

Smaller farms and homes Closest to roads Fewer than 5 household members and a farm size of On average, less than 7.5 kilometers from the around 0.5 hectares nearest main road on average, closer than other crop groups

Most access to youth group 65 percent live in a community with a youth group, more than any other crop group Far from ADMARC outlet An average of almost 10 kilometers from the nearest ADMARC outlet Least storage and consumption of crops Store the least amount of crops per hectare and are the least likely to consume own crops

Different food insecurity coping Least exposed to market shocks Less than 10 percent affected by low output prices 30 percent cope with food insecurity by restricting and around 35 percent affected by high input adult food consumption, more than other groups costs, less than any other crop group

Note: Highlighted differences are statistically significant at the 95% confidence level. 41 Source: Fraym, Fourth Integrated Household Survey, Icons by Noun Project FARMER SEGMENTATION || GROUP 4: MAIZE AND NANDOLO OR NKHWANI

Maize households growing nandolo or nkhwani typically have smaller farms of roughly 0.5 hectares. This group has low per hectare cash costs and low per hectare cash sales.

Most likely to be led by a woman Diversification through livestock 35 percent have a female head of household and Close to 40 percent also work with livestock, more 40 percent have a woman as decision-maker than other non-tobacco households regarding crops, more than any other crop group

Least seasonal work Closest to agricultural market Under 60 percent live in a community where people come to find work for part of the year, less An average of around 20 kilometers from the than other crop groups nearest market, less than any other crop group

Most access to MASAF work program Lowest household expenditure More than 70 percent live in a community with a Households spend an average of 525,000 MWK MASAF work program, more than other crop annually, lower than other crop groups groups, especially tobacco households

Note: Highlighted differences are statistically significant at the 95% confidence level. 42 Source: Fraym, Fourth Integrated Household Survey, Icons by Noun Project FARMER SEGMENTATION || GROUP 5: OTHER MAIZE

Other maize households generally have smaller farms of around 0.5 hectares, with low cash costs and sales per hectare. Compared to the other maize-focused crop groups, these households store the most crops per hectare.

Smaller households Less exposed to drought Fewer than 5 household members and an average of Around 37 percent were affected by drought in the 2 children last year, less than other maize-focused crop groups

More likely to be led by a woman More likely to live near daily market 30 percent have a female head of household and 35 percent live in a community with a daily market, 35 percent have a woman as decision-maker around double the percentage of tobacco regarding crops, more than tobacco-farming households households

Closest to population center Higher household expenditure Households spend an average of 694,000 MWK At an average of 35 kilometers, this group is closest annually, higher than other non-tobacco maize to a population center with at least 20,000 people households

Note: Highlighted differences are statistically significant at the 95% confidence level. 43 Source: Fraym, Fourth Integrated Household Survey, Icons by Noun Project FARMER SEGMENTATION || COMPARISON TABLES

Tobacco, maize Maize and Other tobacco Non-maize Other maize Household crop group and select crops nandolo/nkhwani (n = 214) (n = 902) (n = 3,764) (n = 357) (n = 4,625)

Household composition Average household size* 5.05 5.07 4.22 4.37 4.40 Female household head* 10% 7% 26% 35% 30% Women's empowerment Decision-maker regarding planting of crops is a woman* 11% 10% 36% 44% 39% Decision-maker regarding use of crop output is a woman* 10% 6% 26% 37% 35% Decision-maker regarding revenue from crops is a woman* 10% 7% 20% 36% 27% Nutrition 37% 26% 25% 29% 29% Stunting rate, percent of children under 5 (n = 190) (n = 119) (n = 479) (n = 2,440) (n = 2,000) 9% 4% 10% 9% 12% Wasting rate, percent of children under 5 (n = 186) (n = 116) (n = 460) (n = 2,397) (n = 1,959) Household has a diverse diet* 30% 24% 24% 24% 27% Wealth and assets Average annual household spending (nominal USD 2017)* 812 904 707 722 955 Household owns at least one mobile phone* 59% 54% 34% 40% 42% Household owns a bicycle* 59% 68% 44% 37% 37% Highest educational attainment and literacy Household head is literate* 81% 78% 66% 66% 66% Household head has no education* 9% 13% 21% 20% 20% * Indicates that there are statistically significant differences in group means at the 95% confidence level.

44 Source: Fraym, Fourth Integrated Household Survey FARMER SEGMENTATION || COMPARISON TABLES

Tobacco, maize Maize and Other tobacco Non-maize Other maize Household crop group and select crops nandolo/nkhwani (n = 214) (n = 902) (n = 3,764) (n = 357) (n = 4,625) Community characteristics MASAF¹ work program present in community* 50% 60% 61% 72% 69% People leave community temporarily in search of work 85% 86% 85% 84% 87% People come to community temporarily in search of work* 77% 73% 66% 57% 66% Youth group present in community* 35% 39% 65% 47% 50% Tobacco club present in community* 46% 61% 17% 23% 29% Household shocks in last 12 months Negatively affected by high input costs* 48% 25% 8% 18% 13% Negatively affected by low output prices* 83% 62% 36% 64% 55% Negatively affected by drought* 53% 44% 39% 47% 37% Farm characteristics Average farm size (hectares)* 1.4 1.2 0.5 0.6 0.6 Average number of different types of crops* 3.8 2.3 1 2.8 1.7 10,000 11,300 6,000 2,000 3,600 Average per hectare transportation cost (MWK)2* (n = 349) (n = 200) (n = 317) (n = 3,616) (n = 2,838) 77,000 80,000 34,000 47,000 56,000 Average per hectare farm cost (MWK)3* (n = 349) (n = 200) (n = 316) (n = 3,616) (n = 2,838) 250,000 270,000 221,000 90,000 104,000 Average cash sales (MWK)* (n = 336) (n = 197) (n = 259) (n = 2,053) (n = 1,398) 3.3 4.8 2.5 2.3 2.4 Average number of transactions to sell crops4* (n = 338) (n = 199) (n = 273) (n = 2,003) (n = 1,382) Average per hectare harvest (kilograms)* 1,500 1,300 575 1,260 1,100 Household sells crops at market* 96% 93% 32% 45% 40% * Indicates that there are statistically significant differences in group means at the 95% confidence level.

Note on sample size: Each indicator is at the household level, and each sample size is that listed in the top column unless otherwise noted. Note 1: MASAF refers to the Malawi Social Action Fund. Note 2: Average per hectare transportation cost refers to any transportation cost associated with farming. Only households that had transportation costs were included. Note 3: Only households that had farm costs were included. Note 4: Households outside +/- 3 standard deviations from average number of transactions were not included in these statistics. Source: Fraym, Fourth Integrated Household Survey 45 FARMER SEGMENTATION || COMPARISON TABLES

Tobacco, maize Maize and Other tobacco Non-maize Other maize Household crop group and select crops nandolo/nkhwani (n = 214) (n = 902) (n = 3,764) (n = 357) (n = 4,625)

Household livelihood 50% 31% 23% 27% 22% Percent of population working with livestock1* (n = 1,560) (n = 948) (n = 3,504) (n = 17,595) (n = 14,401) Percent of households consumes at least some of own 92% 86% 71% 86% 80% crops* Percent of households stores some crops* 27% 20% 4% 11% 12% Food security Household food insecure in last 12 months* 73% 67% 77% 79% 75% Household food insecure in last 7 days* 64% 56% 74% 69% 63% Household restricts adult food consumption to allow children 15% 19% 33% 24% 19% to eat* 43% 35% 11% 21% 22% Lack of farm inputs is a top cause of food insecurity (n = 259) (n = 143) (n = 696) (n = 3,668) (n = 2,811) Connectivity Average distance to agricultural market (km)* 33.2 31.2 24.1 21.9 26.6 Average distance to ADMARC2 outlet (km)* 9.2 7.6 7.9 8.4 8.5 Average distance to road (km)* 12.7 13.4 8.0 11.2 9.7 Daily market in community* 21% 13% 39% 31% 35% Main road near community is paved with asphalt or gravel* 55% 35% 41% 33% 41% Average distance to population center with 20,000 people 44.7 42.7 37.9 37.3 35.1 (km)* Household has a bank account* 28% 23% 18% 18% 18% * Indicates that there are statistically significant differences in group means at the 95% confidence level.

Note on sample size: Each indicator is at the household level, and each sample size is that listed in the top column unless otherwise noted. Note 1: This measure is at the individual level, and measures the percent of the total population who reported working at least half an hour on livestock related work in the past seven days. Note 2: ADMARC refers to the Agricultural Development and Marketing Corporation. Source: Fraym, Fourth Integrated Household Survey 46 FARMER SEGMENTATION || HIGHLIGHTS

Similarities between tobacco-farming households and neighboring crop groups emphasize the potential to multiply program impact.

Some differences between the two tobacco-farming groups may be driven by the tobacco, maize, and select crop group 1 having a larger farm on average. A larger farm may enable those households to grow a greater variety of crops, including soya and groundnut, which typically require a separate plot.

The maize and nandolo or nkhwani group and the other maize group are the primary crop groups in areas with tobacco 2 activity. Non-maize households are largely concentrated in the south and along the lake shore where there are few tobacco-farming households.

At least 10 percent of households in both non-tobacco maize groups grow groundnut, soya, and beans each. Efforts to 3 improve yield and agricultural practices related to such alternative crops for tobacco-farming households could have positive spillovers on neighboring farmer groups.

Tobacco-farming households share many challenges with other farmer groups in their communities, including high stunting rates, limited dietary diversity, widespread food insecurity, and limited access to reliable markets. This 4 underscores how programs for tobacco-farming households that focus on shared challenges may have a much greater impact on communities.

Source: Fraym 47 Investigation of potential alternatives POTENTIAL ALTERNATIVES || CURRENT DIVERSIFICATION

Tobacco households are generally diversified beyond tobacco. The majority grow maize and one or more legumes, suggesting the potential to leverage existing knowledge and market linkages to reduce reliance on tobacco.

44% also grow nkhwani Nkhwani (n = 258)

Groundnut 21% also grow groundnut (n = 121) Tobacco

Maize (n = 549) Beans 21% also grow beans (n = 119)

Soya 19% also grow soya (n = 105)

Source: Fraym, Fourth Integrated Household Survey 49 POTENTIAL ALTERNATIVES || MARKET-FACING CROPS

Geographic feasibility, revenue potential, and ease of transition are crucial factors for identifying promising alternative cash crops to tobacco.

Non-tobacco market-facing crops1

Pigeon pea

Geographic feasibility: Groundnut, Groundnut 160,000 MWK sales per ha soyabean, beans, sweet potato, and Beans 70,000 MWK sales per ha 1 sunflower are currently grown in tobacco areas. Sorghum Soya 140,000 MWK sales per ha Rice Revenue potential: Tobacco crops bring in Peas an average of 600,000 MWK in sales per 2 hectare planted. This is significantly higher Sweet potato 190,000 MWK sales per ha than any other market-facing crop. Sunflower 60,000 MWK sales per ha Cotton Ease of transition: 20 percent of tobacco Pearl millet farming households are currently growing groundnut, beans, and/or soya. Of these crops, 0% 5% 10% 15% 20% 25% 30% 3 groundnut and soya have the highest per hectare sales. Percent of farming households growing crop for sale Percent of farming households growing crop for consumption

Note 1: Stacked bar chart shows two distinctive categories of farming households: 1) households growing a crop with the intention to sell at least part of the harvest and 2) households growing a crop for home consumption without the intention to sell. A household can grow multiple market-facing crops. Several crops were removed due to small sample size, such as sugarcane, paprika, onion, and tomato. A household is considered as growing a crop with the intention to sell if they responded yes to selling at least some of their harvested crop in the survey. Some households sold all of their harvested crop without any home consumption, while others had a mix of sales and home consumption. Source: Fraym, Fourth Integrated Household Survey 50 POTENTIAL ALTERNATIVES || EXPANDING CURRENT DIVERSIFICATION

Soya and groundnut have the highest revenue potential1 of market-facing crops currently being grown by tobacco-farming households.

Percent of agricultural households growing Soya Groundnut Beans

140,000 MWK/ha 160,000 MWK/ha 70,000 MWK/ha in sales revenue in sales revenue in sales revenue

0% 25% + 0% 25% + 0% 25% +

Note 1: Revenue potential is evaluated by assessing current sales revenue per hectare planted. 51 Source: Fraym, Fourth Integrated Household Survey POTENTIAL ALTERNATIVES || EXPANDING CURRENT DIVERSIFICATION

Sweet potato is less common than other market-facing crops, but has a high revenue potential.1

Percent of agricultural households growing Sweet potato Sunflower Pigeon peas

190,000 MWK/ha 60,000 MWK/ha 70,000 MWK/ha in sales revenue in sales revenue in sales revenue

0% 10% + 0% 10% + 0% 40% +

Note 1: Revenue potential is evaluated by assessing current sales revenue per hectare planted. 52 Source: Fraym, Fourth Integrated Household Survey INCLUSIVE BUSINESS MODELS || SOYA CASE STUDY

With improved yield, soya production by smallholder tobacco farmers in three central districts could grow to over 12,000 metric tons under current land use.

Estimated soya production by smallholder tobacco In Mchinji, Dowa and Kasungu districts, farmers in Mchinji, Dowa, and Kasungu § Smallholder tobacco farmers currently produce 40 1 35 around 3,000 MT of soya, with an average yield 30 of 0.6-0.8 ton/ha. 25 20 15 With support for agricultural transformation that 10 helps to improve yields2 to around 2 tons/ha,

Metric tons (thousand) 5 - § An additional 9,000 - 12,000 MT of soya could Current Improved yield Potential* be produced in these districts, assuming the same improved yield land use for smallholder tobacco farmers. *Potential assuming a 50% reallocation of tobacco land to soyabean Farmers earn 550,000 MWK per ha of tobacco and 125,000 MWK per hectare of soya in these districts. If demonstrated viability under improved yield facilitates a transition of 50% of smallholder tobacco With new seed varieties, extension services, and market land in these districts to soya, facilitation to ensure a stable price, farmer earnings could reach 375,000 MWK per hectare of soya. With lower per § Around 35,000 MT of additional soya could be hectare costs than tobacco, this may be a competitive income produced. opportunity.

Note 1: Production is estimated using population data in combination with estimations of average yield and plot size from survey data in the three districts. MT refers to metric tons. Note 2: Improved yield potentials for Kasungu district from the Feed the Future Soyabean Innovation Lab trials. 53 Source: Fraym, Fourth Integrated Household Survey INCLUSIVE BUSINESS MODELS || GROUNDNUT CASE STUDY

With improved groundnut yield, smallholder tobacco farmers in these districts could produce over 18,000 metric tons under current land use.

Estimated groundnut production by smallholder In Mchinji, Dowa and Kasungu districts, tobacco farmers in Mchinji, Dowa, and Kasungu § Smallholder tobacco farmers currently produce 60 around 6,000 MT1 of groundnut, with an 50 average yield of 1.2-1.4 ton/ha. 40 30 20 With support for agricultural transformation that helps to improve yields to around 3 tons/ha, 10 Metric tons (thousand) - § An estimated 12,000 – 14,000 MT of additional Current Improved yield Potential* groundnut could be produced in these districts, improved yield assuming land use remains the same. *Potential assuming a 50% reallocation of tobacco land to groundnut

Farmers earn 550,000 MWK per ha of tobacco and 115,000 After demonstrating viability, if tobacco farmers MWK per hectare of groundnut in these districts. decide to use 50% of tobacco land for groundnut, With new seed varieties, extension services, and market § Over 50,000 MT of additional groundnut could facilitation to ensure a stable price, farmer earnings could be produced in these districts. reach 265,000 MWK per hectare of groundnut. With lower per hectare costs than tobacco, this may be a competitive income opportunity.

Note 1: Production is estimated using population data in combination with estimations of average yield and plot size from survey data in the three districts. The Agricultural Transformation Initiative team provided input on the appropriate estimate for improved groundnut yield based on Feed the Future data. Source: Fraym, Fourth integrated Household Survey 54 POTENTIAL ALTERNATIVES || LIVESTOCK

Outside of alternative crops, there are other livelihoods, like dairy production, that offer opportunities for both diversification and improved health outcomes.

Dairy statistics for smallholder agricultural households1

About 11 percent of smallholder tobacco- 95% Confidence Number of farming households own at least one cow, a Interval observations 1 slightly higher proportion than other Household owns at least one cow 6% - 8.5% 4,738 smallholder farming households. Total number of cows per household2 3.5 – 5 322 Percent of households producing milk 25% - 40% 322 Average liters of milk produced per month3 56 – 120 102 Households with cows produce about 40 Average liters of milk produced per month per cow 20 – 60 101 liters of milk per cow per month. As only 15 percent of tobacco-farming households Percent of households selling milk 15% - 25% 322 2 consume dairy more than once per week, Value of milk sold in last 12 months (MWK) 53,000 – 163,000 61 this could present a significant opportunity Value of milk sold in last 12 months 24,000 – 100,000 61 to improve dietary diversity and nutrition. per cow (MWK) Costs associated directly with milk production 0 – 2,700 104 (MWK)4 For households that sold at least some milk, Costs associated directly with milk production per 0 – 1,365 103 the average profit per cow per year is about cow (MWK) 3 60,000 MWK, excluding costs not directly Milk profits (MWK)5 53,000 – 160,000 61 associated with milk production. Milk profits per cow (MWK) 24,000 – 96,000 61 Number of times household consumes dairy per 1.5 – 2 322 week Percent of households consuming dairy daily 8% - 16% 322

Note 1: Statistics only include households that reported having at least one cow in the last year, except for the first statistic which only includes smallholder tobacco and non-tobacco agricultural households. Smallholder is defined as having a total farm size less than two hectares. Note 2: This statistic measures the number of cows that household has had in the past year, including those that have been sold. Note 3: Average liters of milk produced per month is based on farmer recall on the average amount of milk produced per month in the past 12 months. Note 4: Costs associated directly with milk production only include direct costs for milk production, such as additional inputs, transportation, and labor. This measure does not include the cost of feed, veterinary care, hired labor directly related to a dairy cow, nor does it include the cost of the cow itself. Note 5: Milk profits only include the cash sale value of milk, minus the costs directly associated with milk production discussed above. 55 Source: Fraym, Fourth Integrated Household Survey Conclusion and Recommendations CONCLUSION || KEY CHALLENGES AND OPPORTUNITIES

While there are many shared challenges across agricultural households, key differences in tobacco-farming households point to potential high-impact areas.

There are many shared challenges across agricultural households, including malnutrition, poor infrastructure and 1 market access, soil quality and deforestation concerns, and exposure to environmental shocks like irregular rain.

Key differences that present a challenge for tobacco households include high per hectare costs and high exposure to 2 market fluctuations in input costs and output prices. Tobacco farmers have higher literacy, own more basic assets, and are more involved in livestock activities, presenting unique opportunities for economic transition.

There are many areas with the potential to impact vulnerable tobacco households, including facilitating market connections, insulating market-facing households from environmental and price shocks, improving access to alternative 3 livelihoods like livestock, and expanding access to other commercial crops, like groundnut, soya, sunflower, and sweet potato.

Pinpointing concentrations of potential production and domestic demand can help identify priority areas for programs 4 that strive to build inclusive supply chains.

57 Source: Fraym Appendices and sources APPENDIX A || SOIL TYPE

Erosion is a key risk to soil quality in tobacco-farming areas.

The most common soil types in tobacco-farming areas are lixisols and luvisols, which have a 1 higher clay content and base saturation in the subsoil than in the topsoil.1

Control of erosion to preserve the topsoil, which 2 contains important organic matter, is important to prevent deterioration of soil quality.

Soil Type2 Luvisols (high-activity clay, high base status)

Lixisols (Low-activity clay, high base status )

Ferralsols (Red and yellow tropical soils)

Cambisols (Brown soils)

Note 1: Soil type groupings and descriptions are from the FAO World Reference Base for Soil Resources. Note 2: Map shows soil type groupings in Malawi, according to the FAO World Reference Base. Water bodies are shown in blue. Areas with a probability of tobacco farming less than 60% are shown in gray. Source: Fraym, SOTER (Soil and Terrain Database of the Republic of Malawi), World Reference Base for Soil Resources 2014, FAO, 59 http://www.fao.org/3/i3794en/I3794en.pdf. APPENDIX B || DIETARY DIVERSITY INDEX METHODOLOGY

The dietary diversity indicator incorporates the diversity, frequency, and nutrient density of different food groups consumed by households

Overview of Dietary Diversity Food Groups and Weights Indicator is developed using methodologies from 1 the World Food Program1 and USAID.2 Food Items (examples) Food Groups Weight

All food consumed in the past seven days is Cereals, grains 2 grouped into one of nine discrete food groups. Main staples 2 Roots, tubers, plantains

The number of days each food group was eaten in Beans, groundnuts, peas Pulses 3 3 the past seven days is calculated, with a maximum value of seven for each group. Cabbage, tanaposi leaves Vegetables 1

The value obtained by each food group is Mango, banana, avocado Fruit 1 4 multiplied by its weight. More nutrient dense Pork, fish, egg Animal protein 4 foods are given higher weights. Milk, yogurt, cheese Dairy 4 Sum the weighted food groups to get a raw food 5 consumption score. Sugar cane, honey, jam Sugar 0.5 Cooking oil, butter Oil 0.5 Households in the top third of the raw food 6 consumption score are considered to have diverse Tea, coffee, hot sauce Condiments 0 diets.

Note 1: This indicator is based off World Food Program’s (WFP) Food Consumption Score, developed in the manual Food Consumption Analysis (2008). Note 2: This indicator uses USAID’s Food and Nutrition Technical Assistance Manual’s targets for dietary diversity. Note 3: Food group weights based on nutrient density is based off of the WFP’s analysis. Note 4: Using the raw food consumption score of the top third of households as the cutoff point for dietary diversity is based on USAID recommendations. Source: Fraym, Fourth Integrated Household Survey APPENDIX C || DIETARY DIVERSITY DISTRIBUTIONS

16% 14% 14% 12% 12% 10% 10% 8% 8% 6% Households 6% 4% 4%

Percent 2%

2% Percent of Households 0% 0% 4 16 28 40 52 64 76 88 100 112 4 16 28 40 52 64 76 88 100 112 Tobacco households dietary diversity score Agricultural households dietary diversity score

9% 14% 8% 12% 7% 10% 6% 5% 8% 4% 6% of Households 3% 4% 2% Percent 1% Percent of Households 2% 0% 0% 4 16 28 40 52 64 76 88 100 112 4 16 28 40 52 64 76 88 100 112 Non-agricultural households dietary diversity score All households dietary diversity score

Note: Dietary diversity scores are in bins of four. The red line at a score of 47.5 represents the USAID target for a diverse diet in Malawi in 2017 based on LSMS data. Source: Fraym, Fourth Integrated Household Survey 61 APPENDIX D || TOBACCO HOUSEHOLD FOOD GROUP DISTRIBUTION

60

50

40

30

20 Unweighted food consumption Unweighted score

10

0 12 35 47.5 63 112 Weighted food consumption score Main staples Pulses Vegetables Fruits Animal protein Dairy Sugar Oil

Note: Distribution shows the average unweighted food score, a measure of the number of different foods eaten per week, at different levels of the weighted food consumption. Maximum unweighted food consumption score is 56. Red line marks where diverse diets for Malawi start, based on USAID recommendations. Source: Fraym, Fourth Integrated household Survey 62 APPENDIX E || INTERVIEW TIMING

Interview month by household1 Tobacco households Non-tobacco agriculture Percent of Percent of non-tobacco Month of Interview Interviewed households interviewed tobacco households agricultural household ( n = 571 ) ( n = 9,133 ) January 14 2.5% 581 6%

February 91 15.9% 1,260 14%

March 74 13.0% 1,382 15%

April 98 17.2% 1,660 18%

May 44 7.7% 746 8%

June 55 9.6% 837 9%

July 35 6.1% 443 5%

August 1 0.2% 56 1%

September 50 8.8% 856 9%

October 66 11.6% 708 8%

November 14 2.5% 139 2%

December 29 5.1% 465 5%

Note 1: Households were interviewed on a single day from April 15, 2016 until May 1, 2017, with each household only being interviewed once. Households within the same enumeration are were interviewed within a small date range, but there were many enumerators working at the same time across the country, and so any issues with enumeration timing should be homogenous across household types. 63 Source: Fraym, Fourth Integrated Household Survey APPENDIX F || ADDITIONAL SOURCES

Fraym's platform uses proprietary machine learning algorithms to weave together billions of data points. This innovative approach provides hyper-local insights into communities across Africa with an unprecedented level of accuracy.

For this report, Fraym incorporated additional information from the following sources: • Consultations with Viwemi Chavula, Center for Civil Society Strengthening, November 2018. • Malawi Fourth Integrated Household Survey, 2017. • Malawi Demographic and Health Survey, 2016. • Isric World Soil Information, Africa Soil Grids 2015, accessed October 2018. • Global Forest Watch, accessed October 2018. • USAID Food and Nutrition Technical Assistance Guide, 2006. • World Food Program Food Consumption Analysis, 2008. • Feed the Future Soyabean Innovation Lab

64 Source: Fraym APPENDIX G || DATA AND METHODOLOGY

Fraym Data Sources Fraym Methodology

The Fraym platform weaves together the latest satellite imagery Fraym data scientists closely examine representativeness, and geostatistical datasets with professionally enumerated sampling frames, questionnaire coverage, periodicity, and a range household surveys. This allows for the disaggregation and re- of other factors. Fraym obtains microdata, e.g. individual rows of aggregation of large datasets to cover any geographically bounded responses of survey data, in order to avoid any manipulation that area. could potentially occur during the analysis phase.

Indicators are drawn and harmonized from a wide variety of In Malawi, the surveys were implemented by the National household surveys and other data sources. These include the Statistical Office with financial, technical, and managerial support following sources: by large internationally respected organizations, including the USAID: Demographic and health surveys World Bank, USAID, and Millennium Challenge Corporation. These surveys are designed to be representative of both the de jure and United Nations: UN population division database de facto populations. World Bank: Enterprise surveys, living standards, global findex These surveys typically use a stratified, two-stage cluster design surveys, and respective country statistics that ensures representative samples for the national and National Statistical Offices: National censuses and surveys subnational levels. After data collection, post-hoc sampling covering population, businesses, health, housing, agriculture, weights are created to account for any oversampling and ensure and other areas representativeness particularly at hyperlocal levels. International Monetary Fund: World economic outlook databases and respective country statistics National Air and Space Administration: Remote sensing satellite data, such as vegetation, temperature, and precipitation USGS: Landscan, Google Earth, GeoData Institute, OSM

65 Source: Fraym Contact [email protected]