Jump to ContentJump to Main Navigation
Agriculture, Diversification, and Gender in Rural AfricaLongitudinal Perspectives from Six Countries$

Agnes Andersson Djurfeldt, Fred Mawunyo Dzanku, and Aida Cuthbert Isinika

Print publication date: 2018

Print ISBN-13: 9780198799283

Published to Oxford Scholarship Online: February 2018

DOI: 10.1093/oso/9780198799283.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use. date: 17 October 2019

African Smallholder Farmers on the Move: Farm and Non-Farm Trends for Six Sub-Saharan African Countries, 2002–15

African Smallholder Farmers on the Move: Farm and Non-Farm Trends for Six Sub-Saharan African Countries, 2002–15

Chapter:
(p.17) 2 African Smallholder Farmers on the Move: Farm and Non-Farm Trends for Six Sub-Saharan African Countries, 2002–15
Source:
Agriculture, Diversification, and Gender in Rural Africa
Author(s):

Magnus Jirström

Maria Archila Bustos

Sarah Alobo Loison

Publisher:
Oxford University Press
DOI:10.1093/oso/9780198799283.003.0002

Abstract and Keywords

This chapter provides a broad descriptive background of central aspects of smallholder agriculture in six countries in sub-Saharan Africa (SSA). It offers an up-to-date picture of the current trends of crop production, area productivity, levels of commercialization, and sources of cash incomes among 2,500 farming households. Structured around smallholder production, commercialization, and diversification in the period 2002–15, the chapter points on the one hand at persistent challenges such as low crop yields, low levels of output per farm, and a high degree of subsistence farming, and on the other hand at positive change over time in terms of growth in crop production and increasing levels of commercialization. It points at large variations not only between countries and time periods but also at the village levels, where gaps in crop productivity between farms remain large. Implicitly it points at the potential yet to be exploited in the SSA smallholder sector.

Keywords:   smallholder agriculture, sub-Saharan Africa, small farms, crop production, maize, rice, yield gaps, commercialization, income diversification, gender gaps

Introduction

Agricultural development in sub-Saharan Africa (SSA) remains high on the global development agenda. Having predominantly agriculture-based economies, the agricultural sector’s role in the development and transformation of the subcontinent’s economies deserves much attention in view of existing challenges regarding food insecurity and widespread poverty. Despite encouraging economic growth and transformation that has brought millions out of extreme poverty and into food security during the past fifteen years, high levels of poverty and food insecurity continue to plague Africa. With more than 40 per cent of its population, approximately 415 million, living in extreme poverty and approximately a fourth being undernourished, the SSA region faces enormous challenges (FAO 2014, ECA 2016, World Bank 2016).

Several transformational trends are currently affecting the countries of SSA and will contribute in shaping the development of their agricultural sectors in the coming decades. These trends include demographic change, rapid urbanization, and a shift in the labour force from farming to non-farm jobs. In the same way as the more recent years of successful development and change has varied between regions and countries in SSA, so, we may assume, will the impacts of the mentioned trends. Acknowledging vast regional and country differences, the broad trends mentioned can, however, be expected to bring change affecting the whole subcontinent.

The SSA population is estimated to grow from currently more than 950 million people to approximately 2.1 billion by 2050 (OECD/FAO 2016). While (p.18) the rural share of the population is declining and estimated to reach 50 per cent within the next ten years, the absolute number of rural people is expected to increase for another two decades and by 2050 amount to more than 900 million (Hazell 2013, United Nations 2014). Continued rapid population growth will affect agriculture in several ways. As shown by Losch et al. (2012), between 2010 and 2025 some 330 million new labour market entrants need to be absorbed in the SSA economies. The rural share of these—195 million young people—will be looking for rural employment and if unsuccessful will leave the rural areas. In many rural settings, land fragmentation and declining farm sizes are growing phenomena and signs of growing land pressure (Jayne et al. 2014), and questions about current and future viability of minuscular farms are being raised (Hengsdijk et al. 2014).

Urbanization represents another transforming force affecting smallholder agriculture through several mechanisms. Urban consumers purchase much of the food they require, thereby potentially increasing the demand for food produced by smallholders in the rural areas. Although food imports continue to provide a very high share of overall food supply, the increasing global demand for food, fibre, and fuel from agriculture have made food imports a more expensive and risky source of supply for the growing African population. For African smallholders, this development could offer opportunities. Furthermore, urban—and increasingly also rural—people in SSA are changing diets and consuming more high-value foods such as vegetables, fruits, milk, meat, etc. These ongoing dietary shifts may provide opportunities for smallholders supplying such produce.

Urbanization in SSA is mostly characterized by the rapid growth of small and medium-sized cities. This offers opportunities for rural households to diversify their income sources and generate some of their income directly from urban and peri-urban activities without abandoning the rural areas. As will be shown later in this chapter, having access to non-farm income sources can improve smallholder household income substantially. Exiting agriculture in SSA may, as in other parts of the world, be a gradual process and there is currently strong interest in questions related to the improvement of opportunities for smallholders’ income diversification into non-farm income sources (Haggblade et al. 2010, Frelat et al. 2016). Smallholders’ livelihood diversification forms part of the overall structural transformation taking place in SSA. Based on a sample of more than twenty-four countries in SSA, McMillan and Harttgen (2014) estimate the share of the labour force in the agricultural sector to have declined by approximately 10 percentage points between 2000 and 2010.

A large share of the labour force in SSA agriculture is female. In a recent assessment of the contribution of women to labour in crop production across six SSA countries, Palacios-López, et al. (2016), estimate the average share at (p.19) 40 per cent. Female-managed farms (FMFs) play a significant role in SSA and make up close to 30 per cent of the farms in the Afrint III sample. The final section of this chapter presents data pointing at several gender gaps contributing to the challenges facing smallholder agriculture in SSA.

This chapter uses data from the three Afrint surveys described in the methodology in Chapter 1. While panel data from the surveys are used in several chapters, this chapter reports on the data from the cross-sections. For production, the focus is set on two major SSA cereal staples—maize and rice—the production of which are of central importance for the great majority of households in the sample. As will be shown, the share of smallholder households completely depending on farming for their livelihoods remains high. At the same time, for approximately half (55 per cent) of the farm households studied, non-farm income sources provide more than half of total household cash income. While in a cross-sectional analysis of smallholder agriculture for the Afrint I and Afrint II periods, Jirström et al. (2011) could only identify flickering signs of dynamism, the presentation in the following sections will point at several important signs of dynamism, although they are clearly heterogeneous in character.

Increasing Farm Sizes: Relatively Stable Crop Portfolios

For Afrint III, the mean and median farm size, here defined as the area under cultivation, had increased substantially compared to the Afrint II survey round (Table 2.1). The Afrint III mean farm size was 2.3 ha which, compared to the Afrint II figure of 1.7 ha, represents a 36 per cent increase. The increase is significant1 in four countries—Ghana, Tanzania, Zambia, and Mozambique—ranging between 24 per cent to 70 per cent. These countries can be categorized as relatively land-abundant in an African context (Jayne et al. 2014) and we assume that farmers in these countries have brought fallow land into production. In contrast, Kenya and Malawi, where the sampled households report small changes in farm size between the two survey rounds, belong to the growing number of land-constrained African countries. As shown in Table 2.1, the Kenyan and Malawian farm households cultivate substantially smaller farms than farmers in the other four countries.

Table 2.1. Land under cultivation, ha (mean and median)

Change over time (%)

Afrint I

Afrint II

Afrint III

Afrint I to II

Sig.

Afrint II to III

Sig.

Afrint I to III

Sig.

Ghana

Mean farm size (ha)a

2.41

1.99

2.84

−17

***

43

***

18

***

Median farm size (ha)a

2.00

1.60

2.09

No. of cases

412

565

538

Kenya

Mean farm size (ha)a

0.97

0.98

1.02

1

3

5

Median farm size (ha)a

0.80

0.75

0.82

No. of cases

300

298

300

Malawi

Mean farm size (ha)a

1.24

1.18

1.21

−5

3

−2

Median farm size (ha)a

1.01

1.01

1.00

No. of cases

387

393

388

Tanzania

Mean farm size (ha)a

1.96

1.72

2.14

−13

**

24

***

9

Median farm size (ha)a

1.60

1.40

1.62

No. of cases

399

392

390

Zambia

Mean farm size (ha)a

3.01

2.64

3.55

−12

**

35

***

18

***

Median farm size (ha)a

2.30

2.00

2.50

No. of cases

395

397

473

Mozambique

Mean farm size (ha)a

2.02

1.33

2.25

−34

***

70

***

12

*

Median farm size (ha)a

1.50

1.00

1.75

No. of cases

381

400

424

Group total

Mean farm size (ha)a

1.98

1.69

2.30

−15

***

36

***

16

***

Median farm size (ha)a

1.50

1.21

1.60

No. of cases

2,274

2,445

2,513

Notes: (a.) Cases above 25 ha removed.

Table 2.2 shows data for an analysis of the present household landholding size per capita, as well as a closer look at land distribution, by dividing the (p.20) (p.21) (p.22) (p.23) sample into quartiles ranked by per capita farm size and by village. As can be seen, per capita access to land is very limited—in Afrint III it was 0.16 ha per capita or less for the bottom quartile (Table 2.2). For Afrint I, the corresponding figure was 0.15 ha. Under rain-fed conditions and with limited use of other land-augmenting technologies such as adequate application of inorganic and organic fertilizers, the small per capita farm size experienced by a quarter of the sample is worrying. This raises the question of whether such a limited access to farmland is sustainable, not least in light of the growth of the rural African population which is projected to continue for another two to three decades (Masters et al. 2013).

Table 2.2. Landholding size per capita

Change over time (%)

Afrint I

Afrint II

Afrint III

Afrint I–II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Ghana

Mean farm size (ha)a per capita, total sample

0.37

0.30

0.48

−18

**

57

***

29

***

Means for household quartiles ranked by farm size (ha)a per capita by village

Q1

0.09

0.09

0.12

Q2

0.21

0.18

0.26

Q3

0.33

0.30

0.42

Q4

0.82

0.63

1.06

Kenya

Mean farm size (ha)a per capita, total sample

0.20

0.14

0.21

−30

***

50

***

6

Means for household quartiles ranked by farm size (ha)a per capita by village

Q1

0.05

0.03

0.07

Q2

0.10

0.07

0.12

Q3

0.17

0.13

0.19

Q4

0.43

0.31

0.43

Malawi

Mean farm size (ha)a per capita, total sample

0.31

0.29

0.26

−5

−11

*

−16

***

Means for household quartiles ranked by farm size (ha)a per capita by village

Q1

0.12

0.12

0.09

Q2

0.21

0.21

0.17

Q3

0.31

0.30

0.25

Q4

0.57

0.53

0.52

Tanzania

Mean farm size (ha)a per capita, total sample

0.39

0.36

0.50

−8

40

***

28

***

Means for household quartiles ranked by farm size (ha)a per capita by village

Q1

0.14

0.13

0.16

Q2

0.25

0.22

0.29

Q3

0.39

0.34

0.46

Q4

0.75

0.72

1.06

Zambia

Mean farm size (ha)a per capita, total sample

0.43

0.38

0.50

−10

29

***

16

*

Means for household quartiles ranked by farm size (ha)a per capita by village

Q1

0.15

0.12

0.15

Q2

0.27

0.22

0.27

Q3

0.43

0.38

0.43

Q4

0.81

0.77

1.08

Mozambique

Mean farm size (ha)a per capita, total sample

0.38

0.29

0.43

−22

***

48

***

15

**

Means for household quartiles ranked by farm size (ha)a per capita by village

Q1

0.12

0.09

0.12

Q2

0.24

0.18

0.23

Q3

0.36

0.27

0.39

Q4

0.75

0.58

0.95

Group total

Mean farm size (ha)a per capita, total sample

0.35

0.30

0.41

−14

***

36

***

17

***

Means for household quartiles ranked by farm size (ha)a per capita by village

Q1

0.12

0.10

0.12

Q2

0.22

0.19

0.23

Q3

0.34

0.29

0.37

Q4

0.70

0.60

0.88

No. of casesb

2,231

2,384

2,478

Notes: (a.) Cases above 25 ha removed.

(b.) The number of cases in each quartile group varies as all households with the same per capita farm size are automatically placed in the same group in this SPSS routine.

Differences between the four quartiles are large, with a range in Afrint III of six to nine times more land per capita for the highest quartile compared to the lowest one. In relation to the situation in Afrint II, the mean per capita farm size in Afrint III was 36 per cent larger, corresponding exactly to the increase in the mean farm size over the same period. Out of the six countries only the Malawi sample shows falling per capita farm sizes (−11 per cent). To our surprise, the Kenya sample does not confirm the pattern of declining per capita farm size in Kenya (Jayne et al. 2006b). Instead we find that the mean per capita access to land of the Kenyan households has risen by 50 per cent during the period. The explanation perhaps lies in the declining mean household size in the Kenyan sample from 7.0 in Afrint II to 6.1 persons per household in Afrint III.

Turning to the emergence of growing land inequalities as reported in the literature (Jayne et al. 2006a), the increasing difference in per capita farm size between the bottom and the top 25 per cent of the three Afrint samples also points in the direction of gradually growing land inequality, although the change seems moderate in the Afrint villages. For Afrint I, the per capita access to land for the bottom 25 per cent was 0.12 ha per capita and this was also the per capita size in 2013. However, the top 25 per cent expanded from a mean of 0.70 ha to 0.88 ha per capita. As a consequence, the gap between the two extreme groups has increased—in Afrint I the top quartile accessed 6.0 times as much land as the bottom quartile while in Afrint III they accessed 7.3 times as much land (see Table 2.2).

Shifting attention to land use and crop portfolios, the increase in total cultivated area between Afrint II and Afrint III corresponds to the general increase in the cultivated area of the five different categories of crops studied (Table 2.3). For Afrint III, 92 per cent of all households grew maize and the crop remained the most important crop in terms of land allocation, with an average of a little over 1 ha. Both the share of households growing maize and the maize area have increased since Afrint II.

Table 2.3. Land under cultivation (total and per crop) and share of households cultivating by type of crop

Change over time (%)a

Afrint I

Afrint II

Afrint III

Afrint I– II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Maize

Mean crop area (ha)b

0.80

0.81

1.08

1

32

***

34

***

Median crop area (ha)b

0.50

0.60

0.80

Households cultivating (%)

84

86

92

2

**

6

***

8

***

No. of cases

1,951

2,136

2,348

Sorghum

Mean crop area (ha)b

0.73

0.67

0.81

−8

*

21

***

11

Median crop area (ha)b

0.50

0.50

0.50

Households cultivating (%)

21

20

16

−1

−4

***

−5

***

No. of cases

487

495

409

Rice

Mean crop area (ha)b

0.70

0.65

0.83

−8

29

***

18

***

Median crop area (ha)b

0.50

0.40

0.60

Households cultivating (%)

22

23

24

1

1

2

*

No. of cases

508

570

614

Other food crops

Mean crop area (ha)b

0.56

0.51

0.79

−9

55

**

42

***

Median crop area (ha)b

0.40

0.30

0.40

Households cultivating (%)

84

79

90

−5

***

11

***

6

***

No. of cases

1,943

1,954

2,291

Non-food crops

Mean crop area (ha)b

0.79

0.77

0.88

−2

14

12

Median crop area (ha)b

0.40

0.40

0.48

Households cultivating (%)

34

23

32

−11

***

9

***

−2

No. of cases

795

579

817

Total

Mean crop area (ha)b

1.98

1.69

2.30

−15

***

36

***

16

***

Median crop area (ha)b

1.50

1.21

1.60

Households cultivating (%)

100

100

100

n.a.

n.a.

n.a.

No. of cases

2,325

2,473

2,544

Note: (a.) Change in percentage of households cultivating is absolute not percentage change.

(b.) Cases above 25 ha removed.

Approximately a quarter of all households grow rice, a share which remained stable over the study period. Also in the case of rice, there has (p.24) (p.25) been a significant increase, 29 per cent, in mean hectarage. Sorghum growers also allocate more land to the crop, but the share of farmers growing the crop has dropped. Between Afrint II and Afrint III, the share dropped from 20 per cent to 16 per cent of the sample.

The single biggest increase in cropped area pertains to the category ‘other food crops’, which includes all food crops but maize, sorghum, and rice.2 Land allocated to this category grew by 55 per cent between Afrint II and Afrint III, and also the proportion of households growing such crops grew from 79 per cent to 90 per cent. The category includes high-value crops including vegetables and fruits, and the increase in the cultivation of such crops can be seen as a strategy among farm households to increase the value of output per hectare. The non-food crop category, including cash crops such as coffee, tobacco, tea, and sugarcane, engages approximately a third of the sample, up from a little less than a quarter in Afrint II.

The aggregates for the six countries presented here will be explored in more detail at the country level in the subsequent sections on production and productivity for the two staples, maize and rice. We focus on these staple crops because of their strategic importance for African agriculture. The demand for food staples is growing rapidly and there is a potential for farmers in Africa to capture a large share of this growth (Hazell and Poulton 2007). Apart from contributing to growth and poverty reduction in itself, an increase in the productivity of staples can ‘release land, water and labour for the production of other cash and export crops’ (Hazell and Poulton 2007: 3). For African smallholders who are often net buyers of staple foods, increasing productivity in staple crop production can reduce poverty and increase food security.

Production and Area Productivity

Maize

Maize is the most widely grown staple food crop in SSA. It accounts for 30 per cent of cereal production area and about 40 per cent of cereal production (Hazell and Poulton 2007, Cairns et al. 2013). Since the early 1960s, growth in maize production has been driven mainly by area expansion, while yield growth explains only a third of the average annual growth rate of 3 per cent for the period 1961–2008 (Smale et al. 2011). However, maize yields in SSA (p.26) (p.27) (p.28) (p.29) (p.30) remain far behind the global average, and at around 1.8 t/ha only reach a third of the global mean maize yield of around 5.5 t/ha (CIMMYT 2016, USDA 2016). The mentioned average maize yield of 1.8 t/ha stems from Food and Agricultural Organization (FAO) data. Serious criticism of the quality of the national agricultural statistics in SSA on which the FAO data build, has been levelled (Carletto et al. 2014). The farm household survey data on which our figures build also point at significant discrepancies in comparison with the FAO data. When making comparison with national-level data this should be kept in mind.

The three-season average production per farm in our sample of six countries increased substantially from 1.08 t per farm in Afrint II to 1.43 t in Afrint III (Table 2.4). The overall growth in production of 32 per cent is largely the result of an increase in the maize area which grew by 29 per cent during the period. The level of median production volume, 0.60 t in Afrint II and 0.73 t in Afrint III, indicates that, in spite of substantial growth, there is not much, if anything, to market for many households considering that maize is an important food for most households. Except in Zambia where the maize area is larger, the median total production is in the range of 0.33 t to 0.84 t.

Table 2.4. Maize production (t/farm) and cultivated area (ha)

Change over time (%)a

Afrint I

Afrint II

Afrint III

Afrint I–II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Ghana

Proportion of farmers growing maize (%)b

43

53

71

10

***

18

***

28

***

Mean area under maize (ha), 3-season averagec

1.08

0.63

1.11

−42

***

76

***

3

Mean maize production (t/farm), 3-season averaged

0.89

0.61

0.85

−32

***

40

***

−4

Median maize production (t/farm), 3-season averaged

0.63

0.37

0.50

No. of cases

179

301

382

Kenya

Proportion of farmers growing maize (%)b

74

59

85

−15

***

26

***

11

***

Mean area under maize (ha), 3-season averagec

0.52

0.61

0.51

16

−15

**

−2

Mean maize production (t/farm), 3-season averaged

1.19

1.08

0.68

−9

−37

***

−43

***

Median maize production (t/farm), 3-season averaged

0.42

0.45

0.33

No. of cases

222

176

255

Malawi

Proportion of farmers growing maize (%)b

94

94

93

0

−1

−1

Mean area under maize (ha), 3-season averagec

0.30

0.74

0.84

144

***

13

*

176

***

Mean maize production (t/farm), 3-season averaged

0.70

0.83

0.89

19

***

7

27

***

Median maize production (t/farm), 3-season averaged

0.53

0.53

0.62

No. of cases

376

376

375

Tanzania

Proportion of farmers growing maize (%)b

82

82

82

0

0

0

Mean area under maize (ha), 3-season averagec

1.06

0.90

0.88

−15

**

−3

−17

***

Mean maize production (t/farm), 3-season averaged

0.98

1.18

1.29

20

**

9

31

***

Median maize production (t/farm), 3-season averaged

0.63

0.70

0.84

No. of cases

332

328

324

Zambia

Proportion of farmers growing maize (%)b

77

83

88

6

**

4

*

11

***

Mean area under maize (ha), 3-season averagec

1.36

1.30

1.71

−4

32

***

26

***

Mean maize production (t/farm), 3-season averaged

1.58

2.13

3.37

34

***

58

***

113

***

Median maize production (t/farm), 3-season averaged

0.90

1.17

1.93

No. of cases

313

336

423

Mozambique

Proportion of farmers growing maize (%)b

79

63

78

−15

***

15

***

0

Mean area under maize (ha), 3-season averagec

0.95

0.71

1.17

−25

***

64

***

23

***

Mean maize production (t/farm), 3-season averaged

0.48

0.50

0.92

5

83

***

93

***

Median maize production (t/farm), 3-season averaged

0.35

0.35

0.50

No. of cases

313

255

334

Total

Proportion of farmers growing maize (%)b

75

72

82

−3

**

11

***

8

***

Mean area under maize (ha), 3-season averagec

0.86

0.84

1.09

−3

29

***

26

***

Mean maize production (t/farm), 3-season averaged

0.95

1.08

1.43

13

**

32

***

50

***

Median maize production (t/farm), 3-season averaged

0.55

0.60

0.73

No. of cases

1,735

1,772

2,093

Notes:

(a.) Change in proportion of households growing maize is absolute not percentage change.

(b.) Based on a sub-sample including maize growers who cultivated an average of at least 0.1 ha. The sample also excludes cases with a sixfold or higher yield increase between consecutive Afrint rounds.

(c.) Average of area under maize during the last growing season at the time of data collection and the two previous growing seasons before that one. Extreme values removed at the dataset level for each year.

(d.) Average of maize production during the last growing season at the time of data collection and the two previous growing seasons before that one. Extreme values removed based on average area criteria.

The overall mean yield of maize for the entire sample of maize growers of 1.25 t/ha for Afrint III does not indicate any significant change in yields since Afrint I (1.30 t/ha) or Afrint II (1.21 t/ha) (Table 2.5). Variation in production and productivity among the six countries is substantial. In Ghana, Mozambique, and Zambia the mean maize area, mean production, and share of households growing maize increased substantially between Afrint II and Afrint III (Table 2.4).

Table 2.5. Maize yields

Change over time (%)e

Afrint I

Afrint II

Afrint III

Afrint I–II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Ghana

Mean maize yield (t/ha),a 3-season averageb

1.09

1.01

0.84

−7

−17

***

−23

***

Median maize yield (t/ha),a 3-season averageb

0.75

0.95

0.67

No. of cases

175

300

382

20% best-performing farmers’ mean yield (t/ha) maizec

2.54

1.84

1.71

−28

**

−7

−33

***

5% best-performing farmers’ mean yield (t/ha) maizec

4.14

2.72

2.42

−34

*

−11

−41

**

Average yield gap,d compared to 20% best-performing farmers

57

42

44

−15

***

2

−13

***

Average yield gap,d compared to 5% best-performing farmers

74

59

56

−16

***

−2

−18

***

Kenya

Mean maize yield (t/ha),a 3-season averageb

1.70

1.50

1.25

−12

−17

**

−26

***

Median maize yield (t/ha),a 3-season averageb

1.05

1.20

0.90

No. of cases

220

176

255

20% best-performing farmers’ mean yield (t/ha) maizec

3.46

2.79

2.59

−19

−7

−25

**

5% best-performing farmers’ mean yield (t/ha) maizec

4.51

3.62

3.43

−20

−5

−24

Average yield gap,d compared to 20% best-performing farmers

52

47

52

−5

5

0

Average yield gap,d compared to 5% best-performing farmers

64

58

61

−5

*

2

−3

Malawi

Mean maize yield (t/ha),a 3-season averageb

2.22

1.17

1.16

−47

***

−1

−48

***

Median maize yield (t/ha),a 3-season averageb

1.96

0.99

0.92

No. of cases

374

376

372

20% best-performing farmers’ mean yield (t/ha) maizec

4.11

2.19

2.33

−47

***

6

−43

***

5% best-performing farmers’ mean yield (t/ha) maizec

5.11

2.88

3.27

−44

***

13

−36

***

Average yield gap,d compared to 20% best-performing farmers

45

44

51

−1

7

**

6

**

Average yield gap,d compared to 5% best-performing farmers

56

57

62

1

5

**

6

***

Tanzania

Mean maize yield (t/ha),a 3-season averageb

0.97

1.33

1.46

37

***

10

*

51

***

Median maize yield (t/ha),a 3-season averageb

0.79

1.10

1.15

No. of cases

332

322

323

20% best-performing farmers’ mean yield (t/ha) maizec

1.85

2.59

2.83

40

***

9

53

***

5% best-performing farmers’ mean yield (t/ha) maizec

2.57

3.30

3.58

28

**

8

39

***

Average yield gap,d compared to 20% best-performing farmers

47

49

51

2

2

4

Average yield gap,d compared to 5% best-performing farmers

61

60

61

−1

1

0

Zambia

Mean maize yield (t/ha),a 3-season averageb

1.14

1.47

1.82

29

***

24

***

59

***

Median maize yield (t/ha),a 3-season averageb

1.00

1.31

1.61

No. of cases

310

336

423

20% best-performing farmers’ mean yield (t/ha) maizec

2.09

2.72

3.30

30

***

21

***

58

***

5% best-performing farmers’ mean yield (t/ha) maizec

2.72

3.39

4.28

24

***

26

***

57

***

Average yield gap,d compared to 20% best-performing farmers

45

47

45

2

−2

0

Average yield gap,d compared to 5% best-performing farmers

57

57

58

0

1

0

Mozambique

Mean maize yield (t/ha),a 3-season averageb

0.53

0.83

0.89

58

***

6

68

***

Median maize yield (t/ha),a 3-season averageb

0.46

0.67

0.62

No. of cases

313

254

331

20% best-performing farmers’ mean yield (t/ha) maizec

1.04

1.69

2.09

63

***

24

**

102

***

5% best-performing farmers’ mean yield (t/ha) maizec

1.45

2.39

3.31

65

***

38

**

128

***

Average yield gap,d compared to 20% best-performing farmers

49

51

56

2

5

8

**

Average yield gap,d compared to 5% best-performing farmers

63

64

69

1

5

6

**

Total

Mean maize yield (t/ha),a 3-season averageb

1.30

1.21

1.25

−6

**

3

−4

Median maize yield (t/ha),a 3-season averageb

0.94

1.01

0.94

No. of cases

1,724

1,764

2,086

20% best-performing farmers’ mean yield (t/ha) maizec

2.49

2.29

2.49

−8

*

9

**

0

5% best-performing farmers’ mean yield (t/ha) maizec

3.36

3.02

3.39

−10

12

**

1

Average yield gap,d compared to 20% best-performing farmers

48

47

49

−2

3

**

1

Average yield gap,d compared to 5% best-performing farmers

61

59

61

−3

**

2

**

−1

Notes:

(a.) Based on a sub-sample including maize growers who cultivated an average of at least 0.1 ha. The sample also excludes cases with a sixfold or higher yield increase between consecutive Afrint rounds.

(b.) Average of maize yield during the last growing season at the time of data collection and the two previous growing seasons before that one.

(c.) Based on village aggregates.

(d.) The yield gap is calculated as the difference between the household’s 3-season average yield and the average of the best-performing farmers in the village.

(e.) Change in average yield gap is absolute not percentage change.

In Zambia, the mean maize yield increased by 24 per cent between Afrint II and III, a positive change which is close to the national increase of 22 per cent for the same period (FAO 2016). For Mozambique, change during the period was not statistically significant, while for Ghana area expansion outpaced production increase to the extent that the three-year average yield of 0.86 t/ha for the Afrint III period represents a decrease of 17 per cent compared to the Afrint II level. Also, national figures for Ghana show that the average maize yields are among the lowest in the world and remain much lower than the average for SSA (Ragasa et al. 2014). As shown by Dzanku and Sarpong (2014), regional differences in the Ghana sample are substantial, with farmers in the four villages situated in the Eastern Region achieving yields approximately twice as high as those in the Upper East Region in the northern and more arid parts of the country.

In Kenya, the proportion of households planting maize in the Afrint III sample had increased since Afrint II. However, the mean maize area decreased by 15 per cent and the production fell even more (−37 per cent) and, thus, the mean maize yield declined by 17 per cent from 1.5 t/ha to 1.25 t/ha in our (p.31) study regions between Afrint II and III. At a national level, yields have remained steady, at around 1.5 t/ha over the period of our study, setting Kenya, together with Tanzania, apart from neighbouring Eastern African countries such as Ethiopia, Uganda, and Malawi where maize yields have risen since 2003 (Aylward et al. 2015).

In Tanzania and Malawi, change between Afrint II and Afrint III has been limited and is not statistically significant. For Tanzania, the mean yield of 1.46 t/ha in Afrint III is close to the national level of 1.38 t/ha for 2012–14 and, as in our sample, the change in mean yields since 2006–8 has been marginal (FAO 2016, FAOSTAT). In the case of Malawi, the sampled households reported maize yields only about half the level of the national three-year mean yields of approximately 2.2 t/ha for the Afrint III period. A possible explanation for this difference may relate to the sampling criteria for the Malawi sample, where maize was the major crop only in some of the villages.

As previously indicated, average maize yields in SSA are low by international comparison. This is true also in comparison to most other tropical regions where maize is grown under rain-fed conditions. But yield variability also sets SSA apart. According to Byerlee and Heisey (1997), yield variability is almost always higher in SSA countries than in other developing countries with similar mean yields. While climatic factors such as major regional droughts are important explanations, price variability and government policies as well as interventions can be assumed to play a role as well. Yield variability does, however, not only occur between countries and regions. As shown in Table 2.5, differences in maize yields at the village level are significant. The six-country three-season average for Afrint III shows that the yield gap at the village level is substantial. The best-performing 20 per cent of farmers in each village achieve, on average, the double maize yield compared to each individual farmer in the village. When comparing each farmer’s yield with the 5 per cent best-performing farmers’ yield, the gap widens and reaches around 60 per cent. In a subsequent section, we will return to the yield gap discussion in more detail. First, we turn to the second most important cereal among the Afrint farm households, namely rice.

Rice

The importance of rice is growing rapidly in SSA, and demand is growing more rapidly than in any other continent (Nasrin et al. 2015). Although production increased at an annual rate of 3.3 per cent over the period 1970–2009, the gap between consumption and production widened as consumption grew by an annual rate of 4 per cent and thus outpaced production. In 2009, 37 per cent (p.32) (p.33) (p.34) of consumption was imported and SSA has become a major importer of rice with almost a third, 11.8 million tons, of rice traded globally (Nasrin et al. 2015). Rice is now the second largest source of food in SSA and since the turn of the century has gradually come close to maize as a provider of food calories in the subcontinent. The growing popularity and consumption of rice can be traced to a combination of population growth, urbanization and urban growth, and changing diets.

Approximately a fifth of the Afrint sample cultivate rice, the share of which has been stable since Afrint I (Table 2.6). Out of the six countries, two—Kenya and Zambia—lacked households growing rice during the Afrint III period. Among the remaining four, Ghana and Tanzania together accounted for more than 70 per cent of the 527 rice-farming households during the Afrint III period. In the two countries, there has been a significant increase in the mean cultivated rice area, mean farm production, and mean yields over the three-season periods of Afrint II and Afrint III (Tables 2.6 and 2.7). The recorded growth tallies well with the national-level data for the two periods (FAO 2016, FAOSTAT). Data on milled rice consumption per capita for Ghana and Tanzania show a very rapid increase between 1982 and 2012 (Nasrin et al. 2015), and it can be assumed that the strong growth in demand for rice contributes to the substantial increases in production and productivity among the surveyed farms in Ghana and Tanzania.

Table 2.6. Rice production (t/farm) and cultivated area (ha)

Change over time (%)a

Afrint I

Afrint II

Afrint III

Afrint I–II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Ghana

Proportion of farmers growing rice (%)b

40

29

33

−11

***

4

−7

**

Mean area under rice (ha), 3-season averagec

0.66

0.59

0.87

−10

46

***

32

**

Mean rice production (t/farm), 3-season averaged

0.51

0.30

0.88

−41

***

194

***

72

***

Median rice production (t/farm), 3-season averaged

0.35

0.17

0.35

No. of cases

165

163

177

Malawi

Proportion of farmers growing rice (%)b

31

25

26

−6

**

1

−5

*

Mean area under rice (ha), 3-season averagec

0.44

0.48

0.54

10

13

24

**

Mean rice production (t/farm), 3-season averaged

0.76

1.07

0.91

40

***

−15

19

Median rice production (t/farm), 3-season averaged

0.66

0.97

0.67

No. of cases

93

74

77

Tanzania

Proportion of farmers growing rice (%)b

44

43

42

−1

−1

−2

Mean area under rice (ha), 3-season averagec

0.99

0.97

1.13

−2

16

**

13

*

Mean rice production (t/farm), 3-season averaged

1.51

1.89

2.65

25

**

41

***

75

***

Median rice production (t/farm), 3-season averaged

1.17

1.37

2.09

No. of cases

176

171

170

Zambia

Proportion of farmers growing rice (%)b

1

0

0

−1

0

−1

Mean area under rice (ha), 3-season averagec

0.32

n.a.

n.a.

n.a.

n.a.

n.a.

Mean rice production (t/farm), 3-season averaged

0.23

n.a.

n.a.

n.a.

n.a.

n.a.

Median rice production (t/farm), 3-season averaged

0.23

n.a.

n.a.

No. of cases

4

0

0

Mozambique

Proportion of farmers growing rice (%)b

11

15

24

4

9

***

13

***

Mean area under rice (ha), 3-season averagec

0.41

0.29

0.59

−29

***

103

***

45

***

Mean rice production (t/farm), 3-season averaged

0.14

0.23

0.22

63

**

−5

55

**

Median rice production (t/farm), 3-season averaged

0.10

0.15

0.16

No. of cases

44

60

103

Group total

Proportion of farmers growing rice (%)b

21

19

21

−2

2

0

Mean area under rice (ha), 3-season averagec

0.71

0.67

0.85

−5

26

***

19

***

Mean rice production (t/farm), 3-season averaged

0.89

0.99

1.33

11

34

***

49

***

Median rice production (t/farm), 3-season averaged

0.56

0.50

0.53

No. of cases

482

468

527

Notes: (a.) Change in proportion of households growing rice is absolute not percentage change.

(b.) Based on a sub-sample including rice growers who cultivated an average of at least 0.1 ha. The sample also excludes cases with a sixfold or higher yield increase between consecutive Afrint rounds.

(c.) Average of area under rice during the last growing season at the time of data collection and the two previous growing seasons before that one, extreme values removed at the dataset level for each year.

(d.) Average of rice production during the last growing season at the time of data collection and the two previous growing seasons before that one, extreme values removed based on average area criteria.

Table 2.7. Rice yields (ton/ha)

Change over time (%)e

Afrint I

Afrint II

Afrint III

Afrint I–II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Ghana

Mean rice yield (t/ha),a 3-season averageb

0.98

0.48

0.88

−51

***

81

***

−11

Median rice yield (t/ha),a 3-season averageb

0.76

0.35

0.62

No. of cases

164

163

177

20% best-performing farmers’ mean yield (t/ha) ricec

2.03

0.90

1.61

−56

***

79

***

−20

***

5% best-performing farmers’ mean yield (t/ha) ricec

2.64

1.15

1.97

−56

**

71

**

−26

Average yield gap,d compared to 20% best-performing farmers

50

46

44

−4

−2

−6

Average yield gap,d compared to 5% best-performing farmers

61

60

57

−1

−3

−4

Malawi

Mean rice yield (t/ha),a 3-season averageb

1.79

2.25

1.81

26

**

−19

**

1

Median rice yield (t/ha),a 3-season averageb

1.57

2.26

1.56

No. of cases

93

74

77

20% best-performing farmers’ mean yield (t/ha) ricec

3.28

3.53

3.62

8

***

2

10

***

5% best-performing farmers’ mean yield (t/ha) ricec

4.19

3.63

4.86

−13

34

16

Average yield gap,d compared to 20% best-performing farmers

48

49

55

1

6

7

Average yield gap,d compared to 5% best-performing farmers

63

59

65

−3

6

3

Tanzania

Mean rice yield (t/ha),a 3-season averageb

1.52

1.95

2.24

28

***

15

**

47

***

Median rice yield (t/ha),a 3-season averageb

1.45

1.61

2.09

No. of cases

176

171

170

20% best-performing farmers’ mean yield (t/ha) ricec

2.58

3.31

3.55

28

***

7

37

***

5% best-performing farmers’ mean yield (t/ha) ricec

3.08

4.00

4.61

30

15

50

***

Average yield gap,d compared to 20% best-performing farmers

42

40

37

−3

−3

−5

Average yield gap,d compared to 5% best-performing farmers

54

52

52

−2

0

−2

Zambia

Mean rice yield (t/ha),a 3-season averageb

0.47

n.a.

n.a.

n.a.

n.a.

n.a.

Median rice yield (t/ha),a 3-season averageb

0.40

n.a.

n.a.

No. of cases

4

0

0

20% best-performing farmers’ mean yield (t/ha) ricec

0.77

n.a.

n.a.

n.a.

n.a.

n.a.

5% best-performing farmers’ mean yield (t/ha) ricec

0.77

n.a.

n.a.

n.a.

n.a.

n.a.

Average yield gap,d compared to 20% best-performing farmers

34

n.a.

n.a.

n.a.

n.a.

n.a.

Average yield gap,d compared to 5% best-performing farmers

34

n.a.

n.a.

n.a.

n.a.

n.a.

Mozambique

Mean rice yield (t/ha),a 3-season averageb

0.39

0.90

0.52

134

***

−42

***

36

Median rice yield (t/ha),a 3-season averageb

0.25

0.63

0.33

No. of cases

43

60

103

20% best-performing farmers’ mean yield (t/ha) ricec

0.76

1.90

1.20

150

***

−37

*

58

***

5% best-performing farmers’ mean yield (t/ha) ricec

0.72

2.61

1.59

262

***

−39

121

**

Average yield gap,d compared to 20% best-performing farmers

49

49

56

0

6

7

Average yield gap,d compared to 5% best-performing farmers

57

60

64

3

4

7

Total

Mean rice yield (t/ha),a 3-season averageb

1.28

1.35

1.38

5

2

8

Median rice yield (t/ha),a 3-season averageb

1.06

0.97

1.05

No. of cases

480

466

525

20% best-performing farmers’ mean yield (t/ha) ricec

2.30

2.37

2.42

3

***

2

5

***

5% best-performing farmers’ mean yield (t/ha) ricec

2.60

2.78

2.87

7

3

10

Average yield gap,d compared to 20% best-performing farmers

47

45

46

−2

1

−1

Average yield gap,d compared to 5% best-performing farmers

58

57

58

−1

1

0

Notes: (a.) Based on a sub-sample including rice growers who cultivated an average of at least 0.1 ha. The sample also excludes cases with a sixfold or higher yield increase between consecutive Afrint rounds.

(b.) Average of rice yield during the last growing season at the time of data collection and the two previous growing seasons before that one.

(c.) Based on village aggregates.

(d.) The yield gap is calculated as the difference between the household’s 3-season average yield and the average of the best-performing farmers in the village.

(e.) Change in average yield gap is absolute not percentage change.

In the Mozambique sample, the share of households growing rice increased from 15 per cent to 24 per cent between Afrint II and Afrint III. The area under rice approximately doubled from 0.29 ha to 0.59 ha per farm but production remained at around 0.22 t/farm. Consequently, the mean three-season rice yield fell from 0.90 t/ha in the Afrint II period to 0.52 t/ha in the Afrint III period. FAO data indicate a similar situation at the national level, with a dramatic increase in the national rice hectarage accompanied by a sharp fall in the average national rice yield during the period (FAO 2016, FAOSTAT). In Malawi, finally, the fifth of the surveyed households cultivating rice saw their yields declining by, on average, 19 per cent, from 2.25 t/ha to 1.81 t/ha, during the period. National data point at stable rice production and productivity for the period (FAO 2016, FAOSTAT).

Yield Gaps

One out of several ways to conceptualize the term yield gap is to view it as the differences between the average crop yield (area productivity) achieved by farmers and what is achievable using more optimal cultivars, inputs, and other management practices. The level of productivity in staple crop (p.35) production across countries in SSA is generally far below that of developing countries in other regions. While this situation has been observed for many years, more recently attention has also been drawn to the large differences in crop yields achieved within the subcontinent. Yield gaps in sub-Saharan Africa are large, among the largest globally, and particularly so in the case of maize (Mueller and Binder 2015). Tittonell and Giller (2013) estimate current yield gaps for major food crops in Africa, and conclude that yield gaps for almost all crops in all regions remain wide and are likely to increase if soil degradation continues.

One reason for the surging focus on yield gaps relates to the growing awareness that yield growth in already cultivated areas of the world needs to increase at a fast rate in order for global food production to keep pace with demand. Demand is expected to grow substantially in the coming years due to a combination of population and income growth. The realization that yield growth has begun to stagnate in many areas in both developed and developing countries explains the growing research interest in agricultural intensification and, as a part of this complex, in the analysis of existing yield gaps (Beddow et al. 2014, FAO and DWFI 2015). Narrowing yield gaps in SSA might also contribute to the post-2015 global Sustainable Development Goals of ending hunger and achieving food security by 2030.

The Afrint data on yields for the two staple crops, maize and rice, concur with the observations about large yield gaps in SSA. In order to get an indication of what is achievable under local conditions, we have defined the attainable yield as the mean yield of the 20 per cent best-performing farmers per crop and village (outliers excluded). The average difference between that level and each individual farmer’s yield we label the yield gap. Unlike the more common use of the term, we do not refer to the difference between the actual and the agronomic yield potential of the crop (FAO and DWFI 2015). Rather we use the concept yield gap to capture the difference between what is already the attainable yield under local conditions and realities—bio-physical as well as socio-economic—and what is the actual yield for each farmer in the village. This, we think, gives a clear indication of the potential for improvement of yields at the local level.

The aggregates based on village means of the top 20 per cent farms with the highest yields for the three staple cereals are presented in Tables 2.5 and 2.7. As shown, the gaps are substantial in Afrint III—49 per cent for maize and 46 per cent for rice. The percentages indicate the discrepancy between the village attainable yield and the yield of the farmers in the village. For maize, the data for Afrint III show that the top 20 per cent of farms attained yields on average two times higher than the individual farmers’ yield. Comparing the situation over the three different time periods, we observe no marked change. The gaps (p.36) (p.37) (p.38) are within the range of 47 per cent to 49 per cent for maize and 45 per cent to 47 per cent for rice.3

One argument in favour of a stronger focus on yield gaps relates to poverty reduction, and suggests that by closing or reducing yield gaps, poor farmers trapped in low-productivity farming would benefit. As shown by Dzanku et al. (2015), drawing on Afrint panel data for twenty-one regions in eight SSA countries, poverty gaps do increase as yield gaps grow. Whether investment made to close such gaps would be the most efficient way to reduce poverty among African smallholders remains, however, uncertain. The same question also seems relevant when addressing another and related argument, namely that, by closing yield gaps, food security would improve for smallholders, particularly among subsistence farmers. As discussed in a comprehensive analysis of more than 13,000 small farms in seventeen SSA countries,4 a strategy to improve market access and to increase off-farm opportunities would be more efficient in improving food security than a focus on agricultural production and closing yield gaps (Frelat et al. 2016). Among the several possible factors explaining yield gaps, the use of productivity-increasing technologies is often emphasized. Among these, the use of improved seed varieties and fertilizers receive a lot of attention. In the following section, the focus is set on these technologies.

Technology Adoption

Sustainable intensification has become a lead concept in agricultural development generally, and specifically in the discussion of agricultural development in SSA. Although area expansion will remain important for increased food production for several decades to come, it is broadly recognized that there is an urgent need for strategies targeting sustainable intensification of already cultivated land. This recognition is partly the result of two current and simultaneous trends, namely the growing land pressures in many countries and regions within countries and the unsustainability of current low-productivity agriculture which is based on nutrient mining, causing different forms of land degradation and stagnating yields.

(p.39) It is in this context that new technology and technology adoption becomes particularly important in an SSA perspective. Globally, the introduction of new productivity-increasing technologies has been central for agricultural development during the past half century. Large parts of tropical and subtropical Asia and Latin America benefited tremendously from intensification processes popularly referred to as the Green Revolution (Djurfeldt and Jirström 2005). During the same period, development and diffusion of new technologies in Africa generally lagged behind, although there are several well-documented examples of productivity spurts sharing several features with those of the Green Revolution (Holmén 2005). Adoption rates are therefore currently lower in SSA than in countries in Asia and Latin America; nevertheless, many African smallholders are familiar with, for example, new seed and fertilizer technologies (Sheahan and Barrett 2014).

Focusing in this section on the adoption of seed technology and the use of inorganic fertilizer, we show that adoption rates of improved maize seeds and for fertilizer were quite high during the period studied (Table 2.8).5 In the Afrint III period, 50 per cent of maize farmers used improved seeds and 55 per cent applied inorganic fertilizer. The use of improved rice varieties was 20 per cent in Afrint III, a halving since Afrint II. Fertilizer application had, however, become more common with 30 per cent of farmers applying it in Afrint III compared to 19 per cent in Afrint II. For the two crops, the share of households investing in fertilizer use has grown and the usage in the Afrint III period is quite high.

Table 2.8. Seed and fertilizer use (share of famers using)

Change over time

Afrint I

Afrint II

Afrint III

Afrint I–II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Maize

All farmers

Proportion using improved seed (%)

45

53

50

8

***

−3

**

5

***

Proportion using fertilizer (%)

41

41

55

0

14

***

14

***

20% best-performing farmersa

Proportion using improved seed (%)

47

62

54

15

***

−8

**

7

**

Proportion using fertilizer (%)

50

46

57

−4

11

***

7

***

5% best-performing farmersa

Proportion using improved seed (%)

53

61

55

8

−7

1

Proportion using fertilizer (%)

48

50

59

2

9

11

Rice

All farmers

Proportion using improved seed (%)

18

40

20

21

***

−20

***

2

Proportion using fertilizer (%)

17

19

30

1

11

***

12

***

20% best-performing farmersa

Proportion using improved seed (%)

18

45

24

28

***

−21

***

7

Proportion using fertilizer (%)

19

22

31

3

8

12

**

5% best-performing farmersa

Proportion using improved seed (%)

15

42

23

27

***

−19

*

8

Proportion using fertilizer (%)

24

23

33

−1

10

8

Note: (a.) Based on mean yield, 3-season average.

Possibly related, it can also be noted that since the mid-2000s several African countries subsidize fertilizer at high rates. Despite what may seem a somewhat confusing stand by the World Bank as to whether governments should or should not subsidize fertilizer programmes (Mellor 2014), all but Mozambique of the Afrint countries have such programmes. In Mozambique, where the use of fertilizer is very low and most farmers are not familiar with the technology, there is not a general programme but pilot tests have been conducted and an expansion of test areas is taking place (Simtowe 2015).

Technical advances such as the use of improved seed or employment of improved nutrient management can trigger and drive increases in surplus farm production that can be marketed. According to Wiggins et al. (2011), technical advances can be seen as one of two important drivers of the commercialization of small-scale farming, the other being higher prices and better access to markets. We now turn our attention to these issues. (p.40)

(p.41) Commercialization and Market Integration

Although far from all observers consider commercialization as an inherently positive process,6 the most common policy stand taken by governments and international development organizations like the World Bank is to promote the commercialization of the smallholder sector. This standpoint is based on the notion that commercialization generally will benefit the farmers involved, and also contribute to economic growth and improved livelihood opportunities outside the farm economy (World Bank 2007). Considering Africa’s population growth rate and the increasing numbers of urban residents depending on the market for their food needs, it would be hard not to appreciate the importance of agricultural commercialization.

Given the heterogeneity in terms of access to land, capital, labour, and skills found in many smallholder communities, commercialization does not unfold evenly across farms. This should be kept in mind when analysing data on commercialization. When right after the harvest, for example, a low-income farm household sells a high share of its maize in spite of knowing that it will soon have to buy maize but at a higher price, its level of commercialization may indeed have increased, but such a distress sale tells us that it may not represent a positive change for that household.

With this reflection as a cautionary backdrop, data on crop producers’ participation in the market as well as the share of produce marketed are presented in Table 2.9. We also show the absolute amount of staple cereals sold. The great majority (84 per cent) of all households sold some of their crops in Afrint III. The share has grown over the period studied. As expected, non-food crop producers are very market-oriented, with 87 per cent participating in the market. Also, the share of commercialized growers of food crops other than maize, sorghum, and rice is large, with 70 per cent of these producing for the market. By generating immediate income through the year for cash-strapped households, the sale of crops like vegetables, beans, and potatoes is important. Looking at its overall importance as a source of cash income for the entire sample (not only growers), the category made up on average 20 per cent of total cash income in the Afrint III sample, and sale of these types of food crops was thereby the second most important source of cash income.

Table 2.9. Share of farmers selling and amount marketed by type of crop

Change over time (%)a

Afrint I

Afrint II

Afrint III

Afrint I–II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Maize

Proportion of growers who sell the crop (%)

35

45

49

11

***

4

***

15

***

Average amount sold, sellers (t)b

0.63

0.98

1.44

55

***

47

***

127

***

Median amount sold, sellers (t)b

0.30

0.36

0.51

Average proportion of total production sold, all farmers (%)

16

39

23

23

***

−16

***

7

***

Sorghum

Proportion of growers who sell the crop (%)

20

4

10

−16

***

6

***

−10

***

Average amount sold, sellers (t)b

0.21

0.10

0.18

−53

***

86

*

−13

Median amount sold, sellers (t)b

0.11

0.06

0.10

Average proportion of total production sold, all farmers (%)

6

2

3

−5

***

2

**

−3

***

Rice

Proportion of growers who sell the crop (%)

61

51

58

−9

***

7

**

−3

Average amount sold, sellers (t)b

0.59

0.69

1.32

18

89

***

123

***

Median amount sold, sellers (t)b

0.30

0.42

0.60

Average proportion of total production sold, all farmers (%)

26

25

30

−1

5

***

4

**

Other food cropsc

Proportion of growers who sell the crop (%)

65

73

70

8

***

−2

*

6

***

Non-food cropsc

Proportion of growers who sell the crop (%)

96

87

n.a

−9

***

n.a

Any type of cropc

Proportion of growers who sell the crop (%)

71

80

84

8

***

4

***

12

***

Notes: (a.) Change in proportion of growers who sell the crop and average proportion of total production sold is absolute not percentage change.

(b.) Extreme values removed at the dataset level.

(c.) Data on amount sold was not available.

The cereal staples, for which we have richer data, continue to be important sources of cash income for many households. Since Afrint II, the proportion of growers selling cereal crops has increased. Approximately half of all maize growers, which includes 92 per cent of all households, sell maize. For rice (p.42) (p.43) growers the share is higher (58 per cent), while only a tenth of sorghum growers market the crop. Part of the explanation for the low share of marketed sorghum may be the popularity of the crop for beer production, which opens up possibilities for on-farm processing and sale at the local level.

Between Afrint II and III, the average amounts of maize and rice sold increased substantially and, for the Afrint III sample, 1.44 t of maize and 1.32 t of rice were the average amounts sold per farm (all sellers considered). Comparing means and medians, the skewed distribution of marketed output implies that, for the great majority, the amounts marketed are modest, and in most cases limited to a few bags. As noted earlier, this implies that, for many, the sale of these staples is followed by later purchases, eventually turning many sellers into net buyers. The average proportion of total production sold was very low for sorghum, presumably for the reason previously mentioned. For maize and rice we can note that some 23 per cent of all maize produced and 30 per cent of rice were marketed. In the study by Frelat et al. (2016) referred to earlier, covering more than 13,000 small farms in SSA, the share of food crops sold was 20 per cent, a level that our results tally well with.

An indication of the relative importance of cereals staples as a cash earner is that in both Afrint II and Afrint III staple food sales constituted the single most important source of cash income, with an average share of 24 per cent in Afrint II and 26 per cent in Afrint III. For the six countries, only in Kenya were staple crops not the most or second most important cash income source. Taken together, the sale of crops on average made up more than half (54 per cent) of all cash income in the Afrint III sample. Adding the share from the sale of animal products (9 per cent), agricultural production represented 63 per cent of total cash income among the Afrint households. When including agricultural work on others’ farms as a part of farm income, the total share of farm income in relation to total cash income reaches 70 per cent, ranging from 49 per cent in Mozambique to 85 per cent in Zambia. This share comes close to estimates by Haggblade et al. (2007) and Jayne et al. (2006a). The relative importance of farm income sources has not changed since Afrint II, a sobering reminder of the heavy reliance on agriculture in the livelihoods of African smallholders.

Non-Farm Linkages and Diversification

African smallholders in rural areas7 are often farming in regions where economic sectors other than agriculture are significant. Estimating the non-farm share of the rural economies across the developing world, (p.44) Haggblade et al. (2010) show that 35 per cent of rural household income in Africa is derived from non-farm sources. Moreover, in their review, Barrett et al. (2001) indicate an even larger figure of between 40 per cent and 45 per cent for rural Africa. For Asia and Latin America their estimate is approximately 50 per cent. Landless and near-landless households are particularly dependent on the rural non-farm economy, but non-farm income sources are also important for smallholders who have access to more land.

Much hope is presently being pinned on the development and growth of the rural non-farm economy in SSA. By providing pathways out of poverty, such a development seems imperative in order to meet challenges such as pervasive rural poverty, decreasing farm sizes, and the rapid growth in the number of young rural people entering the labour market (Losch et al. 2012). In its analysis of the role of agriculture in contributing to poverty reduction and inter-sectoral growth in SSA, the World Development Report 2008 is explicit when pointing out that agriculture cannot be expected to drive the growth and transformation of the rural economies in all regions where smallholders operate (World Bank 2007). In areas with low potential for agricultural development, other development strategies must, according to the report, be sought, including the provision of safety nets by governments. Not even in the more medium- and high-potential areas (in terms of agro-ecology and market access) can all smallholders be expected to farm their way out of poverty. The World Development Report 2008 does not, however, enter into the challenging questions of how to finance the proposed safety nets.

For resource-poor smallholders, the diversification of livelihoods is often a necessity caused by economic distress, rather than a strategy to seek attractive income opportunities. Moreover, even for the non-farm sector, substantial entry barriers limit access to high-return rural non-farm income to relatively better-off households, while the poor are mainly confined to low-return non-farm activities. In a review of rural livelihood diversification in SSA, Alobo Loison (2015: 1134) concludes that while the benefits of the diversification into non-farm income sources mainly favour the better-off, diversification ‘still provides a safety net for the rural poor and sometimes offers a means for upward mobility’.

The Afrint II and Afrint III surveys provide data on different types of cash incomes sources. In Table 2.10 these are grouped into farm and non-farm sources. Following a classification of income sources according to economic sectors, agricultural labour is here included in the overall category ‘farm income’. On average, non-farm income sources account for 30 per cent of total household cash income.8 There is variation between the country (p.45) aggregates, with the proportion of non-farm income (NFI) in the Afrint III period being the lowest in Zambia (15 per cent) and the highest in Mozambique (51 per cent). In the case of Zambia, it is the dramatic increase in food staple production and sales that have reduced the overall share of NFI from 29 per cent to 15 per cent between Afrint II and Afrint III. In Mozambique, the NFI share was stable over the period and its greater importance is due to the domination of subsistence farming.

Table 2.10. Non-farm income shares

Change over time

Afrint I

Afrint II

Afrint III

Afrint I–II

Sig.

Afrint II–III

Sig.

Afrint I–III

Sig.

Ghana

NFI as a share of total income (%)

28

39

n.a.

11

***

n.a.

Share of households having NFI (%)

49

54

75

5

22

***

26

***

Kenya

NFI as a share of total income (%)

18

26

n.a.

8

***

n.a.

Share of households having NFI (%)

70

32

56

−38

***

24

***

−14

***

Malawi

NFI as a share of total income (%)

32

23

n.a.

−8

***

n.a.

Share of households having NFI (%)

67

60

52

−7

*

−8

**

−15

***

Tanzania

NFI as a share of total income (%)

33

26

n.a.

−7

***

n.a.

Share of households having NFI (%)

71

50

50

−21

***

0

−21

***

Zambia

NFI as a share of total income (%)

29

15

n.a.

−14

***

n.a.

Share of households having NFI (%)

52

56

39

4

−17

***

−13

***

Mozambique

NFI as a share of total income (%)

48

51

n.a.

3

n.a.

Share of households having NFI (%)

54

47

60

−6

*

13

***

6

*

Total

NFI as a share of total income (%)

31

30

n.a.

−1

n.a.

Share of households having NFI(%)

60

51

56

−9

***

5

***

−4

***

While on average NFI makes up an important share of total cash income, still many households report not having any NFI at all (Table 2.10). As much as 44 per cent of the Afrint III sample did not report any NFI. Ranging on average between 50 per cent and 60 per cent, the share of households which do generate NFI has been stable during the period. In Afrint III, two countries diverged from this level. In Ghana, 75 per cent of households engaged in non-farm activities, and in Zambia only 39 per cent reported having NFI. In Zambia, the proportion of households participating in the non-farm sector dropped from 56 per cent in Afrint II to 39 per cent in Afrint III, a change which may reflect the boom in maize production. In Ghana, the share of households engaging in micro-business increased significantly (16 per cent), contributing to the growing importance of that source of income, which increased from 8 per cent to 16 per cent of total household cash income between the Afrint II and Afrint III periods.

Production, Commercialization, and Income Diversification by Gender and Income Group

Following the two themes of this volume, this section of the chapter presents findings from a cross-tabulation of indicators for production, productivity, commercialization, and diversification comparing FMFs and male-managed farms (MMFs) as well as four income groups. The analysis depicts income patterns and income diversification among the sampled farm households on the basis of these comparisons. More detailed analysis of gender differences of production, commercialization, and income diversification at the country and regional levels follows in Chapters 3 and 4.

Gender Differences

As noted in Chapter 1, FMFs constitute 29 per cent of the Afrint III sample. The survey data show that landholdings held by women were only approximately (p.46) (p.47) half the size (56 per cent) of those held by men (Table 2.11). For maize, the mean cultivated area of FMFs of 0.86 ha constitutes 69 per cent of that of MMFs. This difference explains most of the gender difference in maize production with male-headed households producing 1.64 t per farm compared with 0.86 t for their female counterparts. In terms of area productivity, the gap is not big. In Afrint III, FMFs achieved 90 per cent of the maize yield of MMFs, a gap which was the same in Afrint I.

Gender-based differences in the adoption rate of seed and fertilizer technology in maize cultivation were not very large in Afrint III (Table 2.11). The share of FMFs using improved seeds and inorganic fertilizers was 45 per cent and 49 per cent respectively. For male-headed households, the corresponding proportion of users was 51 per cent and 56 per cent. However, lacking detailed data on use, for example, amounts and types of fertilizers, we cannot further explore potential differences in technology application.

Table 2.11. Production, commercialization, and diversification by sex of farm manager

Afrint I

Afrint II

Afrint III

Male

Female

Diff.

Sig.

Male

Female

Diff.

Sig.

Male

Female

Diff.

Sig.

No. of households

1,609

712

1,811

661

1,803

734

Proportion of sample

69

31

73

27

71

29

Mean farm size (ha)

2.21

1.45

0.77

***

1.90

1.13

0.77

***

2.63

1.49

1.13

***

Mean area under maize (ha), 3-season averagea

0.96

0.67

0.28

***

0.93

0.61

0.32

***

1.19

0.82

0.37

***

Mean maize production (t/farm), 3-season averageb

1.07

0.72

0.35

***

1.21

0.72

0.50

***

1.64

0.86

0.78

***

Mean maize yield (t/ha),c 3-season averaged

1.34

1.20

0.14

**

1.23

1.17

0.07

1.29

1.15

0.14

***

Proportion using improved maize seed (%)

46

43

3

55

49

6

**

51

45

6

***

Proportion using fertilizer on maize (%)

39

45

−6

**

41

39

3

56

49

7

***

Proportion of maize growers selling maize (%)

39

26

13

***

48

38

11

***

53

40

13

***

Proportion of total maize production sold (%)

18

11

7

***

41

32

9

***

25

17

8

***

Proportion of growers of other food crops selling (%)

68

56

12

***

75

67

8

***

74

60

14

***

Proportion of growers of non-food crops selling (%)

96

96

0

88

81

7

**

Proportion selling any crop (%)

76

62

14

***

83

72

11

***

88

73

15

***

Sale of staple crops as a proportion of total cash income (%)

26

19

7

***

28

21

7

***

Agricultural wage labour as a proportion of total cash income (%)

5

9

−4

***

5

9

−4

***

NFI as a proportion of total cash income (%)

29

39

−10

***

28

37

−10

***

Mean household income per adult equivalente, 2010 PPP-adjusted USDf

281

297

−16

317

226

91

***

Notes: (a.) Average of area under maize during the last growing season at the time of data collection and the two previous growing seasons before that one.

(b.) Average of maize production during the last growing season at the time of data collection and the two previous growing seasons before that one.

(c.) Based on a sub-sample including maize growers who cultivated an average of at least 0.1 ha. The sample also excludes cases with a sixfold or higher yield increase between consecutive Afrint rounds.

(d.) Average of maize yield during the last growing season at the time of data collection and the two previous growing seasons before that one.

(e.) The use of adult equivalents takes the age composition of the household into account, by assigning adult household members (16–60) a value of 1, whereas children (0–15 years of age) are given value of 0.50 and elderly household members (61 years and above) are given a value of 0.75, when calculating the number of adult equivalents per household.

(f.) Afrint II incomes are adjusted from 2008 values; Afrint III incomes are adjusted from 2013 values for Ghana, Kenya, Malawi, and Zambia, and 2015 values for Tanzania and Mozambique. Extreme values removed at the village level.

Turning to the indicators of commercialization, the difference is larger with only 40 per cent of female farm managers selling maize while 53 per cent of the male respondents sold some of their maize in the Afrint III period. Furthermore, FMFs tend to sell a smaller share of their maize—17 per cent compared to 25 per cent for the MMFs. In comparison to the Afrint II period, the share of maize sold declined substantially to 32 per cent and 41 per cent for FMFs and MMFs respectively, but the gap between the two remained at the same level. The gap in the level of commercialization can perhaps be related to the much smaller total production on FMFs and the fact that, in many households, maize is a popular and important food staple. In this context, we can also note that while male landholders on average set 45 per cent of the farm area to maize, FMFs used 55 per cent of their land for maize in the Afrint III period. The pattern for maize is similar for the crop category of other food crops. In the case of non-food cash crops, the difference is smaller with 81 per cent of the FMFs marketing their produce compared to 88 per cent of the MMFs.

Gender differences in income and income diversification are substantial, with FMFs generating annual cash incomes/per adult equivalent to only approximately 70 per cent of those of MMFs (see also Chapter 4). For MMFs, on average 72 per cent of total cash income derives from farm sources while for FMFs the share is 63 per cent. Being more dependent on NFIs for their total income, FMFs also depend more on agricultural wages. For Afrint III, work on others’ land generates 9 per cent of total income for FMFs while only 4 per cent for MMFs.

Diversification by Income Group

By dividing the farm households into four groups according to their cash income sources—farm only, non-farm only, both, or none—several important (p.48) (p.49) observations indicating substantial differences in terms of production, productivity, and commercialization and total cash income can be made (Table 2.12). The great majority of households belong either to the group relying completely on agriculture for cash income (40 per cent) or the group combining farm and NFI sources (49 per cent). Only 7 per cent of households report that they depend completely on non-farm sources of income. The remaining 5 per cent report no cash income at all.

Table 2.12. Production, commercialization, and diversification by income group

Afrint I

Afrint II

Afrint III

Only farm income

Only non-farm income

Both income sources

No income

Only farm income

Only non-farm income

Both income sources

No income

Only farm income

Only non-farm income

Both income sources

No income

No. of households

831

305

1,082

107

1,096

252

1,006

119

1,006

182

1,240

116

Proportion of sample

36

13

47

5

44

10

41

5

40

7

49

5

Mean farm size (ha)

2.30

1.31

1.98

1.39

1.72

1.07

1.86

1.27

2.37

1.47

2.44

1.38

Mean area under maize (ha), 3-season averagea

0.94

0.70

0.88

0.70

0.80

0.67

0.91

1.01

1.11

0.92

1.10

0.87

Mean maize production (t/farm), 3-season averageb

0.96

0.48

1.13

0.58

1.03

0.59

1.25

1.18

1.68

0.64

1.37

0.62

Mean maize yield (t/ha),c 3-season averaged

1.25

0.94

1.48

0.87

1.22

0.91

1.31

1.01

1.35

0.85

1.26

0.76

Proportion using improved maize seed (%)

43

24

54

38

55

24

61

30

57

22

49

28

Proportion using fertilizer on maize (%)

38

26

48

28

39

16

51

21

61

20

56

24

Proportion of maize growers selling maize (%)e

43

5

40

13

51

13

51

16

56

5

54

6

Proportion of total maize production sold (%)e

20

2

17

6

43

5

45

17

28

1

24

4

Proportion of growers of other food crops selling (%)e

76

10

73

21

78

22

79

39

78

18

74

35

Proportion of growers of non-food crops selling (%)e

96

100

96

95

92

45

88

35

Proportion selling any crop (%)e

85

13

82

19

89

25

86

45

93

25

89

35

Sale of staple crops as a proportion of total cash income (%)

36

0

18

0

39

0

19

0

Agricultural wage labour as a proportion of total cash income (%)

10

0

3

0

10

0

4

0

NFI as a proportion of total cash income (%)

0

100

49

0

0

100

45

0

Mean household income per adult equivalent,f 2010 PPP-adjusted USDg

243

141

371

0

213

179

373

0

Notes: (a.) Average of area under maize during the last growing season at the time of data collection and the two previous growing seasons before that one.

(b.) Average of maize production during the last growing season at the time of data collection and the two previous growing seasons before that one.

(c.) Based on a sub-sample including maize growers who cultivated an average of at least 0.1 ha. The sample also excludes cases with a sixfold or higher yield increase between consecutive Afrint rounds.

(d.) Average of maize yield during the last growing season at the time of data collection and the two previous growing seasons before that one.

(e.) The data are based on households’ answers as to whether they had sold or intended to sell any maize following the most recent crop year. The classification of household in the four categories shown in the table is based on respondents’ answers about incomes from different income sources during the past year. This explains why a number of households belonging to the income categories ‘only non-farm income’ and ‘no income’ are included as sellers of crops. The number of cases in Afrint III for maize was 8 and 6 respectively, for other food crops 28 and 34, for non-food crops 14 and 8, and for any crop sold 44 and 39. The number of cases in Afrint II for maize was 28 and 17 respectively, for other food crops 31 and 36, for non-food crops 10 and 20, and for any crop sold 62 and 54.

(f.) The use of adult equivalents takes the age composition of the household into account, by assigning adult household members (16–60) a value of 1, whereas children (0–15 years of age) are given value of 0.50 and elderly household members (61 years and above) are given a value of 0.75, when calculating the number of adult equivalents per household.

(g.) Afrint II incomes are adjusted from 2008 values; Afrint III incomes are adjusted from 2013 values for Ghana, Kenya, Malawi, and Zambia, and 2015 values for Tanzania and Mozambique. Extreme values removed at the village level.

Starting with the biggest group, households combining farm and NFIs, the average total household income of this group is approximately double that of the groups relying completely on either farm or NFI. This finding is broadly in line with previous research pointing to the benefits of supplementing farm incomes with NFIs among African smallholders (Alobo Loison 2015). Compared to the group relying completely on NFI to generate cash, the group combining income sources has larger mean farm size (2.44 ha compared to 1.47 ha), attains higher mean maize yields (1.26 t/ha compared to 0.85 t/ha), and more commonly uses improved maize seeds and fertilizer on maize.

The 40 per cent of farm households completely dependent on agriculture for cash income do not show statistically significant differences from the ones combining farm and NFI sources in terms of production, productivity, and commercialization. The only significant difference is in the proportion using improved maize where the farm-only group has a significantly higher proportion of households using improved maize seeds than any of the other groups.

Likewise, the group of 116 households reporting not to have any cash income shows no significant difference from the group that depends on NFI in several characteristics, including farm size, maize yield, and a less common usage of improved seeds and fertilizers. Both groups, however, show significantly lower results for all of these factors as compared to the groups of households whose cash incomes are specialized in agriculture and those having both agricultural and NFIs. Conclusions about this group of households reporting to have no cash income must, however, be drawn with care. Several of the households reported that they sold or intended to sell some of their crop produce, and thus the true number of households that belong to this group can be questioned (see note in Table 2.12). Nonetheless, we note that this group of households appears to be very resource weak.

Concluding this section, we emphasize the finding that the large group of households that are completely dependent on farm activities for cash income, on average have low cash incomes in spite of good access to land, a relatively high share of households using improved seeds and inorganic fertilizer for maize, and are comparatively commercialized. This prompts a question which decision-makers and policymakers need to consider: would the farm households in this group be able to farm their way out of poverty if they had better (p.50) (p.51) (p.52) access to affordable technologies, better access to markets, and better credit and insurance facilities?

Conclusion

Since Afrint II the approximately 2,500 farm households studied have increased their agricultural production substantially. On average, the production of maize, sorghum, and rice increased by more than 30 per cent. This was made possible by a roughly corresponding expansion of the area cultivated with these crops, but also other food crops such as vegetables, roots and tubers, and fruits. Thus, the mean farm area cultivated increased by 36 per cent and reached 2.30 ha in 2013. FMFs, which make up approximately 30 per cent of all farms in the sample, are significantly smaller than male-held landholdings, however. Cultivating only nearly half as large areas, and, in the case of maize, attaining about 90 per cent of the yield per hectare of male-headed households, female-headed households produce substantially less farm output.

Productivity measured in terms of output per area did not change significantly on average. There is variation between countries but broadly for the whole sample and for the whole period between Afrint I and Afrint III, changes in production depend generally on corresponding changes in the area cultivated. While yields for the main cereal staples remain low, the gap between the top performers in each of the fifty-six villages and the great majority is substantial, with the top 20 per cent of farms producing approximately twice as much per area. This indicates that higher yields are attainable under local conditions and realities—biophysical as well as socio-economic—and thus there is a clear potential for raising local yields.

The general increase in crop production resulted in significant increases in the maize and rice volumes marketed. The overall proportion of farm households who marketed some share of their different crops harvested increased since the Afrint II survey, reaching 84 per cent in Afrint III. Taken together, the sale of crops, on average, makes up more than half of total household cash income and, adding to this other agricultural income sources, the proportion of farm income to total income reaches 70 per cent.

To the extent that this description of increases in production and the levels of commercialization signals positive change since Afrint II, a broader view of smallholders’ livelihoods shows that farm households relying entirely on income from agricultural sources are faring much worse than those able to combine farm and non-farm cash income sources. Making up some 40 per cent of the total sample, smallholder farms lacking NFI generate only about half as much cash income as those farm households also able to find (p.53) employment and opportunities in the non-farm economy. In spite of cultivating, on average, more than 2.3 ha and participating in the different crop commodity markets, these households do not, under current conditions, seem to be able to farm their way into the income levels of households with more diversified income sources.

Returning to this chapter’s introduction, and the several major transformatory trends affecting the predominantly agrarian economies of SSA, the picture emerging from our broad description of development in the smallholder sector of six SSA countries clearly points to the need for reinforced support to the sector. At the same time, we have shown that by supplementing farm income with NFI, half of the households studied are able to increase their total income significantly. This points to the opportunities for supportive policies and initiatives targeting the development of the rural non-farm economy to contribute to improved conditions for African smallholders.

Notes:

(1) Statistical levels of significance are set at below 1 per cent (***), 5 per cent (**), and 10 per cent (*), respectively, throughout the book, but extreme cases have been handled differently by the authors and the treatment of extreme cases are discussed individually in each chapter.

(2) In the Afrint III survey, cassava was included in the category ‘other food crops’, whereas in Afrint I and Afrint II it was not. This was due to the uncertainties and problems encountered when collecting data on cassava production in the Afrint I and II surveys. It is assumed that this change has contributed to the increasing area under ‘other food crops’.

(3) If comparing each farmer’s yield with the 5 per cent top-yielding farms the yield gaps increase to between 59 per cent to 61 per cent for maize and 57 per cent to 58 per cent for rice. Because the Afrint village samples contain between thirty and fifty farms, a 5 per cent subsample will contain one to three farms only and the statistical error could be high, so we prefer to use the 20 per cent level when analysing the village sample yield gaps; however, we provide the 5 per cent values here to maintain comparability with previous publications.

(4) The Afrint dataset forms part of data used in the study.

(5) Some caution is warranted in the interpretation of the figures presented in Table 2.8. Farmers reporting on the use of improved seed sometimes refer to recirculated seeds. On the other hand, farmers sometimes refer to improved seed as traditional because they have used it for several years and consequently label it traditional technology.

(6) Stockbridge (2006) and Wiggins et al. (2011) offer reviews of agricultural commercialization in Africa which include discussions of different and sometimes competing perspectives on the social and economic impacts of commercialization.

(7) Long neglected, African urban farmers are increasing in number and importance. In this volume, however, the focus is on smallholders in rural areas.

(8) Cash income only forms part of total household income for households retaining part of their agricultural output for their own consumption, payment for hired labour, seeds, etc. If this part of the production was to be valued at the price of marketed output and added to the amount of cash income, the overall share of farm income would be markedly higher.