Household Determinants of the Adoption of Improved Cassava

Journal of Agricultural Economics doi: 10.1111/1477-9552.12247

Household Determinants of the Adoption of Improved Cassava Varieties using DNA Fingerprinting to Identify Varieties in Farmer Fields: A Case Study in Colombia Victorino O. Floro IV, Ricardo A. Labarta, Luis A. Becerra pez-Lavalle, Jose M. Martinez and Tatiana M. Ovalle1 Lo (Original submitted November 2016, revision received May 2017, accepted July 2017.)

Abstract We examine factors affecting the adoption of improved cassava varieties of 217 households in the Cauca Department in southwest Colombia. Using DNA fingerprinting through Single Nucleotide Polymorphisms (SNPs), we identified different cultivars in farmers fields. We also used this information to remove possible bias in the adoption model that could have resulted from a misclassification of improved varieties (IVs). As a result, we found that farmers substantially overestimate their use of IVs and there are important differences in the determinants of adoption between farmer self-identification and DNA fingerprinting. This finding implies that the incorporation of DNA fingerprinting in IV adoption studies is important to

1

Victorino Floro IV is a CIAT Visiting Researcher from the Global Human Development Program at Georgetown University, USA. E-mail: [email protected] for correspondence. Ricardo Labarta is Senior Scientist and Impact Assessment Research Leader, International Center for Tropical Agriculture (CIAT), Colombia. Luis Becerra Lopez-Lavalle is Program Leader, Agrobiodiversity Research Area Cassava Program, International Center for Tropical Agriculture (CIAT), Colombia. Jose Martinez is a Research Assistant, International Center for Tropical Agriculture (CIAT), Colombia and an adjunct professor at the Economics Department of Universidad del Valle (Colombia). Tatiana Ovalle is a Research Assistant, Cassava Genetics Program, International Center for Tropical Agriculture (CIAT), Colombia. The authors would like to thank the CGIAR Research Program on Roots Tubers and Bananas and the Global Human Development Program at Georgetown University for providing financial support for this research. The authors are also grateful to Greg Traxler, Derek Byerlee, Jeff Alwang, Mywish Maredia, Byron Reyes and three anonymous referees for their helpful comments, and to Jairo Gomez for outstanding data cleaning assistance. The authors would also like to express their gratitude to Ruben Dario Rojas, Dominique Dufour, Fernando Calle, John Ocampo, Consuelo Montes, Diego Gomez, William Espinoza, Maria Constanza Perez, Kevin Sanchez, Juan Carlos Gallego and Diego Cifuentes for their contribution during the study. Ó 2017 The Authors. Journal of Agricultural Economics published by John Wiley & Sons Ltd on behalf of Agricultural Economics Society. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.

2

Victorino O. Floro IV et al. ensure the accuracy of future agricultural economic research and the relevance of subsequent policy recommendations. Keywords: Cassava; adoption; Colombia; DNA fingerprinting. JEL classifications: O13, Q10, Q16.

1. Introduction Cassava (Manihot esculenta Crantz) is a key crop in tropical countries because it can be cultivated in marginal conditions, with principal traits including: tolerance to low or unpredictable rainfall; cultivability in acidic or nutrient-deficient soils; easy and low-cost propagation; ability to be grown all-year round (Henry and Hershey, 2002; Howeler et al., 2013). In Colombia, smallholders have traditionally grown cassava but high demand for cassava starch has increased the number of large plantation style farmers. However, nearly all the pests and diseases known to affect cassava are present in the country (Henry and Hershey, 2002) making the adoption and studies on the adoption of improved varieties (IVs)2 critical. Up to now, nearly all studies on the factors affecting adoption of IVs have relied on farmers’ self-reported cultivar names to estimate varietal adoption (Labarta et al., 2015). Despite efforts by recent studies to engage crop experts with photographs and morphological descriptors, many studies have encountered difficulty in verifying crop varieties cultivated by farmers (Larochelle et al., 2015; Walker, 2015). These difficulties are particularly acute in cassava, which is exclusively maintained by vegetative propagation (stem cuts or in vitro culture). Vegetative propagation allows easy exchange of planting materials among farmers, but maintaining the full genetic integrity of these materials is challenging as farmers do not use uniform naming conventions or uniform cultivar arrangements on their farms. In recent years, the incorporation of DNA fingerprinting in agricultural economic studies has been proposed to resolve IV identification issues. We use DNA fingerprinting and econometric analysis to compare the determinants of adoption of improved cassava varieties between farmer self-reported identification and identification by geneticists in the Cauca department of Colombia. 2. Cassava Production and the Use of Improved Varieties in Cauca In Colombia, cassava remains a staple food, retaining a special status in food and livelihoods of the poor, particularly for cassava starch making (Hershey et al., 2000; Howeler et al., 2013). In addition, cassava genetic diversity in the Americas provides numerous opportunities for improvement through research and development. In Colombia alone, there are hundreds of cassava landrace varieties and at least 15 IVs bred by the Centro Internacional de Agricultura Tropical (International Center for Tropical Agriculture, ‘CIAT’) and the Corporaci on Colombiana de Investigaci on Agropecuaria (Colombian Corporation for Agricultural Research, ‘CORPOICA’).

2

We define an improved variety as a cultivar developed through a cassava-breeding programme by the agriculture research and development organisations in Colombia. However, we acknowledge that some farmers may have bred their own versions of IVs over the years through crossing of preferred cultivars for desired traits. Ó 2017 The Authors. Journal of Agricultural Economics published by John Wiley & Sons Ltd on behalf of Agricultural Economics Society.

3

Adoption of Improved Cassava using DNA Testing Table 1 Difference between improved and non-improved by breeding characteristic Characteristic Fresh root yield (t/ha) Dry matter yield (t/ha) Dry matter content (%) Plant type score (1–5) Harvest index (0–1)

SM 1495-5 (Improved variety)

MCOL 1522 (Algodona)

Significance

33.9 11.9 36.8 2.8 0.53

18.4 5.8 33.5 3.23 0.39

** ** ** ** **

Notes: Plant Type Score is scored from 1–5 with 1 as good and 5 as bad; Harvest Index refers to the root weight/total biomass. ** denotes significance at the 5% level. Source: Corporaci on Colombiana de Investigaci on Agropecuaria & Centro Internacional de Agricultura Tropical (CIAT) (2014).

Each of these IVs was developed over at least 8 years through selective crossing of landraces3 and older IVs for five key characteristics: fresh root yield, dry matter yield, dry matter content, plant type score (i.e. resistance to disease and pest tolerance),4 and harvest index5 (H. Ceballos, personal communication, 25 July 2016). Further, these IVs maintain the starch gelatinisation and pasting6 traits of landraces that millers/bakers prefer, while often having superior characteristics relative to landraces under either poor or good planting conditions (CORPOICA-CIAT, 2014). For example, Table 1 shows the five characteristics of the improved variety SM 1495-5 (released in 2014) against MCOL 1522 known as ‘Algodona’ by smallholder farmers in Cauca, which is the most commonly used landrace in the region. In terms of production, cassava harvests across the country have increased by an average of 2.0% per year since 1994 to over 2.6 million metric tons in 2014 (Food and Agriculture Organization, 2014). In Cauca, 30,000 hectares were used for cassava in 2015, representing 6.8% of total of cassava acreage in the country (Departamento Administrativo Nacional de Estadıstica/DANE, 2015). Within the department, the crop is grown in three distinct agro-ecological geographic zones: the low zone, intermediate zone, and the high plateau zone (Jaramillo, 2008).7 Economically, cassava is critical to the starch-making industry in Cauca. Many farmers sell to rallanderıas8 or cassava starch mills, providing an additional source of 3

A landrace refers to a commonly propagated variety that was not developed by CIAT or CORPOICA. 4 Plant type score includes plant architecture and health (resistance to pests and disease). 5 Harvest index refers to the proportion of root weight to overall plant biomass. 6 Starch gelatinisation refers to the process of breaking down the intermolecular bonds of starch molecules with heat and water (occurs during cooking/baking). 7 The low zone is mostly flat at 850–1,150 m above sea level (MASL) with annual precipitation between 1,000–1,700 mm. The intermediate zone is characterised by rolling hills at 1,200–1,500 MASL and an annual precipitation rate of 1,500–1,800 mm. The high plateau zone has smooth slopes at 1,600–1,900 MASL with precipitation around 1,800 mm per year. 8 There are an estimated 150 to 200 cassava starch/flour processing mills in Cauca with majority of them concentrated in the northern 12 municipalities of the department with 69% located in Santander de Quilichao (Corporaci on Regional Del Cauca, 2005; Torres et al., 2007; Jaramillo, 2008). Ó 2017 The Authors. Journal of Agricultural Economics published by John Wiley & Sons Ltd on behalf of Agricultural Economics Society.

4

Victorino O. Floro IV et al.

income. Given its importance, some of Cauca’s farmers have acquired IVs through government extension services over the past 30 years. Jaramillo (2008) conducted a comprehensive study on the state of the cassava sector in Cauca and found that, as intended, released IVs had higher yields, better root quality, better growth, and were more tolerant to disease (namely ‘cuero de sapo’ or frog skin disease). 3. Previous Adoption Studies As IVs are crucial in improving agricultural productivity, studies on their adoption are important in directing policy and increasing knowledge of on-farm conditions and farmer decision-making. Likewise, accurate identification of crop varieties is critical to correctly assess the impact of crop improvement research. Yet, two commonly used identification approaches, elicitation of variety names from farmer interviews and morphological plant descriptors, involve much uncertainty (Rabbi et al., 2015). To mitigate this, numerous studies have used various methods in attempts to accurately estimate the prevalence of IV adoption. For example, Agwu and Anyaeche (2007) utilised both variety names in the local language and morphological characteristics to identify improved cassava varieties in Nigeria. Others such as Muhammad-Lawal et al. (2012) employed multistage sampling techniques focused in areas with recent dispersion of improved cassava to ensure identification of IVs. A more recent study by Awotide et al. (2015) utilised data collected for cassava from the CGIAR’s Diffusion and Impact of Improved Varieties in Africa (DIIVA) project,9 which involved the validation of experts’ estimates of adoption against household and community survey data. In spite of these efforts, cassava varietal identification remains difficult since there are often differences between genetically unique cultivars and variety names as elicited from farmers (Rabbi et al., 2015). Furthermore, other confounding factors that may hinder proper varietal identification include: (1) some farmers’ inability to identify varieties by names, (2) the inconsistency in the names of the varieties as identified by the farmers and those of official government/agriculture extension records (i.e. varieties may have locally adapted names), and (3) the loss of genetic identity. Importantly, this loss of genetic identity is due to the incorporation of volunteer seedlings from cross-pollinated seed generated by previously planted cultivars. This is a mechanism traditional cassava farmers practice to increase genetic variability and avoid genetic erosion (Pujol et al., 2005). Morphological identification of cassava also poses substantial challenges. Cassava can be identified by 54 morphological descriptors with farmers most familiar with stem branching habit and colour of the root cortex (Alves, 2002; Fukuda et al., 2010). Since genotype interacts with environmental conditions (M€ uhlen et al., 2000; Alves, 2002; Peroni et al., 2007), and with the large number of commercially cultivated cassava genotypes grown in a highly diverse set of ecosystems, it is difficult to make precise characterisations of morphological descriptors. Moreover, farmers in Cauca have been found to mix genotypically distinct but phenotypically similar seed (i.e. seed of

9

The Diffusion and Impact of Improved Varieties in Africa (DIIVA) project was directed and co-ordinated by the CGIAR’s Standing Panel on Impact Assessment and administered through Bioversity International from 2009 to 2013. The US$ 3 million project was the first major study to focus on the diffusion and adoption of improved crop varieties in Sub-Saharan Africa (See Walker and Alwang, 2015). Ó 2017 The Authors. Journal of Agricultural Economics published by John Wiley & Sons Ltd on behalf of Agricultural Economics Society.

Adoption of Improved Cassava using DNA Testing

5

the same named variety) into a single seed lot, inadvertently increasing diversity (Dyer et al., 2007). In view of these difficulties, assuming that farmers’ identified names or expert opinion match genetic differentiation, may well be misleading, with yields, pest-tolerance, and other traits of different varieties being wrongly attributed. It is also plausible that misidentification in previous studies has led to measurement errors. For the dependent variable (ID use) this can lead to biased and inconsistent estimators, particularly if using binary models where the bias can be more evident (Hausman, 2001; Meyer and Mittag, 2014). To demonstrate this bias, with no loss of generalisation, consider a survey where farmers are asked about the use of a j variety of a specific crop. Now allow the farmers to report, on average, an incorrect answer with a false negative at a rate a or a false positive at rate b, yielding: T DR i ¼ ð1 a bÞDi

ð1Þ

with a + b ≤ 1. Here DR i is the binary adoption variable reported by the farmer and DTi is the true condition of adoption. It is straightforward to show that this would yield a biased value for the sample adoption rate: 1X R 1X T Di ¼ ð1 a bÞð Di Þ AdoptionR ¼ n i n i ð2Þ ¼ð1 a bÞAdoptionT Also, Hausman et al. (1998) have already shown that this setting would mean a biased measure for marginal effects in a probability model, where: @PðDR @PðDTi ¼ 1jXÞ i ¼ 1jXÞ ¼ ð1 a bÞ : @xij @xij

ð3Þ

Meyer and Mittag (2014) prove that in many cases with an absence of further information regarding the process of misreporting, something can still be learned from the biased coefficients. However, this depends on the extent that the researcher can feel confident about the direction of the expected bias. Using DNA fingerprinting, with only a 3% margin of error in classification, essentially rules out the presence of bias in our econometric analysis on the determinants behind crop varietal adoption. Since we have both the reported data on crop variety and the ‘true’ DNA data, we can identify the misclassification bias in the reported data. 4. Methodological Approach and Data Used Since DNA fingerprinting has not yet been widely used in adoption studies, the methodology of our study represents a fairly new multidisciplinary approach and synergy of efforts between agriculture scientists and economists. Some pilot studies designed to test the use of DNA fingerprinting in estimating the adoption of improved crop varieties have been recently completed.10 Results from these studies show a 10

The use of DNA fingerprinting in agricultural economics was first presented during the 2015 International Conference of Agriculture Economists in Milan by CIAT researchers Ricardo Labarta, Byron Reyes and Mywish Maredia of Michigan State University and Gregory Traxler from the University of Washington.

Ó 2017 The Authors. Journal of Agricultural Economics published by John Wiley & Sons Ltd on behalf of Agricultural Economics Society.

6


common misidentification of crop varieties in farmer fields and highlight a large number of varieties incorrectly identified by farmers as improved (Type I error) and varieties incorrectly identified by farmers as local cultivars (Type II error). The results of these studies underscore the importance of using DNA fingerprinting for correcting the misidentification of crop varieties (Labarta et al., 2015; Rabbi et al., 2015; Kosmowski et al., 2016; Maredia et al., 2016). As estimating the level of adoption of improved varieties implies the combination of Type I and Type II errors, a potential under or over estimation of the real adoption of improved varieties will depend on the magnitude of both types of error. Some pilot studies related to using DNA analysis in cereal crops (wheat, maize and rice) and beans found an underestimation of the real adoption of improved cultivars (Labarta et al., 2015; Yirga et al., 2015; Maredia et al., 2016) while other studies related to root and tuber crops (sweet potato, cassava and potato) reported an overestimation of the level of adoption of improved cultivars (Hareau et al., 2016; Kosmowski et al., 2016; Maredia et al., 2016). In light of these studies, this study aims to: (1) establish protocols on how to better incorporate DNA fingerprinting in cassava adoption studies, and (2) estimate the level and determinants of adoption of IVs in the Cauca department of Colombia. We compare the determinants of adoption of IVs through two varietal identification strategies: (1) a household survey of farmers, and (2) genetic identification of varieties through DNA fingerprinting. The household survey was conducted in late 2014 and 2015 in the Cauca Department in Southwestern Colombia, an area where cassava has traditionally been grown and continues to be important in the local diet.11 A total of 305 households were surveyed across 19 of 42 municipalities, representing main cassava-producing locales.12 The questionnaire included household characteristics, perceptions of varieties, and field-level data, particularly the name and acreage of each variety planted and reported by interviewed farmers. These same households provided crop samples for DNA fingerprinting. 5. DNA Fingerprinting using Cassava Nanofluidic Dynamic Arrays (SNPY-chip13 ) Cassava genetic profiling has traditionally been assessed by the study of single sequence repeat (SSRs) (Mba et al., 2001; Hurtado et al., 2008). However, there are very few SSRs that produce multi-allelic markers in cassava, making studies of genetic variation or genetic identification costly and time consuming. On the other hand, single nucleotide polymorphisms (SNPs) are biallelic type markers that are more abundant than SSRs, and easy to automate and share among laboratories. One SNP resolving platform that provides a cost effective genotyping system is based on nanofluidic dynamic arrays, developed by FluidigmÒ (San Francisco, CA, USA). In 2014, the first cassava FluidigmÒ

11

To note, though cassava is the second most important crop for indigenous peoples who cultivate up to 14% of their land with the crop (DANE, 2015), we did not survey nor take cassava samples from indigenous households due to strict regulations and the intricate nature of interactions with indigenous community members in Colombia. 12 At the time of the survey, DANE had not released the latest agriculture census and there was little information on the acreage and distribution of the crop in Cauca to aid the surveyors. 13 SNPY-chip refers to Single Nucleotide Polymorphisms-Yuca (cassava) chip. Ó 2017 The Authors. Journal of Agricultural Economics published by John Wiley & Sons Ltd on behalf of Agricultural Economics Society.


7

SNP array (SNPY-chip) of this kind was developed at the CIAT cassava genetics laboratory using next generation sequencing information generated on over 150 LAC landraces. This array was successfully used for the first time in the molecular characterisation of 173 Amazonia landraces from Colombia (Pe~ na-Venegas et al., 2014). This study had a substantial initial cost to develop the RAD-seq library, SNPY-chip, and the initial DNA sequencing, around US $250,000 while the cost of DNA analysis for each sample from Cauca is between US $25–30. Despite the initial cost and the relatively high fixed cost for this study, subsequent studies are expected to be significantly more cost effective. For a similar CIAT study in Vietnam using the same methodology and same RAD-seq library, each sample cost between US$ 10–15. We confirm that 93 SNPs were sufficient to identify duplicated genetic materials in this study. These same materials were also consistently classified under the same class according toxicity, root colour and use. The use of SNP data for accurate genotype identification has also been demonstrated in cacao, grapevine, strawberry and pomelo (Cabezas et al., 2011; Ge et al., 2013; Ji et al., 2013; Wu et al., 2014). The design of the study initially aimed to collect cassava samples from each variety identified by cassava growers and for every household interviewed. However, as cassava is harvested year round, many farmers did not have readily available samples of all the varieties they plant. As a result, the team was able to collect a total of 436 stem samples from 217 households.14 We conducted our analysis on only these 217 households.15 For the households that had planting material at the time of the interview, we first asked farmers to identify all cassava varieties that they had grown in the last growing season and to estimate the area under each cassava variety. We then requested farmers during the same visit to provide a small quantity of cassava stems to represent those varieties identified during the interview. These were brought back to CIAT headquarters in Palmira (2 to 3 hours from the collection areas) where the stems were grown in greenhouses. After being cultivated for 6 to 8 weeks, tissue samples were taken to the Cassava Genetics Laboratory at CIAT for processing and DNA testing. The entire DNA sampling process took 3 months.16 14

The team collected cassava stem samples from 305 households but due to duplication, deterioration of the samples, and data cleaning, only 436 stem samples from 217 households could be used for data analysis. Many of these 217 households had multiple cassava plants and so the team collected multiple stem samples from each household for DNA fingerprinting. 15 Though it is difficult to extrapolate the results of the regression using DNA fingerprinting, we extended the econometric model for farmer-self identification of IVs to the full set of 305 observations. We found the results of the regression between the smaller set of 217 households and the larger dataset to be consistent. 16 Plant molecular analysis was done under the framework of new legislation for research institutes associated with Colombia’s Ministry of Environment and Sustainable Development (see MINIAMBIENTE, 2013) through which the Universidad Nacional de Colombia (Palmira campus) does not need permission for genetic resources assessment when the material is collected without commercial interest and for research purposes only. The samples were lyophilised overnight using an Alpha 2-4 LD plus Martin Christ Freeze-dryer (Germany). DNA was then extracted according to the CTAB-based DNA extraction protocol described by Doyle and Doyle (1990) with minor modifications. The disrupted tissues were incubated at 65°C for 1 hour followed by one organic extraction using chloroform-isoamyl alcohol (24:1); mixing gently but thoroughly for 30 minutes at 0°C. The DNA resulting from this extraction was assessed for quality in a 1% agarose gel and quantified using a Synergy HT Multi-detection microplate reader (BioTekÒ , USA). TM


8


For the DNA finger printing, we used 434 stem samples collected from the 217 households and genotyped these using the cassava SNPY-CHIP as described in Pe~ naVenegas et al. (2014). Duplicate accessions were determined by pairwise multilocus matching among all 434 samples, 70 CIAT high-land breeding lines, 85 GRU highland landraces, and 50 accessions randomly collected in Cauca based on morphological variation. Descriptive statistics were calculated for 93 SNPs. The key descriptive statistics included minor allele frequency, observed homozygosity, observed heterozygosity and expected heterozygosity. Computations were then carried out using the same NGSEP program (Perea et al., 2016). DNA samples that were fully matched at the genotyped SNP loci were declared the same cultivar or duplicated clones by calculating the number and percentage of homozygous and heterozygous differences between every pair of samples. Finally, a cluster analysis using the neighbour-joining (NJ) method was used to further examine the genetic relationship among accessions or group of accessions (see online Appendix B for the dendrogram). 6. A Model of the Determinants of Adoption of Modern Varieties Following various studies that have identified and discussed determinants of technology adoption in developing countries (Feder et al., 1985; Doss et al., 2003; Pattanayak et al., 2003), we assume a model considers the farmers’ decision of adoption as driven by the utility that they would gain from it. Let y0i* and y1i* be the benefits that the ith farmer would obtain when he/she does not adopt IVs and when he/she does, respectively. In order to estimate the relationship between adoption decision and the relevant variables discussed in the next section, binary regression is necessary to define a series of marginal effects on the probability of adoption. Given yi the model estimates the expected value: E½yi ¼ 1jxi ¼ pi ¼ Fðx0i bÞ:

ð4Þ

Under this framework the estimated parameters reveal the direction of the partial correlation; marginal effects may vary depending on the values taken by the vector xi usually analysed for mean values. Adoption is considered as a one trial binomial (or ‘Bernoulli’) experiment. Estimates are derived from maximum (log-) likelihood approximation because of the intrinsic non-linearity imposed by logit and probit models (Greene, 2012). Our dependent variables are binary with one (1) signifying self or geneticist identification of an improved variety, and zero (0) if not. A standard probit model was used to estimate the likelihood of adoption. Further, we incorporated fixed effects in the probit regression model by clustering municipalities based on their geographic proximity to one another within the department (North, Centre-West, Centre-East, South). Descriptive statistics, variables and results of the regressions are discussed in the next section. 7. Data Used and Characteristics of Adopters and Non-Adopters Based on descriptive statistics shown in Table 2 (see online Appendix A for the full set of statistics), we see a similar pattern of household characteristics for adopters of IVs, regardless of identification method. Utilising both farmer identification and Ó 2017 The Authors. Journal of Agricultural Economics published by John Wiley & Sons Ltd on behalf of Agricultural Economics Society.

Table 2

Self-identify: No Yes Geneticist: No Yes

24

1 24

1

179

37 196

20

15.00

32.40 28.06

25.70

0.55

0.72 1.19

1.22

65.00

54.05 48.47

49.16

25

24.32 3.6

1.7

14

30 149

133

70.00

81.08 76.02

74.30

2.89

2.12 1.06

1.03

4.69

4.34 2.97

2.88

Percentage of No. of Ave. total Ave. size Percentage of Percentage of households household heads Identification land size of cassava Percentage households households Ave. with at least a with a method (in plots of with access No. of with access No. of primary female on IV Total to extension landowners landowners (in hectares) hectares) to credit dependents education head adoption No.

Descriptive statistics of adopters and non-adopters, by identification method



9

10


DNA fingerprinting, we see that adopters of IVs have smaller households, better access to credit, and better access to extension services. Surprisingly, only 15% of adopters identified using DNA fingerprinting had a primary education compared with 32.40% for adopters based on self-identification. In terms of land ownership, we see that a majority of adopters and non-adopters are landowners, in line with the overall sample average of 75.46%. However, it is clear that adopters had larger cassava plots as well as more land, in general. The average self-identified adopter had about twice the acreage for cassava compared to nonadopters while geneticist-identified adopters had almost thrice the acreage compared to non-adopters. 8. Explanation of Variables We consider a comprehensive list of possible determinants, in line with related literature on similar determinant studies (Table 3). The variable ‘farmeridentify’ or farmer self-identification is based on the name farmers use for the variety. Often farmers would associate numbers or codes with IVs such as ‘ICA’ which stands for the Instituto Colombiano Agropecuario (Colombian Agricultural Institute). There are also some few well-known IVs used by farmers such as ‘Cumbre-3’. On the other hand, some common names for non-improved varieties include ‘Sata’, ‘Algodona’ and ‘Amarga’. If a household self-identified use of an improved variety, the variable takes the value of one (1) or if otherwise, zero (0). For ‘geneticist’ or varietal identification through DNA fingerprinting, if a household was found to be using an IV, the variable takes the value of one (1) or if otherwise, zero (0). 9. Results Based on the interviews with farmers, we found 117 unique names while the clustering patterns used in DNA fingerprinting allowed the identification of 60 main varietal types and 60 unique genotypes that were not represented in CIAT’s genebank. Of these 120 varietal types, CIAT geneticists only identified nine (9) to be improved cassava varieties while the remaining were landraces (see online Appendix B for cassava dendrogram). Some of the names for more popular landraces include ‘Algodona’, ‘Amarga’ and ‘Sata’. From these varieties, we examined farmers’ perceptions of IVs and found that there were differences in opinions between adopters and non-adopters of IVs. In particular, we compared the rankings of nine (9) characteristics for varieties used by adopters (IVs) to those varieties used by non-adopters. Table 4 shows our results. According to the rankings by farmers, we find three traits where IVs are considered better: tolerance to pests, size of roots, and root quality. We then checked which variety was most reported to have these desirable traits and found that farmers preferred “ICA-48”, an IV developed by the Colombian Institute for Agriculture. As shown, adopters perceive benefits in IVs, which would motivate adoption. To estimate the level of adoption of modern improved varieties in Cauca, we define it as the percentage of the total cassava acreage that was planted with modern varieties in 2015. Table 5 describes the level of adoption following farmers’ self-identification of varieties and following the results of the DNA fingerprinting.



11

Table 3 List of household determinant variables Variable name Dependent variables farmeridentify geneticist Household characteristics female_head other_employment

Definition Farmer self-identified use of improved variety Use of improved variety verified by DNA fingerprinting

Household head is female More than 2 household members (50% of mean household size) have income-generating non-agriculture jobs hh_education Education level of household head (primary school finisher or not) hh_age Age of household head dependent_proportion Percentage of household members not in the age range of 14–70 years old (working age) logexperience Natural log of the number of years farmers have planted cassava; also functions as a proxy control for age of household head Land characteristics and production landownership* Household owns land used for farming credit Household has access to credit logcassavasize Natural log of the plot area used for cassava cultivation logtotalsize Natural log of the total size of household’s farmland extension Household has received extension services in the past year Market factors direct_selling Farmers directly sell cassava at cassava starch/flour mills Control variables group_alt† Altitude zone of household farmland (low, mid, high) group_mun Municipal clusters (North, Centre-West, Centre-East, South) *

Some households cultivate more than one plot where the second or third plot may be rented. We consider households with ownership of at least one lot as landowners whereas if a household does not own a single plot (rents all farmed land or relies on communal/borrowed land) this variable takes the value of zero. † We include altitude, as some varieties are more suited to certain altitude zones. Further, Dyer et al. (2007) found that controlling for elevation shows that seed diffusion is higher for landraces than for improved varieties.

As shown in Table 5, we find that farmers overestimate their use of IVs, in line with Kosmowski et al. (2016), Maredia et al. (2016) and Hareau et al. (2016) on other root crops and tubers. On the household survey, 37 households (17.05%) believed that they were using improved cassava but after varietal identification by geneticists, only 20 households (9.22%) were found to be using IVs. Even accounting for a 3% margin of error for DNA fingerprinting, the difference between survey data and actual genetic data is still observable. Further, geneticists were able to confirm that only 10 of the 37 households that self-identified as using IVs were correct, which means that 27 households were mistaken in their self-identification (an error rate of 73%, with almost ¾ of farmers thinking they were using IVs when in fact they were not). In addition, geneticists found that another 10 households believed they were using landraces but


12

Victorino O. Floro IV et al. Table 4 Perceptions of IVs by farmers

Characteristic

p-value (significance of difference between IVs and non-IVs)†

1. Yield 2. Tolerance to diseases 3. Tolerance to pests

0.83 0.46 0.07*

4. Size of roots 5. Market acceptance 6. Market price 7. Root quality 8. Time to harvest 9. Taste

0.02** 0.17 0.87 0.05** 0.82 0.19

Preferred improved variety (based on frequency of reported name per trait)§ N/A‡ N/A‡ ICA 48, SM1495-5, CUMBRE-3 ICA 48, Sata N/A‡ N/A‡ ICA 48 N/A‡ N/A‡

Notes: ***P < 0.01, **P < 0.05, *P < 0.1. † These are Kruskal–Wallis results comparing farmer rankings of IVs and non-IVs. The null hypothesis is that there is no difference between variety rankings. ‡ If there is no difference between IVs and non-IVs, no preferred IV is listed. § Preferred varieties are based on the frequency of names of the varieties that ranked the highest for each characteristic. For example, ICA-48 had the most farmers reporting it as the variety with the highest tolerance to pests.

were in fact using IVs. This discrepancy confirms that there is difficulty on farmers’ parts to identify or recall the type of cassava they are growing. More importantly, the discrepancy in farmer knowledge carries through to a difference in the amount of land with IV cultivation. In terms of acreage, DNA fingerprinting found that only 12.6% of land was cultivated with IVs while farmer selfidentification pegged this figure at 24.4% or an overestimation of the number of hectares by 11.8%. This finding for DNA fingerprinting falls within the results of the cassava survey by Jaramillo (2008) that, depending on the altitude zone, between 10– 20% of land of Cauca was planted with IVs. Looking at the breakdown by municipality (see online Appendix A for descriptive statistics and Appendix C for the map of households), we note that in some municipalities, the number of households self-identifying the use/non-use of IVs closely

Table 5 Improved variety identification Farmer self-identification

Number of households Percentage of households Percentage of acreage

Identification though DNA fingerprinting

Improved

Landrace

Improved

Landrace

37 17.05 24.39%

180 82.95% 75.61%

20 9.22% 12.63%

197 90.78% 87.37%



13

matches the results of the DNA fingerprinting by CIAT geneticists. These municipalities include Caldono, Timbıo, Rosas, Caloto, El Bordo and Morales. The municipality of Morales itself had the highest number of self-identified adopters of IV (14 households) as well as the highest number of confirmed adopters by DNA fingerprinting (13 households).17 This close match in between the two results is unique in Morales possibly due to the municipal government’s maintenance of an active agricultural extension programme that has disseminated improved agricultural technologies and cultivars to resident farmers. On the other hand, the municipality of Santander de Quilichao had the highest number of farmers mistakenly identifying the use of IVs. 35.7% of farmers in Santander de Quilichao believed they were using IVs when, in fact, none of the 14 households surveyed were using improved cultivars. Table 6 presents the results of two probit-model specifications, using both self-identification and identification by DNA fingerprinting to identify the determinants of the adoption of modern cassava varieties. In the first specification the dependent variable defines an IV using farmers’ selfidentification of the cassava varieties while in the second specification, the dependent variable defines an IV using the identification derived from DNA fingerprinting. For each dependent variable, we present the marginal effects of two models. The first model on the first and third columns (1) and (3) factors in only household characteristics while the second model on the second and fourth columns (2) and (4) shows the full set of variables. In order to ensure that the findings below are not localised, all models control for two fixed effects: altitude and clusters of municipalities.18 Our results indicate that very few factors influence farmers’ decision to adopt improved cassava varieties. Using the farmer self-identification models, we find no significant variable when only including household characteristics but three significant using the full model: proportion of dependents (5%), total land size (5%), and access to extension services (1%). Of these, a key factor is a household’s access to agricultural extension services. As improved cassava varieties have been developed and disbursed for over 30 years in Colombia, extension is clearly an influence on whether or not farmers choose to adopt IVs and their ability to recognise them. Compared with the farmer’s self-identification of IVs, our estimates using DNA fingerprinting show weak results. Unlike the full model using farmer self-identification, access to extension services shows no significance. This finding is key as it shows that despite decades of agriculture extension activities in Cauca, actual IV diffusion remains largely unaffected. Interestingly, the results for DNA fingerprinting show some significance for both land ownership and land area, with the former showing a negative effect on the probability of adopting IVs. Nine percent (9%) of landowners adopted IVs while 13.2% of renters did so.19 There are many possible explanations for this. Based on CIAT’s experience working with farmers, farmers who rent land are more likely to adopt IVs, with higher yields,

17

Seven (7) out of the 14 households that self-identified as adopters of IVs were confirmed to be using IVs by DNA fingerprinting. Five (5) out of the 13 confirmed adopters were not aware/did not self-identify as adopters. 18 Municipalities geographically located near each other have similar eco-agricultural characteristics. Each cluster also had a similar number of households. 19 Farmers that both owned and rented land, as well as only renters, held more land dedicated to planting cassava compared to farmers who only owned land. Ó 2017 The Authors. Journal of Agricultural Economics published by John Wiley & Sons Ltd on behalf of Agricultural Economics Society.

14

Victorino O. Floro IV et al. Table 6

Marginal effects of probit model on determinants of adoption for farmer self-identification and DNA fingerprinting Farmer self identification Variables Female head of household Other non-farm employment Education level of household head Age of household head Access to credit Proportion of dependents Log of years of farm experience

(1) –0.16 (0.10) 0.01 (0.09) –0.02 (0.06) 0.00 (0.00) –0.02 (0.05) (0.14) (0.11) –0.01 (0.04)

Land ownership Log size of cassava plots Log of total land area Size of cassava plots/Total land size Access to extension services Selling to Starchmills

(2)

(3)

0.05 (0.08) –0.05 (0.06) 0.00 (0.00) –0.06 (0.06) 0.20** (0.10) 0.01 (0.03) 0.00 (0.06) –0.01 (0.03) 0.08** (0.03) 0.08 (0.05) 0.28*** (0.09) –0.07 (0.05)

–0.08 (0.05) –0.00 (0.00) 0.06 (0.05) –0.21* (0.12) 0.01 (0.03)

– 0.0917 0.1251 176 Yes Yes

(4)

–0.08 (0.09)

(Omitted) Other non-farm employment Prob. > Chi-squared Pseudo R-squared Observations Regional fixed effects Altitude fixed effects

DNA fingerprinting

0.0136 0.2358 166 Yes Yes

0.0001 0.3261 163 Yes Yes

–0.06 (0.05) –0.00 (0.00) –0.02 (0.05) –0.14 (0.12) 0.02 (0.03) –0.09* (0.05) –0.03 (0.03) 0.08* (0.04) 0.08 (0.11) 0.10 (0.08) 0.08 (0.05) – 0.0005 0.4016 153 Yes Yes

Notes: Standard errors in parentheses; The difference in the number of observations between each column is due to the model dropping observations due to missing data points but we have conducted checks to ensure model accuracy. ***P < 0.01, **P < 0.05, *P < 0.1.

because of the need to pay back the rent on the land and other inputs.20 On another note, cassava tends to be a second choice to other crops for farmers in many areas. If growers have better land or have access to inputs, they will often plant higher-value 20

A report by the Massachusetts Institute of Technology Jameel Poverty Action Lab (J-PAL) states that ‘relative to output sharing, land rental markets should create strong investment incentives by making the renter the full beneficiary of increases in productivity. . . Consequently, rental contracts generate the greatest incentive to invest’ (Jack, 2013).



15

crops (Hershey et al., 2016) though we have found some farmers who prefer cassava to cash crops like coffee due to labour constraints. Nonetheless, owning the land may cause farmers to invest more in inputs or in planting other, higher value crops. Land area is marginally significant, suggesting that farmers with more land were more likely to cultivate cassava. To better understand the role of land ownership and land area, we looked at which farmers were wealthier and who owned more land. In terms of wealth,21 there is no significant difference between renters and non-renters. However, we found that renters cultivated multiple tracts of land (i.e. rented and owned). As landowners, renters’ properties are smaller than the property of non-renters but combined with their rented properties, renters had more total land. Interestingly, we found no significance in the difference in cassava plots on properties owned by renters and non-renters but renters had more land overall cultivated with cassava. It is apparent that there are clear differences in the identification of adoption factors between farmer self-identification and DNA fingerprinting, albeit the results only show marginal significance. Nonetheless, the ramifications are significant. Most importantly, the attribution of significance to extension in farmer self-identification seems to be misguided. As shown in Table 5, farmers did not clearly know whether or not they were planting IVs. These findings imply that misidentification of IVs in other studies may have led to errors in determining the factors of adoption. 10. Implications and Conclusion The use of DNA fingerprinting in econometric analyses of household determinants of adoption of IVs is still nascent. As most studies implicitly assume the accuracy of expert or farmer self-identification of varieties, there is little literature discussing the possibility of misidentifying IVs. In line with previous studies incorporating DNA data on root and tuber crops, our results show that farmers overestimate their use of improved cassava varieties, suggesting that the use of DNA fingerprinting corrects for overestimation in adoption studies. Importantly, our findings show that using farmer self-identification in regression analysis can falsely attribute significance to determinants of adoption that DNA fingerprinting may not otherwise find significant. Based on the discrepancy between farmer self-identification and DNA fingerprinting, we conclude that though knowledge of names of IVs may spread across a region, actual diffusion of an IV through the effects of extension services remains localised. This implies that better extension services are needed not only to properly educate farmers on the types of cassava they are growing but also to better disperse IVs. Moreover, farmer misidentification of the varieties they are cultivating may lead farmers to use on-farm practices that are not well suited for certain IVs or landraces. In general, farmers need to know their varieties correctly if they are to apply the correct amount of fertiliser, or to plant varieties in the best soil type to maximise yields. From an econometric perspective, the incorporation of DNA fingerprinting removes the consequences of misclassifying the dependent variable (i.e. the adoption decision). This being said, it is vital that future adoption studies consider the use of DNA fingerprinting in order to ensure that the determinants of adoption are correctly

21

The survey did not include a discrete measure of income and to measure economic status we used the Progress Out of Poverty Index as well as a constructed measure based on reported assets.


16


identified. Moving forward, the trend of reducing costs in DNA analysis especially for large samples should increase the feasibility of including DNA fingerprinting in more adoption studies. Additionally, the accurate identification of crop varieties also provides a stepping-stone for impact assessment studies to verify the effectiveness of breeding programmes on varietal improvement. From an agronomic perspective, the identification of 60 unique varieties in the Cauca region, despite CIAT’s extensive gene bank of cassava germplasm, demonstrates the genetic diversity of the crop in Latin America. This genetic diversity also helps explain the difficulty farmers have in distinguishing between IVs and landraces. Due to the genetic diversity of cassava, we also emphasise that DNA fingerprinting is imperative in proper varietal identification. Supporting Information Additional Supporting Information may be found in the online version of this article: Appendix A Descriptive Statistics. Table A1. Total sample descriptive statistics. Table A2. Household characteristics by municipality. Table A3. Land characteristics by municipality. Figure A1. Altitude variations by municipality. Figure A2. Percentage of farmers selling at Rallenderıas by municipality. Appendix B Dendrogram of identified cassava varieties in Cauca. Appendix C Map of households studied in select municipalities of the Cauca department, Colombia. References Agwu, A. E. and Anyaeche, C. L. ‘Adoption of improved cassava varieties in six rural communities in Anambra State, Nigeria’, African Journal of Biotechnology, Vol. 6, (2007) pp. 89–98. Alves, A. ‘Cassava Botany and Physiology’, in: R. J. Hillocks and J. M. Thresh (eds,), Cassava: Biology, Production and Utilization (Wallingford, UK: CABI Publishing, 2002). Awotide, B. A., Alene, A. D., Abdoulaye, T. and Manyong, V. M. ‘Impact of agricultural technology adoption on asset ownership: the case of improved cassava varieties in Nigeria’, Food Security, Vol. 7, (2015) pp. 1239–1258. ~ez, J., Lijavetzky, D., Velez, D., Bravo, G., Rodrıguez, V., Carre~ Cabezas, J. A., Iban no, I., Jermakow, A. M., Carre~ no, J., Ruiz-Garcıa, L. and Thomas, M. R. ‘A 48 SNP set for grapevine cultivar identification’, BMC Plant Biology, Vol. 11, (2011) pp. 1–12. Corporaci on Colombiana de Investigaci on Agropecuaria & Centro Internacional de Agricultura Tropical (CIAT). Liberaci on de nuevas variedades de yuca para las zonas de altura del Departamento Cauca. (Palmira, Colombia: CIAT, 2014). Corporaci on Regional Del Cauca Rallandero Limpio: Cartilla educativa (Popay an, Colombia: Corporaci on Regional Del Cauca, 2005). Departmento Administrativo Nacional de Estadıstica (DANE). Censo Nacional Agropecuario Novena entrega de Resultados, 2015. Available at: http://www.dane.gov.co/index.php/Ce nso-Nacional-Agropecuario-2014 (last accessed 10 August 2016). Doss, C. R., Mwangi, W. M., Verkuijl, H. and De Groote, H. Adoption of Maize and Wheat Technologies in Eastern Africa: A Synthesis of the Findings of 22 Case Studies, CIMMYT



17

Working Paper No. 2003-01, 2003. Available at: http://ageconsearch.umn.edu/bitstream/ 46522/2/wp030001.pdf (last accessed 30 July 2016). Doyle, J. and Doyle, J. ‘A rapid total DNA preparation procedure for fresh plant tissue’, Focus, Vol. 12, (1990) pp. 13–15. Dyer, G. A., Gonzalez, C. and Lopera, D. C. ‘‘Informal “seed” systems and the management of gene flow in traditional agroecosystems: The case of cassava in Cauca’, Colombia’, PLoS One, Vol. 6, (2007) p. e29067. Feder, G., Just, R. E. and Zilberman, D. ‘Adoption of agricultural innovations in developing countries: A survey’, Economic Development and Cultural Change, Vol. 33, (1985) pp. 255– 298. Food and Agriculture Organization. Statistics (Rome: FAO, 2014). Available at: http://faosta t3.fao.org/ (last accessed 30 August 2016). Fukuda, W. M. G., Guevara, C. L., Kawuki, R. and Ferguson, M. E. Selected Morphological and Agronomic Descriptors for the Characterization of Cassava (Ibadan, Nigeria: International Institute of Tropical Agriculture, 2010). Ge, A. J., Han, J., Li, X. D., Zhao, M. Z., Liu, H., Dong, Q. H. and Fang, J. G. ‘Characterization of SNPs in strawberry cultivars in China’, Genetics and Molecular Research, Vol. 12, (2013) pp. 639–645. Greene, W. H. Econometric Analysis, 7th ed (Boston, MA: Pearson, 2012). Hareau, G., Pradel, W., Xie, K., Qin, J., Forbes, G., Ellis G., Barkley, N., Alwang, J., Norton, G., Larochelle, C., Myrick, S. and Li, C. H. Adoption and Diffusion of Potato Variety Cooperation 88 (C88) in China. Presentation to the CGIAR Standing Panel of Impact Assessment (SPIA) (Boston, MA: CGIAR, 2016). Hausman, J. A. ‘Mismeasured variables in econometric analysis: Problems from the right and problems from the left’, The Journal of Economic Perspectives, Vol. 15, (2001) pp. 57–67. Hausman, J. A., Abrevaya, J. and Scott-Morton, F. M. ‘Misclassification of the dependent variable in a discrete-response setting’, Journal of Econometrics, Vol. 87, (1998) pp. 239–269. Henry, G. and Hershey, C. ‘Cassava in South America and the Caribbean’, in: R. J. Hillocks and J. M. Thresh (eds.), Cassava: Biology, Production and Utilization (Wallingford, UK: CABI Publishing, 2002). Hershey, C., G Henry, R. Bes and Iglesias, C. Cassava in Latin America and the Caribbean: Resources for Global Development. Background Document (Cali, Colombia: Cassava Biotechnology Network Regional Planning Meeting, 2000). Hershey, C., Alvarez, E., Aye, T., Becerra, L., Belotti, A., Ceballos, H., Fahrney, K., Howeler, R., Lefroy, R., Ospina, B. and Parsa, S. Eco-Efficient Interventions to Support Cassava’s Multiple Roles in Improving the Lives of Smallholders (Palmira, Colombia: CIAT, 2016). Howeler, R. L., Lutaladio, N. and Thomas, G. Save and Grow: Cassava. A Guide to Sustainable Production Intensification: Produire Plus Avec Moins. Ahorrar Para Crecer, Working Paper No. FAO 633.6828 S266 (Rome, Italy: Food and Agriculture Organization, 2013). Hurtado, P., Olsen, K. M., Buitrago, C., Ospina, C., Marin, J. A., Duque, M. C., de Vicente, M. C., Wongtiem, P., Wenzl, P., Killian, A., Adekele, M. and Fregene, M. A. ‘Comparison of simple sequence repeat (SSR) and diversity array technology (DArT) markers for assessing genetic diversity in cassava (Manihot esculenta Crantz)’, Plant Genetic Resources: Characterization and Utilization, Vol. 6, (2008) pp. 208–214. Jack, K. Market Inefficiencies and the Adoption of Agricultural Technologies in Developing Countries, Literature Review (Cambridge, MA: J-PAL MIT, 2003). Jaramillo, G. Diagnostico del cultivo de la yuca y su agroindustria en el departamento del Cauca (Palmira, Colombia: CIAT, 2008). Ji, K., Zhang, D., Motilal, L. A., Boccara, M., Lachenaud, P. and Meinhardt, L. W. ‘Genetic diversity and parentage in farmer varieties of cacao (Theobroma cacao L.) from Honduras and Nicaragua as revealed by single nucleotide polymorphism (SNP) markers’, Genetic Resources and Crop Evolution, Vol. 60, (2013) pp. 441–453.


18


Kosmowski, F., Aragaw, A., Kilian, A., Ambel, A., Ilukor, J., Yigezu, B. and Stevenson, J. Varietal Identification in Household Surveys: Results from an Experiment Using DNA Fingerprinting of Sweet Potato Leaves in Southern Ethiopia, Policy Research Working Paper No. 7812 (Washington, DC: The World Bank, 2016). Labarta, R., Martinez, J., Lopera, D., Gonzalez, C., Quintero, C., Gallego, G., Viruez, J. and Taboada, R. Assessing Impacts of the Adoption of Modern Rice Varieties using DNA Fingerprinting to Identify Varieties in Farmer Fields: A Case Study in Bolivia (Palmira, Colombia: CIAT, 2015). Larochelle, C., Alwang, J., Norton, G. W., Katungi, E. and Labarta, R. A. ‘16 impacts of improved bean varieties on poverty and food security in Uganda and Rwanda’, in: T. Walker (ed.), Crop Improvement, Adoption and Impact of Improved Varieties in Food Crops in SubSaharan Africa (Wallingford, UK: CABI, 2015). Maredia, M. K., Reyes, B. A., Manu-Aduening, J., Dankyi, A., Hamazakaza, P., Muimui, K., Rabbi, I., Kulakow, P., Parkes, E., Abdoulaye, T., Katungi, E. and Raatz, B. Testing Alternative Methods of Varietal Identification Using DNA Fingerprinting: Results of Pilot Studies in Ghana and Zambia. International Development Working Paper No. 149 (East Lansing. Michigan: Michigan State University, 2016). Mba, R. E. C., Stephenson, P., Edwards, K., Melzer, S., Nkumbira, J., Gullberg, U., Apel, K., Gale, M., Tohme, J. and Fregene, M. ‘Simple sequence repeat (SSR) markers survey of the cassava (Manihot esculenta Crantz) genome: Towards an SSR-based molecular genetic map of cassava’, Theoretical Applied Genetics, Vol. 102, (2001) pp. 21–31. Meyer, B. and Mittag, N. Misclassification in Binary Choice Models, Working Paper No. 20509 (Cambridge, MA, USA: National Bureau of Economic Research, 2014). Minambiente, M.d.A.y.D.S. Decreto por el cual se reglamenta el permiso de recoleccioń de especı´menes de especies silvestres de la diversidad bilolo´gica con fines de investigacioń cientı´fica no comercial. (Bogota, Colombia: MINAMBIENTE, 2013). Muhammad-Lawal, A., Salau, S. A. and Ajayi, S. A. ‘Economics of improved and local varieties of cassava among farmers in Oyo State, Nigeria’, Ethiopian Journal of Environmental Studies and Management, Vol. 5, (2012) pp. 189–194. M€ uhlen, G. S., Martins, P. S. and Ando, A. ‘Variabilidade genetica de etnovariedades de mandioca, avaliada por marcadores de DNA’, Scientia Agricola, Vol. 57, (2000) pp. 319–328. Pattanayak, S. K., Mercer, D. E., Sills, E. and Yang, J. C. ‘Taking stock of agroforestry adoption studies’, Agroforestry Systems, Vol. 57, (2003) pp. 173–186. Pe~ na-Venegas, C., Stomph, T., Verschoor, G., Becerra Lopez-Lavalle, L. A. and Struik, P. ‘Differences in manioc diversity among five ethnic groups of the Colombian Amazon’, Diversity, Vol. 6, (2014) pp. 792–826. Perea, C., De La Hoz, J. F., Cruz, D. F., Lobaton, J. D., Izquierdo, P., Quintero, J. C., Raatz, B. and Duitama, J. ‘Bioinformatic analysis of genotype by sequencing (GBS) data with NGSEP’, BMC Genomics, Vol. 17, (2016) p. 498. Peroni, N., Kageyama, P. and Begossi, A. ‘Molecular differentiation, diversity, and folk classification of “sweet” and “bitter” cassava (Manihot esculenta) in Caicßara and Caboclo management systems (Brazil)’, Genetic Resources and Crop Evolution, Vol. 54, (2007) pp. 1333–1349. Pujol, B., M€ uhlen, G., Garwood, N., Horoszowski, Y., Douzery, E. J. and McKey, D. ‘Evolution under domestication: Contrasting functional morphology of seedlings in domesticated cassava and its closest wild relatives’, New Phytologist, Vol. 166, (2005) pp. 305–318. Rabbi, I. Y., Kulakow, P. A., Manu-Aduening, J. A., Dankyi, A. A., Asibuo, J. Y., Parkes, E. Y., Abdoulaye, T., Girma, G., Gedil, M. A., Ramu, P., Reyes, B. and Maredia, M. K. ‘Tracking crop varieties using genotyping-by-sequencing markers: A case study using cassava (Manihot esculenta Crantz)’, BMC Genetics, Vol. 16, (2015) p. 115. Torres, P., Perez, A., Cajigas, A., Juardo, C. and Ortiz, N. ‘Selecci on de in oculos para el tratamiento anaerobio de aguas residuales del proceso de extracci on de almid on de yuca’, Ingenierıa de Recursos Naturales y del Ambiente’, Vol. 6, (2007) pp. 105–111.



19

Walker, T. S. Validating Adoption Estimates Generated by Expert Opinion and Assessing the Reliability of Adoption Estimates with Different Methods (Boston, MA: CABI, 2015). Walker, T. S. and Alwang, J. (eds.) Crop Improvement, Adoption and Impact of Improved Varieties in Food Crops in Sub-Saharan Africa (Boston, MA: CGIAR and CABI, 2015). Wu, G. A., Prochnik, S., Jenkins, J., Salse, J., Hellsten, U., Murat, F., Perrier, X., Ruiz, M., Scalabrin, S., Terol, J., Takita, M. A., Labadie, K., Poulain, J., Couloux, A., Jabbari, K., Cattonaro, F., Del Fabro, C., Pinosio, S., Zuccolo, A., Chapman, J., Grimwood, J., Tadeo, F., Estornell, L., Mu~ noz-Sanz, J. and Ibanez, V. ‘Sequencing of diverse mandarin, pummelo and orange genomes reveals complex history of admixture during citrus domestication’, Nature Biotechnology, Vol. 32, (2014) pp. 656–662. Yirga, C., Traxler, G., Kim, M. and Alemu, D. Using DNA Fingerprinting to Estimate the Bias of Farm Survey Identification of the Diffusion of Improved Crop Varieties in Ethiopia. Conference Paper 9–14 August (Milan, Italy: International Conference of Agricultural Economists, 2015).


Household Determinants of the Adoption of Improved Cassava

Recommend Documents