Variety selection and diversification are climate change adaptation practices pursued by Colombian common bean producers. We investigate the drivers behind common bean variety selection and diversification in one of the most important common bean production regions in Colombia —Santander. The effects of climate change on this region are expected to be elevation driven. Exploiting the relationship between elevation-driven weather variations and climate change perception in Santander, we estimate an alternative-specific conditional logistic regression model to identify the determinants of common bean variety selection from a survey of producers. Using an ordered-logistic regression model, we also investigate the drivers behind common bean variety diversification within this farming community. We find that farms’ elevation, household composition, and seed certification are some of the most important drivers behind farmers’ common bean variety selection in Santander. We also find that varieties that sell at higher prices and have shorter vegetative cycles tend to be more preferred by farmers. Finally, farmers who receive more help from family members and own a tractor tend to grow more than one variety in the same production cycle. Common bean breeding programmes can exploit these drivers to design communication strategies to maximize uptake of newly developed common bean phenotypes.
1. Introduction
Colombian weather patterns are elevation-driven and strongly influenced by El Nino Oscillation (ENSO) (Cepal, 2012; IPCC, 2014; Buhr et al., 2018). El Niño phase tends to generate higher daily temperatures, less daily precipitation, and more recurrence of droughts, especially in low-elevation geographical areas. In contrast, la Niña phase tends to generate lower daily temperatures, more daily precipitation, and more instances of flood and extreme temperature variations, especially in high-altitude geographical areas (Poveda and Mesa, 1996; Poveda et al., 2011; Henao et al., 2020). This weather instability generates an asymmetric impact on agricultural productivity. While el Niño phase increases the prevalence of abiotic stresses in crops, la Niña phase increases the prevalence of biotic stresses and the risk of crop destruction due to floods and landslides (Santos, 2006; Duque et al., 2013; IDEAM, 2013). These patterns are expected to get worsened by climate change, negatively impacting 60% of the current Colombian agricultural production areas and 80% of the crops that Colombian farmers currently cultivate (Feola and Binder, 2010; Ramirez-Villegas et al., 2012; Eitzinger et al., 2014; IPCC, 2014).
Climate change is expected to particularly affect the production of common beans (Phaseolus vulgaris) in Colombia. Some of the traditional common bean varieties cultivated by Colombian farmers, such as Calima, are highly vulnerable to extreme temperatures and reduced levels of rainfall (Schoonhoven and Voysest, 1991; CIAT, 2008; Troyo-Diéguez et al., 2010). Consequently, a worsened ENSO is expected to reduce the profitability of the cultivation of traditional common bean varieties (CIAT, 2008). In addition, low-elevation geographical areas are expected to become less suitable for the cultivation of traditional varieties of common beans because they will have an increased prevalence of biotic and abiotic stresses with a worsened ENSO (Ramirez-Villegas et al., 2012; Eitzinger et al., 2014; Güiza-Villa et al., 2020). Finally, producers located at higher elevations have a limited capacity to adapt to climate change since a hilly topography limits the use of heavy machinery or bulky technology to cultivate common beans (Ramirez-Villegas et al., 2012; Feola et al., 2015; Acevedo and Martinez, 2016).
There are several adaptation strategies that Colombian common bean growers can pursue at farm-level (Smit and Skinner, 2002; Clements et al., 2011; Asfaw et al., 2013a, Asfaw et al., 2013b; Niles et al., 2015; Islam et al., 2020). One of the most common adaptation strategies proposed for this farming community is the use of improved seeds (Hailu et al., 2015). There are several companies supporting common bean breeding programs in Colombia (Schoonhoven and Voysest, 1991; Blair, 2003; Muñoz et al., 2008; CIAT, 2008; Hershey and Neate, 2013). This work has mainly focused on making common beans more resistant to the most acute and prevalent biotic stresses present in Colombia (Leon and Jimenez, 1997; Leon and Jimenez, 2002; FENALCE, 2011; Beebe et al., 2011). However, the uptake of these new varieties has been low and Colombian farmers continue to grow traditional varieties, which are mostly exchanged in informal or non-market settings (Sperling and McGuire, 2010; FENALCE, 2020).
The development of common bean varieties in Colombia has been mainly based on expert opinion about the needs of farmers, partially disregarding the determinants of farmers’ demand for particular common bean attributes (Chauhan et al., 2020; Jochua et al., 2020; Ribeiro et al., 2020). Some studies in Africa have shown that demand depends on agronomic and economic attributes and farmers’ socioeconomic characteristics (Katungi et al., 2011; Katungi et al., 2015; Sichilima et al., 2016). To our knowledge, no research has been performed on identifying the drivers behind farmers’ demand for the attributes of common beans in a Latin American context. According to the international evidence, farmers’ demand for phenotypes under development by plant breeding institutions depends on factors that also help determine the demand for current commercial phenotypes (Sichilima et al., 2016; Shikuku et al., 2017; Vanegas, 2017; Onzima et al., 2019). These factors are also expected to influence farmers’ response to extension and commercial programs that promote the voluntary uptake of the new varieties. Consequently, identifying the factors that determine farmers’ demand for common bean varieties allows agricultural companies, extension service providers, and seed suppliers to create commercial and communication strategies aimed at maximizing the uptake of the new varieties under development (CIAT, 2008; Sichilima et al., 2016; Eitzinger et al., 2018). This is particularly relevant to Colombia where farmers’ demand for common beans is expected to be influenced by the elevation in which farms are located, which implies that seed suppliers and companies in charge of common bean breeding programs should consider elevation as an important factor to develop new varieties and design commercial and engagement strategies.
Consequently, the aim of this paper is to provide the first identification of the factors that determine Colombian farmers’ demand for common bean varieties. By analysing the responses to a revealed-preference survey of 566 common bean producers in the department of Santander, this paper performs the first econometric estimation of the determinants of the demand for common bean varieties in Colombia. These determinants are identified employing an alternative-specific conditional logistic regression model. In addition, this paper also performs the first identification of the factors that determine variety diversification in Colombia using an ordered logistic regression model.
The department of Santander has been selected for this study because it is the fifth most important common bean producing region in Colombia (DANE, 2014) and it is expected to be the worst affected by climate change among the most important common bean production regions in Colombia (Ramirez-Villegas et al., 2012; Eitzinger et al., 2014; Eitzinger et al., 2018). Moreover, Santander’s municipalities are mainly located on the Andean mountains, which implies that Santander has the archetypal mountainous landscape of the Colombian Andes and any inference based on this region is easily extendable to other Colombian regions with similar agroecological environments and elevation-driven weather variations (Perez et al., 2019; Botero et al., 2020).
2. Conceptual framework
Valuation methods applied to the stated-preference exercises estimate farmers’ willingness to pay for common beans’ attributes and the magnitudes of the trade-offs that farmers are willing to accept to exchange one attribute for another (Katungi et al., 2011a; Katungi et al., 2015). Stated-preference experiments have been employed to measure farmers’ preferences for agricultural products’ and seeds’ attributes in different parts of the world (Drucker and Anderson, 2004; Asrat et al., 2010; Mahadevan and Asafu-Adjaye, 2015; Sánchez-Toledano et al., 2017; Acheampong et al., 2018; Jin et al., 2020). This methodology has been used extensively in Africa to determine consumers’ and farmers’ demand for the attributes of common beans (Lambrecht et al., 2013a, Lambrecht et al., 2013b; Lambrecht et al., 2015; Assete et al., 2018). One important advantage of stated-preference experiments is that the experimenter may manipulate the attributes offered to consumers and farmers to study their willingness to pay for each attribute based on their variety selection. Its most important disadvantage in agricultural settings is that experiments are usually applied to varieties that are not in the market yet, which impedes the experimenter to utilise the market value of the options offered. Consequently, these studies usually rely on hypothetical economic values and rewards to elicit behaviour, which may have important consequences on the consistency of the answers (Kuhberger et al., 2002; Locey et al., 2011; Luchini and Watson, 2014).
Revealed-preference or market methods are used as an alternative approach to stated-preference experiments (Louviere et al., 2000). In contrast to stated-preference experiments, revealed-preference methods do not rely on controlled experiments to obtain information on variety selection but on actual input choices. This methodology can be used to estimate the determinants of variety selection based on the actual seed choices made by commercial farmers. It utilises market information on variety choices to draw conclusions on the factors that determine the common bean variety selection observed in the market. Apart from differing in the source of information employed, revealed- and stated-preference methodologies utilise identical estimation methods and their estimated parameters have similar interpretations. This implies that both methodologies are able to determine a ranking of preference for common bean varieties, with the only difference that one relies on hypothetical selections and the other on market ones.
The main drawback of the revealed-preference method is that market choices do not include the whole universe of potential choices available to farmers, whereas in stated-preference experiments all existing varieties may potentially be included. Consequently, conclusions resulting from a revealed-preference estimation only apply to the varieties actually selected in the market, whereas the conclusions drawn from stated-preference experiments apply to the whole sample of varieties presented to farmers during the experiment. The main advantage of the revealed-preference method is that information on farmers’ socioeconomic characteristics is more reliable since this information is collected through face-to-face interviews with commercial farmers, which usually takes place in the farms where common beans are grown. In contrast, farmers’ socioeconomic information collected through stated-preference experiments depends on farmers’ willingness to participate in the experiments, which in turn depends on transportation costs and farms’ distance to the study site where experiments usually take place, generating a sample selection bias that may potentially affect the generalization of the results drawn from these experiments. As a result, the revealed-preference methodology tends to do a better job in identifying farmer-specific determinants of variety selection than the stated-preference one, but a poorer job in identifying variety-specific determinants (Louviere et al., 2000; Katungi et al., 2011a; Katungi et al., 2015).
In this study, we employ a revealed-preference approach to identify the factors that determine variety selection of common beans in Colombia. We further employ an ordered logistic regression model to investigate the drivers behind seed diversification. Following to Katungi et al. (2015), we use a combination of variety-specific and farmer-specific characteristics as the determinants of variety selection. Following to Onzima et al. (2019), we use a set of farm-specific factors as the explanatory variables of seed diversification. Two important determinants of variety selection introduced in this study are farms’ elevation and distance to the nearest urban centre. According to Feola et al. (2015), future weather variations in Colombia are expected to be elevation-driven, which will have a differentiated effect on common bean production regions. Ramirez-Villegas et al. (2012) estimate that high-elevation farms will experience more extreme temperature variations and unpredictable seasons and low-elevation farms will experience more droughts and lower rainfall levels. Botero et al. (2020) find that elevation is an important driver of climate change perception in this region of Colombia. Farmers located at low elevations tend to perceive more droughts and water deficits, whereas farmers located at high elevations tend to perceive extreme temperature variations, even though they consider that they have enough water for their bean production. As a result, farms’ elevation is expected to be an important driver behind farmers’ variety selection. In turn, farms’ distance to the nearest urban centre is expected to be an important driver behind variety selection because distance determines farms’ accessibility in the Andean mountains (Feola et al., 2015; Botero et al., 2020), affecting farmers’ transportation costs. In turn, these two variables are also expected to be important determinants of seed diversification. Elevation is expected to reduce seed diversification because it is more complicated to cultivate several crops in a steep land field. Distance is also expected to reduce seed diversification because more varieties grown imply a more spatially scattered demand, increasing transportation costs (Feola et al., 2015; Eitzinger et al., 2018). The whole set of regressors utilised in this study is introduced and explained in detail in the next section.
3. Data and descriptive analysis
3.1. The study site
According to IPCC (2014), the northeast of Colombia will be the most affected region with climate change. Santander is selected for this study because this region is expected to be the most affected common bean production area in Colombia and it is one of the most affected areas by ENSO (Ramirez-Villegas et al., 2012; Eitzinger et al., 2014; IPCC, 2014). Most municipalities in Santander have a hilly topography because they are located along the eastern side of the Colombian Andes. In Colombia, temperature and rainfall variations are elevation-driven. Low-elevation farms tend to have higher temperatures and lower rainfall levels and high-elevation farms tend to have lower temperatures and higher rainfall levels. This elevation-driven weather variation also affects the types of products grown in each thermal floor1 and the varieties grown of the same crop (IDEAM, 2013; Eitzinger et al., 2018; FENALCE, 2020).
Four municipalities are selected for this study: Barichara (6.6358° N, 73.2234° W), Curití (6.6063° N, 73.0687° W), San Gil (6.5548° N, 73.1341° W), and Villanueva (6.6709° N, 73.1748° W). Two criteria were used to select the study area. First, these municipalities are among the most important common bean production areas in Santander according to the 2014 Colombian Agricultural National Census. This allows having in the sample farmers with extensive knowledge on common bean production and on adaptation strategies to tackle the negative effects of ENSO on common bean production. Second, these municipalities have different elevations, which results in a different temperature and rainfall level depending on the elevation in which farms are located. Barichara (with an average elevation of 1266 masl) and Villanueva (1288 masl) tend to have higher temperatures and less rainfall than Curití (1568 masl) and San Gil (1452 masl). Consequently, common bean variety selection is expected to depend on each municipality’s elevation. Fig. 1 in appendix shows a map of the study site.