Adoption of modern varieties, farmers' welfare and crop biodiversity: Evidence from Uganda

Ecological Economics 119 (2015) 346–358

Contents lists available at ScienceDirect

Ecological Economics journal homepage: www.elsevier.com/locate/ecolecon

Analysis

Adoption of modern varieties, farmers' welfare and crop biodiversity: Evidence from Uganda Manuela Coromaldi a, Giacomo Pallante b,c,⁎, Sara Savastano b a b c

University of Rome Niccolò Cusano, Italy University of Rome Tor Vergata, Italy Sustainability Environmental Economics and Dynamics Studies (SEEDS), Italy

a r t i c l e

i n f o

Article history: Received 1 March 2015 Received in revised form 30 July 2015 Accepted 13 September 2015 Available online 6 October 2015 Keywords: Crop biodiversity Rural welfare Modern varieties Agricultural intensification Uganda Treatment effects

a b s t r a c t This paper assesses the impact of modern varieties adoption on farmers' welfare and crop biodiversity conserved in-situ. Using nationally representative data collected in 2009/2010 in Uganda, an endogenous switching regression model estimates the net economic and environmental effects of switching from local landraces to modern species. Results show that, after controlling for market and agro-ecological factors, the local varieties perform better than modern ones in marginalized and climatic vulnerable areas. Crop biodiversity shows to play a fundamental role in farmers' risk minimizing strategies when the available modern varieties are not adaptable to the local context and not supported by the required level of agro-intensification. Rural development policies should consider the heterogeneity in the adoption returns and support diversity conservation as a national strategic asset for a suitable bioprospecting and a best-fitting agricultural system implementation. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Evidence from the agricultural and development economics literature has proven that the adoption of modern varieties (MVs) has had a positive impact on productivity growth and food security in Asia and Latin America (Evenson and Gollin, 2003; Pingali, 2012). However, the increasing availability of modern hybrid species has been recognized as one of the main causes of a narrowing cultivation of local landraces (LLs), also called traditional varieties, causing in turn, a rapid declination of inter and intra agricultural genetic diversity conserved on-farm (Harlan, 1972; Altieri, 1999; Pascual and Perrings, 2007). As a matter of fact, the second report on the State of the World's Plant and Genetic Resources for Food and Agriculture finds that, since the beginning of the Green Revolution, the number of the worldwide consumed crop varieties have gradually decreased and, currently, only four crops provide 60% of human food energy (FAO, 2010). The trade-off between agricultural productivity and the conservation of crop genetic diversity has historically been resolved in favour of the former. Especially in sub-Saharan Africa, where 26.8% of the population is afflicted by chronic undernourishment (FAO, 2012), prioritizing food security through the diffusion of new technologies seemed to

⁎ Corresponding author at: University of Rome Tor Vergata, Department of Economics and Finance, Via Columbia 2, post code 00133 Rome, Italy. E-mail addresses: [email protected] (M. Coromaldi), [email protected] (G. Pallante), [email protected] (S. Savastano).

http://dx.doi.org/10.1016/j.ecolecon.2015.09.004 0921-8009/© 2015 Elsevier B.V. All rights reserved.

be the only rational strategy. To this end, sub-Saharan African countries followed the example of other developing countries and fostered a shift from a traditional agriculture to an intensive one. National programmes and international organizations have concentrated their efforts on providing marginalized smallholders with high-yielding varieties of cash crops by assisting them in capacity building in new agro-technology use (Tripp and Rohrbach, 2001). In these cases, the results have been uneven (Otsuka and Larson, 2013). For instance, while maize yields increased by 60 and 56% in South-East Asia, and Latin American and Caribbean countries respectively, between 1970 and 2012, sub-Saharan Africa has only seen a 22% increase (FAOSTAT, 2014). Among the causes of these productivity differentials, structural market failures have been demonstrated to play a fundamental role as they prevent farmers from optimally combining all the elements necessary for the best agricultural responsiveness of MVs (Dercon and Gollin, 2014; Collier and Dercon, 2013). In fact, modern seeds in isolation do not necessarily improve yields; rather, they are expected to outperform LLs only if accompanied by simultaneous use of complementary inputs such as chemicals fertilizers, pesticides or herbicides (Narloch et al., 2011; Teklewold et al., 2013). However, in sub-Saharan Africa, the access to these inputs is hampered by important and long lasting market frictions. High transport costs, failures to deliver credit to producers, price fluctuations, informational barriers and, in general, poor market infrastructures (Conley and Udry, 2010; Croppenstedt et al., 2003; Liverpool and Winter-Nelson, 2010), can radically reduce the returns of an investment in intensive agriculture. This translates in an extremely low adoption rate of agro-chemicals as pictured by the 12.9 kg per

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

hectare utilized in the sub-Saharan African region against the 174.3 kg per hectare employed, on average, in South-East Asia (WBG, 2013). A second driver of productivity gaps has recently been identified in non-market elements. Contradictory findings on the potential yield of MVs in sub-Saharan African have emerged as a result of their unsuitability to extreme agro-climatic conditions regardless of the use of recommended rates of agro-chemicals (Lipper and Cooper, 2009; Tittonell and Giller, 2013; Cavatassi et al., 2011). In fact, while the new varieties are genetically uniform and developed for high responsiveness and simplified monoculture systems (Perrings et al, 2006; Smale, 2005), LLs are the outcome of an evolutionary selection process driven by local agroecological characteristics as well as the farmers' subsistence requirements (Altieri, 1999). The heterogeneity of this selection process makes LL incredibly adaptable to degraded and poor soils, water scarcity, droughts, and biotic/abiotic stresses. This motivates marginalized farmers to still cultivate a diversified portfolio of traditional crops as a strategic asset to face agricultural shocks and climate change (Bellon, 2004; Jarvis et al., 2008; Mercer and Perales, 2010). Exploiting the ecological services supplied by the crop variability, farmers can minimize agricultural risks (Di Falco and Chavas, 2009; Di Falco and Perrings, 2005) and indirectly operate as “custodians” of the local agricultural biodiversity stock for future bioprospecting activities (Narloch et al., 2011; Quaas and Baumgärtner, 2008). The described behavioural pattern might be extremely relevant in the creation of development policies for sub-Saharan African countries, where around the 63% of the population still live in rural areas. In fact, except for the promotion of modern agriculture, the questions of how to increase food security and reduce poverty in such areas remain at the top of international agenda. In view of agro-environmental insights, the “silver-bullet” approach of incentivizing the intensification and the adoption of standardized MVs, irrespective of the socio-economic constraints and/or the ecological context, is increasingly criticized in favour of a best-fitting sustainable strategy to be implemented at farm level (Conway and Barbier, 2013; Giller et al., 2009). Since the distribution of costs and benefits associated with a new agricultural technology is heterogeneous (Suri, 2011; Narayanan, 2014), it is therefore likely to observe groups of adopters facing real or financial returns lower than their expectations (Duflo et al., 2008; Dercon and Christiaensen, 2011). The heterogeneous outcome can be the direct consequence of a not-fully informed adoption of new technologies by the farmers. For instance, farmers can be unaware of how a MV will perform in the long term or under a specific agro-ecological and market access framework, or they may be unable to consistently adjust the input utilization rates to soil nutrient requirements (Barham et al., 2014; Isik and Khanna, 2003; Feder, 1980). In this paper we empirically explore under which circumstances the adoption of MVs is a strictly optimal strategy for smallholder farmers in terms of welfare and, further, we verify the effects of such adoption decisions on the conservation of crop diversity. The assessment of nonadopters' and adopters' relative performances, in the counterfactual scenario of the observed individual strategy, has scope for providing suggestions to design agricultural policies that could address rural development and crop biodiversity conservation problems simultaneously. To address this issue, we use the nationally representative Uganda Panel Survey of the Living Standard Measurement Survey on Agriculture (LSMS-ISA), consisting of 3123 households, carried out in 2009–2010. As we do not have a randomized control experiment, we employ an endogenous switching regression model. This empirical framework allows us to overcome the challenges associated with the unobserved heterogeneity and the potential endogeneity that may affect the consistent estimation of welfare and crop biodiversity outcome variables. Therefore, while this paper adds to the growing recent literature on drivers of MVs adoption in sub-Saharan Africa (Asrat et al., 2010; Cavatassi et al., 2011; Kassie et al., 2011; Asfaw et al., 2012; Teklewold et al., 2013; Shiferaw et al. 2014), our first contribution is the investigation of the cross-impact of adoption on crop diversity conservation at the farm

347

level by simultaneously verifying the effects of market and agroecological constraints. There is still insufficient empirical studies conducted on the impact of crop diversity on welfare (Di Falco et al., 2007, Di Falco and Chavas, 2009; Di Falco et al, 2010), however, in respect to existing studies, we provide further evidence by treating crop diversity as an indirect outcome of the farmers' livelihood strategies, and for this reason, directly affected by the adoption decisions. We examine the implication of MVs adoption on different outcome variables. More specifically, net crop income and food expenditure per capita are used as welfare indicators, and the crop richness and evenness as indicators of crop diversity. Second, we explore and discuss the results' sensitivity to variations of an aggregate relative index of intensification. This index is an adaptation from agronomic studies (Herzog et al., 2006) and has never been utilized in empirical economic assessments. Nonetheless, it is particularly relevant to determine the complementary effects of a mixed combination of external inputs on a MV responsiveness and on the depletion of diversity. Third, we supply an assessment of the impact of adoption according to different Ugandan agro-ecological zones as well as soil quality levels. Finally, by characterizing the nature of the beneficiaries and losers of the adoption of MVs, we can identify, on one hand, the target population for a proper sustainable intensification approach and, on the other hand, those farmers who face no opportunity cost for the cultivation of LLs, and should be sustained for their conservation of crop biodiversity. The rest of this paper is organized as follows. The next section presents the conceptual framework and explains the empirical strategy, Section 3 illustrates the data, Section 4 explores and discusses the results, while Section 5 concludes. 2. Conceptual Framework and Estimation Procedure The adoption of a MV can be viewed as a binary voluntary decision by farmers who maximize their expected utility according to their individual heterogeneous characteristics, as well as according to local structural factors (Suri, 2011). Adopting farmers should optimize this objective through the intensification of their land endowment so as to obtain the best average yield response, expressed in terms of agricultural welfare, from a single or few MVs. However, while the adoption of the new technology is a decision that creates two mutually exclusive groups, the properly implemented intensification level is not purely deterministic. For example, we can observe MV cultivation in low intensive agroecosystems or a higher than average crop diversification in high intensive farms. In fact, as pointed out by the seminal paper of Feder (1980), the choice to adopt a MV is subject to a certain degree of uncertainty determined by the fact that a farmer builds an expectation on which will be the optimal intensification “package” for the best responsiveness of the new technology. The extent to which such expectations will be consistent with the achievable real performance depends on market failures and agro-ecological factors (Altieri, 2004; Kijima et al., 2011). The larger the market constraints, and the less the suitability of the MV to the agricultural local system and poor soil quality, the greater will be the gap between expectations and current returns. Alternatively, a farmer can decide to not-adopt the MV and continue to rely on the cultivation of a diversified LLs portfolio in a framework of risk minimization behaviour. Farmers, aware of the potential negative effects stemming from their inability to optimally support the MV cultivation, cultivate the LLs to avoid facing financial vulnerability associated to the intensification strategy (Weitzman, 2000; Di Falco and Chavas, 2009). In this context, the individual MV adoption pattern not only affects the overall agricultural outcome, but also indirectly drives the crop diversity that is conserved on-farm. In fact, while non-adopters maintain a high crop richness to minimize the impacts of market and climatic shocks, if adopters could perfectly adjust the inputs rates to the new variety requirements, hence, just one MV should be cultivated to obtain the best yield response (Omer et al., 2010). The diversity of crops cultivated is therefore an outcome of the decision to adopt and, like

348

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

agricultural welfare, is dependent on all the factors that affect such decisions and/or limit the implementation of the proper intensification (Altieri, 2009). Farmers are therefore expected to select a strategy according to their observable and unobservable characteristics, and yet it is likely that other endogenous factors could randomly bias this selfselection (Kabunga et al., 2014; Brugnach et al., 2011). As a result, not only adopting farmers risk obtaining an agro-economic outcome that is lower than the one they could have achieved by maintaining a diversified set of LL, but the stock of local unique crop biodiversity would be reduced or lost. 2.1. Econometric Approach Since we want to investigate under what circumstances and constraints does the MVs adoption become an effective means of enhancing farmers' welfare, as well as what is the complementary magnitude of the negative impact on crop variability, a counterfactual scenario of the potential result achievable in the opposite case of the observable individual behaviour is required. This can be assessed empirically by utilizing conditional expectations from an endogenous switching regression model (ESR). The ESR analyses the binary adoption decision, and the implications it has on welfare and diversity in a two-stage framework. The use of the ESR to evaluate the technology adoption in agriculture is quite diffused (Alene and Manyong, 2007; Noltze et al., 2013; Abdulai and Huffman, 2014; Cavatassi et al., 2011). In fact, the adoption decision, in a context of cross sectional analysis and without a randomized controlled experiment, might suffer of sample selection and endogeneity biases. Sample selection bias refers to the case where the voluntary decision to adopt is observed only by a restricted, nonrandom sample. The adoption status may be endogenous when the decision to adopt or not adopt is correlated with unobservable factors that affect the outcome variables. The failure to control for this correlation yields an estimated downward biased adoption effect on outcomes. These factors are unknown to researchers, but accounted for in farmers' expectations, affecting both the decision to adopt and the outcome variables. Moreover, since the outcome gap between adopters and nonadopters is assumed to be systematic, two different outcome equations are estimated in the ESR. The covariates are assumed to have different impacts on the two groups of farmers while a pooled sample would have considered the difference between groups as just intercept shifters. Therefore, with an ESR model, endogeneity and sampleselection (Hausman, 1978; Heckman, 1979) are both taken into account. The econometric specification is as follows: δ
¼ α 0 ðyMV −yLL Þ þ z0 γ þ ε


δ ¼ 0 if δ
≤0 δ ¼ 0 if δN0

ð1Þ ð2Þ

Eqs. (1) and (2) are the specification of a probit model for the dichotomous adoption decision (criterion function) in the first stage (Maddala, 1983). δ⁎ is the latent variable that determines if a farmer is a MV adopter or not, and is based on the farmers' expectations regarding the relative performance of the new technology in respect to the LLs, expressed in terms of an outcome variable y; δ⁎ is not observable but we observe δ, which is the MV adoption dummy; z′ is a vector of covariates that are relevant for the adoption decisions; α and γ are unknown parameters vectors to be estimated and ε is a random disturbance term with zero mean and σ2 variance. Eqs. (3) and (4) represent the regime equations, in the second stage, that we observe conditional to adoption decisions made at the first stage: yLL ¼ φ0 βLL þ η if δ ¼ 0

ð3Þ

yMV ¼ φ0 βMV þ ϵ if δ ¼ 1;

ð4Þ

where φ' is a vector of covariates that affects y and may overlap with z', but with the caution, for the model identification purpose, to have at least one instrument in the criterion equation that is not in the regime equations; βMV and βLL are vectors of parameters to be estimated, ϵ and η are random disturbances terms with zero mean and σ2ϵ and σ2η variance. The covariance matrix is:   σ2  ϵ X  ðε; ϵ; ηÞ ¼  σ ηε   σ ϵε

σ ηε σ 2η σ ηε

 σ ϵε   σ ηε   2  σε

ð5Þ

where σε equals 1 since α and γ are estimable only up to a scale factor (Greene, 2008). Moreover, σηε = 0 because it is not possible to observe adoption and non-adoption outcomes contemporary (Maddala and Nelson, 1975). Estimation of the covariance terms can provide a test for the endogeneity through the significance of the following correlation coefficients: ρϵε ¼ σ ϵε =σ ϵ σ ε ; ρηε ¼ σ ηε =σ η σ ε

ð6Þ

These correlations have also an economic interpretation that will be explained in the description of results. The expected values of the truncated errors are equal to: 

 ξ σε   Eðηjδ ¼ 0Þ ¼ −σ ηε λη ¼ −σ ηε ξ 1−F σε f

 ξ σ  ε ξ F σε

ð7Þ



f

Eðϵjδ ¼ 1Þ ¼ σ ϵε λϵ ¼ σ ϵε

ð8Þ

where λϵ and λη are the Inverse Mill Ratios estimated at ξ = α ' (yMV − yLL) + z ' γ and f and F are, respectively, the density and the cumulative distribution function. As explained in Lokshin and Sajaia (2004), the ESR can efficiently be estimated with the full information maximum likelihood (FIML) approach, ensuring the simultaneous estimation of the probit model and regime equations with consistent standard error. 2.2. Treatment Effect The conditional expectations from the ESR can be used to estimate the average treatment effects (ATE) of the counterfactual scenario for both the groups. Expectations conditional to adoption decision are estimated as follows (Di Falco et al., 2011): EðyMV jδ ¼ 1Þ ¼ φ0 βMV þ σ ϵε λϵ

ð9Þ

EðyLL jδ ¼ 1Þ ¼ φ0 βLL þ σ ηε λϵ

ð10Þ

EðyMV jδ ¼ 0Þ ¼ φ0 βMV þ σ ϵε λη

ð11Þ

EðyLL jδ ¼ 0Þ ¼ φ0 βLL þ σ ηε λη

ð12Þ

Eqs. (9) and (12) are the actual outcome expectations conditional to the adoption status chosen by farmers. These represent the expected outcome of MVs adopters when they adopt and the non-adopters outcome when they do not adopt. Eqs. (10) and (11) evaluate the outcomes in the counterfactual case that adopters did not adopt and that nonadopters adopted thereby providing a measure of the relative performance of the status for which the farmer has opted. Thus, the ATE of

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

adoption on adopters (TT) and the ATE of adoption on non-adopters (TU) are equal to:   TT ¼ EðyMV jδ ¼ 1Þ−EðyLL jδ ¼ 1Þ ¼ φ0 ðβMV −βLL Þ þ σ ϵε− σ ηε λϵ ; ð13Þ   TU ¼ EðyMV jδ ¼ 0Þ−EðyLL jδ ¼ 0Þ ¼ φ0 ðβMV −βLL Þ þ σ ϵε− σ ηε λϵ : ð14Þ 3. Data Description We use data from the 2009/2010 LSMS-ISA, which was gathered by the Development Economics Research Group of the World Bank. Information on rural households (HHs) characteristics, crops cultivated, agricultural input use, production costs and availability of extension services were gathered over the two cropping seasons by means of a structured questionnaire that covers the four regions and the 111 districts of Uganda. The LSMS-ISA survey data also recorded geo-referenced enumeration area level latitude and longitude coordinates by using handheld global positioning system (GPS) devices, which creates the possibility of linking household level with geo-referenced climatic and soil data by means of the Thin Plate Spline method. Climatic variables are analysis data on monthly rainfall and temperature from the National Oceanic and Atmospheric Administration (NOAA) and the European Centre for Medium Range Weather Forecasts (ECMWF), respectively.1 Considering the data limitation for some of the utilized variables, the final sample consists of 2124 national representative HHs. Uganda represents a suitable case to verify our conceptual framework for multiple reasons. First, Uganda is a “hot spot” in terms of biodiversity richness. It is an important global centre of plant diversity that supports a large variability of crop landrace diffusion (Tumuhairwe et al., 2003; Hartter et al, 2012) along different agro-ecological zones (AEZ2) (Frisvold and Ingram, 1995). Second, as the largest country among the sub-Saharan African (SSA) countries, Uganda faces a severe problem of agricultural productivity and food insecurity. From 2000 to 2010 the proportion of the population that was undernourished had increased from 26.5 to 33% (WBG, 2013). In 2001 the Government of Uganda launched a plan for the modernization of national agriculture (PMA) with the aim to shift production from a subsistence-based towards a modern marketoriented agricultural system (MAAIF, 2000). For this purpose, several MVs of staple crops, particularly of maize, were released in the market (Sserunkuuma, 2005). The adoption decision is purely voluntary as the purpose of the PMA was enabling a bottom-up demand driven environment for the technology diffusion that should, then, be based on HH's private evaluations3 (NAADS, 2001; Pamuk et al., 2014). 1 Rainfall data are extracted from the Africa Rainfall Climatology version 2 (ARC2) for each decade (i.e. 10 day intervals). ARC2 data are based on the latest estimation techniques on a daily basis and have a gridded spatial resolution of 0.1° (~10 km). Our temperature data are surface temperature measurements (in degrees Celsius) at decadal intervals obtained from the ECMWF's ERA-Interim (ECMWF re-analyses) database at a spatial resolution of 0.25° (about 28 km). As pointed out by Auffhammer et al. (2013), gridded datasets are a good source of rainfall and temperature data since they potentially adjusts for issues like missing station data, elevation, and the urban heat island bias in a reasonable way. 2 According to FAO (1978) “an agro-ecological Zone (AEZ) refers to the division of an area of land into smaller units, which have similar characteristics related to land suitability, potential production and environmental impact”. 3 A top-down technology pushing approach could lead to inadequacy of an ESR model since it would exclude the possibility of farmers' self-selection. This approach has been massively utilized in the past in SSA, but many countries began to revise it in the past decade. Uganda is an exemplar way from this point of view, jointly with other SSA countries as Tanzania, Ethiopia, Nigeria, Mali and Malawi (Pamuk et al., 2014). The modern varieties adoption in these countries is not purely forced by the national governments, but encouraged through a participatory and decentralized system of access to agricultural markets, inputs and new technologies availability facilitation as well as the provision of extension services for a capacity building based on farmers' needs and autonomous decisions of participation (WB, 2007). The NAADS, the National Agricultural Advisory Services'programme launched in 2001 by the Government of Uganda, aimed at developing a demand driven, farmer-led agricultural service delivery system. Although the programme was a public intervention, farmers have to decide whether to participate in the programme or not (Benin et al., 2012).

349

Descriptive statistics with statistical significance tests on equality of means for the groups of non-adopters and adopters and the total sample are illustrated in Table 1. The adoption status, the dependent variable of the criterion equation, is a dummy with value 1 if the farmer adopts a MV. The adopters represent the 26.1% of the total sample. Four variables were selected as outcome of the regime regressions. The annual total profits per hectares from crops cultivated (profits_ha) and the annual food consumption expenditure per capita (consumption_pc) are proxies for the HH welfare, while the number of crops species per hectares (crop_richness) and the Shannon index4 (crop_evenness) represent a measurement of the crop diversity conserved on-farm.5 On average, the adopter group shows a lower value of crop profits and a not statistical difference of food consumption expenditure, while the two diversity indicators exhibit a higher measure in the non-adopter group. We also selected five aggregates of covariates, expected to influence adoption and outcomes, including: HH characteristics, geographic and climatic factors, soil and agro-stresses, agricultural inputs, and finally, market constraints and availability of extension services (Di Falco et al., 2011; Bryan et al., 2013; Teklewold et al., 2013; Noltze et al., 2013; Shiferaw et al., 2014). Significant differences between a large number of covariates are observable between groups. Three variables that are worth being described in greater detail include info, radio and the intensification index I. First, info is a dummy that assumes value 1 if HH accessed agricultural information from extension workers/cooperatives/government in the last year. The 39% of adopters declared to have access to this service against the 25% of non-adopters. Second, radio is a dummy for the HH radio ownership. The adopters show a higher proportion of possession. As in Di Falco et al. (2011), we follow a simple falsification test to verify if both info and radio can be valid instruments for the ESR model identification. The hypothesis is that the two variables are correlated with the adoption probability through access to information such as from agricultural programmes or trainings, but unlikely to directly influence the outcome variables.6

C

4 The Shannon diversity index is calculated as: H j ¼ −∑c¼1 pc lnpc, where pc is the proportion of area cultivated with crop c on the total cultivated area of the farmer j. It measures the uncertainty to predict the identity of an individual that is randomly taken from a community. The higher is the index, the higher the uncertainty and consequently the evenness in the dataset is lower. The use of Shannon index in measuring the different components of the agricultural biodiversity is diffused in literature (Mäder et al., 2002; Di Falco and Perrings, 2005; Mouysset et al., 2012). Also the crop richness has been widely used in literature (see Di Falco et al., 2010). For all the pro and cons of each index, see Duelli and Obrist (2003). 5 As recently noted (Chavas and Di Falco, 2012), the analysis of the determinants of diversification at farm level should be based on a multi-crop framework since it helps at capturing the impact of risk mitigation strategies and control for productive crop complementarity effects. Farmers observe the overall effect of the crop mix allocated, included the new technologies, on her total agricultural welfare as an investor who evaluates the overall returns from a diversified composition of assets in the classic financial portfolio theory. Moreover, on average, in our sample HHs cultivate less than two plots, with around the 45% of HHs owning only one plot. By turning the analysis to a single crop, other than losing the possibility to empirically utilize the diversity indices as outcome variables, we should also face the issue of inputs allocation to each crops on one plot. Recently, Carpentier and Letort (2011) addressed such issue by demonstrating as the methodologies mainly utilized for allocation fails in producing consistent estimations when there is unobserved heterogeneity in the sample. Nonetheless, we address the potential existing bias from pooling several crops by controlling for the share of maize on the HH cultivated land, as usefully suggested by an anonymous referee. 6 The information source has been already used as instrument for technology adoption in Di Falco et al. (2011) and Shiferaw et al. (2014). The falsification test is conducted via probit and OLS. Results show that the instruments are significant in the selection equation. Info does not affect profits_ha, crop_richness and the crop_evenness, but have an impact on consumption_pc, while radio is never statistically significant. We also tested distance, but the results does not allow to accept it as relevant instrument. Estimations of falsification test are available upon request.

350

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

Table 1 Descriptive statistics. Non-adopters (73.9%) Variable Outcome variables profits_ha consumption_pc crop_richness crop_evenness

Adopters (26.1%)

Total (N = 2124)

Description

Mean

(SD)

Mean

(SD)

Mean

(SD)

Annual HH crop profits (US$/ha) Annual pc food consumption expenditure (US$) Crops cultivated (number of crops/ha) Exp of Shannon index

279.504⁎⁎⁎ 298.912

(328.73) (194.44)

237.919⁎⁎⁎ 316.055

(294.73) (213.13)

268.228 302.825

(320.78) (205.20)

6.607⁎⁎⁎ 8.196⁎⁎⁎

(10.90) (13.72)

5.060⁎⁎⁎ 6.649⁎⁎⁎

(6.98) (8.35)

6.275 7.796

(6.20) (12.57)

47.277⁎⁎ 5.295⁎⁎⁎ 7.626⁎⁎⁎ 5.425⁎⁎⁎ 1.774⁎⁎⁎

(15.07) (2.20) (7.01) (2.84) (2.23)

45.437⁎⁎ 5.670⁎⁎⁎ 9.260⁎⁎⁎ 5.840⁎⁎⁎ 2.497⁎⁎⁎

(13.43) (2.49) (8.21) (2.94) (3.08)

46.801 5.392 8.048 5.532 1.962

(14.68) (2.28) (7.38) (2.87) (2.50)

0.219 0.224⁎⁎⁎ 0.257⁎⁎⁎ 0.300⁎⁎⁎ 0.526⁎⁎⁎

0.201 0.343⁎⁎⁎ 0.366⁎⁎⁎ 0.090⁎⁎⁎ 0.643⁎⁎⁎

0.287 0.187⁎⁎⁎ 1259.615⁎ 21.753⁎⁎ 1214.170⁎

(0.41) (0.41) (0.43) (0.45) (0.49) (0.45) (0.39) (250.68) (1.811) (192.57)

1268.855⁎

(0.40) (0.47) (0.48) (0.28) (0.48) (0.45) (0.24) (247.09) (1.645) (165.996)

0.214 0.255 0.285 0.246 0.556 0.288 0.155 1241.67 21.911 1228.273

(0.41) (0.43) (0.45) (0.43) (0.49) (0.45) (0.36) (251.74) (1.79) (187.37)

0.293 0.557 0.015⁎⁎⁎ 0.580⁎ 0.419⁎ 0.054⁎

(0.45) (0.49) (0.12) (0.42) (0.35) (0.23)

0.311 0.558 0.038⁎⁎⁎ 0.631⁎ 0.369⁎ 0.065⁎

(0.46) (0.49) (0.19) (0.41) (0.38) (0.24)

0.299 0.557 0.021 0.599 0.401 0.058

(0.45) (0.49) (0.14) (0.42) (0.36) (0.23)

0.253⁎⁎⁎ 6.489⁎⁎⁎ 234.530⁎⁎ 91.455 2.397⁎

(0.19) (9.63) (242.93) (394.40) (16.90)

0.311⁎⁎⁎ 8.651⁎⁎⁎ 296.616⁎⁎ 103.691 9.770⁎

(0.21) (10.85) (303.56) (404.52) (168.66)

0.269 7.005 250.592 94.62 4.315

(0.20) (9.91) (261.60) (396.985) (87.23)

0.023⁎ 0.021⁎ 0.427 0.624⁎⁎⁎ 0.248⁎⁎⁎

(0.15) (0.14) (0.49) (0.48) (0.43)

0.038⁎ 0.032⁎ 0.423 0.707⁎⁎⁎ 0.388⁎⁎⁎

(0.19) (0.17) (0.49) (0.44) (0.48)

0.027 0.024 0.428 0.646 0.284

(0.16) (0.15) (0.49) (0.47) (0.45)

Covariates 1) HH age Age of HH head (years) edu Education of HH head (years) distance Distance from main road (Km.) hh_size Number of HH components area Area cultivated (ha) 2) Geographic and climatic region_c Dummy central region region_e Dummy eastern region region_n Dummy northern region region_w Dummy western region warm_aez Dummy warm AEZ cool_aez Dummy cool AEZ subhumid/cool_aez Dummy subhumid cool AEZ elevation Elevation (metres o.s.l) temperature Avg annual temperature (celsius) precipitation Avg annual rain precipitation (mm) 3) Soil and agro- stresses erosion Dummy if experienced soil erosion drought Dummy if experienced drought floods Dummy if experienced floods soil_good Dummy if soil is fair quality soil_poor Dummy if soil is poor quality crop_stress Dummy if experienced crop stress/disease 4) Agricultural inputs maize_share Share of maize on cultivated land I Herzog intensification index family_lab Annual family labour (man days/ha) organic Annual organic fertilizer (Kg/ha) livestock Tropical livestock unit 5) Market constraints and extension services input_constr Dummy if experienced high inputs prices output_constr Dummy if experienced low outputs prices credit Dummy if have access to credit radio Dummy if own radio info Dummy if received agro-information

0.293 0.065⁎⁎ 1190.28⁎ 22.362⁎⁎

⁎ Significant at 10%. ⁎⁎ Significant at 5%. ⁎⁎⁎ Significant at 1%.

Third, I is an agricultural intensification index adapted from the agronomic literature (Herzog et al., 2006; Blüthgen et al., 2012). The variable I aggregates different indicators of intensification as follows: X5 Ij ¼

n¼1



   h j −hi;mindis = hi;maxdis −hi;mindis

n for each farmer j ¼ 1; :::; J;

 100;

ð15Þ

where, for each HH j, hj is the observed value of an intensification indicator n and hi ,mindis and hi ,maxdis are, respectively, the lowest and highest values of the intensification indicator observed for HH i in each district where HH j resides. The index goes from 0 to 100 supplying a relative overview of the local intensification level and holds the advantage of controlling for unobservable agro-climatic and market constraints characteristics that affect the use and the effectiveness of inputs on outcome variables. This index has the original aim to account for the synergetic negative impact on diversity conservation that the simultaneous use of different inputs or modern agro-practices can jointly determines (Billeter et al., 2008; Hendrickx et al., 2007). In our analysis, the

complementarity of inputs is fundamental also with regard to the impact on welfare outcomes. Since MVs are highly responsive in intensified agro-ecosystem, in order to increase the probability of a good productive performance an inclusive “package” of inputs should be combined with their adoption.7 On the other hand, an unbalanced use of agricultural external inputs could generate unexpected adverse or insignificant impacts on MVs productivity (Narloch et al., 2011; Teklewold et al., 2013). Mean and SD statistics of the four intensification indicators utilized for the estimation of I are showed in Table 2 for both the groups and by clustering the districts summaries at the AEZ8 level. In the same table is also reported the most recent and relevant literature citing these variables as intensification indicators. 7 By utilizing the inputs separately we would lose the possibility to control for such synergetic impacts on both welfare and diversity outcomes. We could have estimated our model with all possible interaction terms between inputs and practices, but failing to control for the complementarity effects between more than two inputs (Place et al. 2003). 8 We aggregate the subhumid/warm zone with the tropic-warm zone because of the low proportion of HH living in the former (subhumid/warm non-adopters = 3.54%, subhumid warm adopters = 0.88%).

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

351

Table 2 Indicators of agricultural intensification.

Variable of intensification

Literature

Chemicals (Kg/ha)

Foley et al. (2011)

Pesticides (Kg/ha)

Flohre et al. (2011)

Hired labour (man days/ha)

Takeshima et al. (2013)

Average length of fallow (years)

Rossi et al. (2010)

Tropic warm

Tropic cool/sub-humid

Tropic cool/humid

Mean (SD)

Mean (SD)

Mean (SD)

Non-adopters

Adopters

Non-adopters

Adopters

Non-adopters

Adopters

0.376⁎⁎ (5.202) 1.239 (20.822) 15.856⁎⁎⁎

1.288⁎⁎ (7.926) 1.172 (8.078) 58.786⁎⁎⁎

0.100⁎⁎⁎ (1.215) 0.051⁎⁎⁎

1.326⁎⁎⁎ (6.807) 0.422⁎⁎⁎

0.787⁎⁎⁎ (9.437) 0.743⁎

2.770⁎⁎⁎ (17.426) 2.294⁎

(0.438) 13.323⁎⁎⁎

(1.613) 38.008⁎⁎⁎

(8.966) 11.777⁎⁎⁎

(10.883) 22.685⁎⁎⁎

(31.752) 1.194 (2.591)

(38.281) 1.155 (2.836)

(24.785) 0.683⁎ (2.222)

(59.309) 0.219⁎ (0.834)

(31.622) 1.031⁎⁎ (2.729)

(38.749) 0.618⁎⁎ (2.123)

⁎ Significant at 10%. ⁎⁎ Significant at 5%. ⁎⁎⁎ Significant at 1%.

As expected, in all the AEZs, with the exception of pesticides in the tropic warm AEZ, the non-adopters use less agro-chemicals, and hired labour compared to the adopters, while non-adopters apply a longer period of fallow. By using the individual values of the intensification variables, we estimate the average intensification index reported in Table 1. 4. Results and Discussion 4.1. Criterion and Regime Regressions Tables 3 and 4 present the results obtained from the ESR model, which is estimated at HH level using the FIML approach with a log-log functional form, district clustered standard errors and the inclusion of district fixed effects to address unobserved heterogeneity at the district level (Di Falco and Veronesi, 2013; Di Falco and Bulte, 2013).9 Estimations of criterion equations are illustrated in columns (1), (4), (7) and (10). The regime equations of non-adopters and adopters are respectively illustrated in models (2) and (3) for crop profits, (5) and (6) for per capita food consumption expenditure, (8) and (9) for crop richness, and (11) and (12) for crop evenness. All the criterion equations feature similar impact of covariates in terms of signs and significance on the MV adoption probability. In line with the literature, we note that having a higher educational level increases the probability of adoption. This is due to the increased access to information regarding the real performance of new technology (Doss, 2006) that goes along with a higher education. On the other hand, we note that increasing transaction and transportation costs, approximated by distance to the main road, reduces adoption of MV (Minten et al., 2014). The climatic variables are determinants of the decision to adopt; temperature, specifically, shows a non-linear concave behaviour. We also find that while farmers who experienced floods or crop diseases are more likely to switch to MV, households with poor soil quality are less likely to adopt. Agricultural intensification, as expected, is associated with a high probability of adoption, and this probability increases with intensification. We also highlight that the two variables radio and info, used as instruments, are both positive and significant drivers of MV adoption. 9 Estimations at plot level are also available upon request (on a sample of 8046 plots). The analysis at plot levels allows to include pseudo-fixed effects à la Mundlak (1978), which address the issue of unobserved heterogeneity at farm level (agricultural skills) by introducing the mean of plot variant explanatory variables (intensification index, family labour, organic fertilizer, soil quality and erosion). Nevertheless, in our sample, given the low level of fragmentation these controls are not significant as confirmed by an F test for the joint significance of the plot variant variables (Non-adopters equation: chi2(5) = 6.36, Prob N chi2 = 0.2729. Adopters equation: chi2(5) = 8.49, Prob N chi2 = 0.1310). Moreover, results at plot level corroborate estimations at HH level and strengthen our confidence in the robustness of the results.

The fact that in all regime regressions there are remarkable differences between groups for most of the coefficients, suggests to us that the ESR model is preferable to the pooled model estimation with common slope coefficients. There are three main results in the crop profits outcome regression (columns 2 and 3). First, the economic performance of LLs is affected neither by agricultural risk-factors, with the exception of droughts, nor by high inputs and low prices of output. By contrast, the adopters' equation reveals that all the agricultural stresses and market constraints are significant and negative drivers of profits. These findings, per se, seem to imply that a diversified mix of LLs allows the conservation of a set of ecosystem soil services that operate as natural substitutes for agro-chemical nutrients. Adopters, who are not able to optimally substitute those natural services with high-quality external inputs, because of high input prices and shortage of credit (see the related coefficients), face greater vulnerability to agricultural shocks (Cavatassi et al., 2011). Second, while an increase in temperature negatively affects the MV performance, it has a positive impact on non-adopters' profits. As pointed out by Altieri et al. (2012), LLs are more adaptable to environments characterized by weather variability. Such results are corroborated by two findings: farmers living in colder AEZ (cool and sub-humid/cool) benefit compared to those living in the warm zone, and while adopters' profits are positively dependent by rain precipitation, LL are resistant to water scarcity. Another interesting point is that the intensification coefficients of non-adopters follow a U-shape path with a turning point at a very low level of I.10 A small level of intensification signifies a positive effect also on LL performance albeit this seems mainly caused by the hired labour component of I. This statement is the consequence of a test where we run our empirical specification by disentangling the hired labour from the aggregate intensification index. We obtain that the effect of I is not significant, and the impact of hired labour is significantly positive. As expected, hence, the labour increases crop profits irrespective of the seed type utilized, as also suggested by the family labour coefficient. The concave behaviour of I on adopters' outcomes could be somewhat surprising whether we ignore the traditional poor quality of agro-chemicals in SSA. This impedes to satisfy the increasing soil micronutrients11 requirements caused by the cultivation of MVs not adapted to local agro-climatic conditions (Block, 2013; Bellon, 2004), and, therefore, produces severe shortcomings on the productive soil formation that would, instead, naturally supplied by a sustainable

10 The turning point of non-adopters is equal to 1.384, while for adopters is equal to 5.707. 11 It must be underlined that, albeit we do not control explicitly for the quality of inputs, the market constraints variables and the distance are suitable proxies for such control (Johnson et al., 2003).

352

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

Table 3 ESR estimates — agricultural welfare. A) Crop profits

B) Food pc consumption expenditure

Criterion (1)

Non-adopters (2)

Adopters (3)

Criterion (4)

Non-adopters (5)

Adopters (6)

Variable

Coeff

(SE)

Coeff

(SE)

Coeff

(SE)

Coeff

(SE)

Coeff

(SE)

Coeff

(SE)

HH age age_square edu distance hh_size

−0.509⁎⁎ 0.242⁎⁎ 0.138⁎⁎ −0.081⁎⁎⁎

(0.240) (0.111) (0.059) (0.035) (0.056)

0.211⁎⁎⁎ −0.105⁎⁎ 0.028 0.021⁎⁎ 0.165⁎⁎⁎

(0.074) (0.038) (0.026) (0.010) (0.050)

−0.186 0.071 0.091 −0.027⁎⁎⁎ 0.147⁎⁎⁎

(0.144) (0.066) (0.081) (0.010) (0.059)

−0.488⁎⁎ 0.239⁎⁎ 0.101⁎ −0.056⁎⁎ 0.079⁎

(0.211) (0.106) (0.056) (0.020) (0.045)

0.330⁎⁎⁎ −0.167⁎⁎⁎ 0.145⁎⁎⁎ −0.072⁎⁎

(0.101) (0.045) (0.033) (0.031) (0.103)

0.405⁎⁎ −0.174⁎ 0.128⁎ −0.118⁎⁎ 0.122⁎⁎

(0.196) (0.086) (0.069) (0.047) (0.054)

0.075 −0.048 0.251⁎⁎⁎ 0.065 0.018 −0.105 3.783⁎⁎

−0.118⁎ −0.111 0.511⁎⁎⁎ 0.195⁎⁎ 0.091⁎ 0.256 −3.430⁎⁎

−0.356⁎⁎ 0.245⁎⁎ 0.020 −1.249⁎⁎ −1.299⁎⁎⁎ 0.415⁎⁎ 0.681⁎⁎⁎

(0.152) (0.127) (0.178) (0.098) (0.025) (0.555) (0.333) (0.171) (0.141)

−0.267⁎⁎⁎ −0.375⁎⁎⁎ −0.144⁎⁎ 0.081 0.075⁎ −0.794⁎⁎⁎ 3.327⁎⁎ 0.305 0.143

(0.071) (0.086) (0.061) (0.054) (0.044) (0.205) (1.672) (0.206) (0.229)

−0.199⁎⁎ −0.412⁎⁎⁎ −0.123 0.163⁎⁎ 0.052 0.233 −3.495⁎

0.215 0.039⁎⁎⁎

(0.069) (0.083) (0.079) (0.098) (0.053) (0.303) (1.671) (0.206) (0.015)

0.273⁎ 0.258⁎⁎

0.358 0.021

(0.055) (0.053) (0.047) (0.045) (0.052) (0.261) (1.501) (0.293) (0.089)

0.252 0.038⁎⁎⁎

(0.094) (0.105) (0.138) (0.086) (0.041) (0.385) (2.088) (0.171) (0.010)

0.080

Geographic and climatic region_e 0.361⁎⁎ region_n 0.226⁎

−0.101

region_w cool_aez subhumid_cool_aez elevation temperature temperature_square precipitation

−0.586⁎⁎⁎ 0.059⁎ 0.007 −1.207⁎⁎ −1.173⁎⁎⁎ 0.305⁎⁎⁎ 0.792⁎⁎⁎

(0.173) (0.124) (0.196) (0.031) (0.010) (0.587) (0.282) (0.106) (0.185)

Soil and agro-stresses erosion drought floods soil_poor crop_stress

−0.049 0.018 0.189⁎⁎⁎ −0.171⁎⁎ 0.052⁎⁎

(0.060) (0.041) (0.060) (0.064) (0.016)

0.023 −0.052⁎ −0.171 −0.042 −0.118

(0.028) (0.027) (0.175) (0.043) (0.107)

−0.114⁎⁎⁎ −0.137⁎ −0.208⁎ −0.275⁎⁎⁎ −0.383⁎⁎

(0.033) (0.071) (0.107) (0.076) (0.151)

−0.018 0.051 0.110⁎⁎ −0.128⁎⁎ 0.031⁎⁎

(0.073) (0.054) (0.045) (0.052) (0.018)

−0.005 −0.024 −0.084 0.064 −0.148

(0.027) (0.098) (0.197) (0.058) (0.127)

−0.188⁎⁎ −0.112⁎⁎ −0.184⁎ −0.158⁎⁎ −0.101⁎

(0.089) (0.057) (0.108) (0.062) (0.055)

Agricultural inputs maize_share I I_square family_lab organic livestock

0.411⁎⁎⁎ 0.641⁎⁎⁎ 0.191⁎⁎⁎ 0.010 0.018 0.173⁎⁎⁎

(0.148) (0.134) (0.043) (0.015) (0.013) (0.034)

−0.101 0.161⁎⁎ −0.041⁎ 0.051⁎⁎⁎ 0.027⁎⁎

−0.199⁎⁎⁎ 0. 403⁎⁎⁎ −0.034⁎ 0.042⁎⁎ 0.028⁎⁎ −0.053⁎⁎

(0.071) (0.124) (0.017) (0.017) (0.013) (0.021)

0.389⁎⁎⁎ 0.405⁎⁎⁎ 0.075⁎⁎ 0.102 0.021⁎ 0.122⁎⁎⁎

(0.132) (0.116) (0.026) (0.072) (0.018) (0.034)

−0.155 0.322⁎⁎⁎ −0.115⁎⁎⁎ −0.074⁎⁎⁎ 0.032⁎⁎⁎ 0.035

(0.101) (0.088) (0.031) (0.017) (0.005) (0.028)

−0.244⁎⁎ 0. 381⁎⁎ −0.066 −0.098⁎⁎⁎ 0.049⁎⁎⁎

−0.025

(0.098) (0.084) (0.023) (0.011) (0.012) (0.019)

0.022

(0.101) (0.166) (0.044) (0.034) (0.011) (0.041)

0.084 0.165 0.168⁎⁎

(0.167) (0.175) (0.064)

−0.341⁎⁎⁎ −0.208⁎⁎ 0.095⁎⁎⁎

(0.108) (0.105) (0.033)

(0.159) (0.167) (0.055)

−0.136⁎⁎⁎ −0.162⁎⁎⁎ 0.075

(0.019) (0.036) (0.047)

(35.21) (0.079) (0.136)

68.411⁎⁎⁎ 0.544⁎⁎⁎ −0.642⁎⁎⁎

(22.14) (0.032) (0.021)

(0.201) (0.182) (0.031) (0.061) (0.024) (5.001)

0.174 −0.088 0.209⁎⁎⁎

81.004⁎⁎ 0.104 −0.295⁎⁎

−0.107 −0.005 0.090⁎⁎⁎ 0.323⁎⁎⁎ 0.181⁎⁎ 15.554⁎⁎⁎

57.065⁎ −0.601⁎⁎⁎ −0.531⁎⁎⁎

(32.87) (0.078) (0.102)

24.116 −0.401⁎⁎⁎ −0.415⁎⁎⁎

(41.13) (0.066) (0.155)

Market constraints and extension services input_constr −0.161 (0.178) output_constr −0.013 (0.186) credit 0.078⁎ (0.038) radio 0.172⁎⁎⁎ (0.063) info 0.242⁎⁎⁎ (0.061) Constant 11.244⁎⁎ (3.370) lnσi ρiε N 2124 ll −2877.321 χ2 Walda 74.49⁎⁎⁎

2124 −2431.004 63.87⁎⁎⁎

Note: Robust standard errors clustered at the district level in parenthesis and in italic. Fixed effects at the district level are included. a LR test for independent equations. ⁎ p b 0.1. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.

traditional farming or best-fitting agricultural practices on LLs (Conway and Barbier, 2013). The food consumption expenditure per capita regime equations (columns 5 and 6) show little differences in respect to the crop profits equation. The distance from the main road and the altitude over the sea level have a negative effect on non-adopters food consumption expenditure. This is consistent with the hypothesis that farmers in marginalized areas face higher transaction costs. Moreover they have limited off-farm employment opportunities (Richardson et al. 2012); while this should not have effect on the crop profits, it is straightforward to expect a negative impact on the per capita expenditure capacity. Finally, it is worth noting that there is a negative impact of family labour on both the groups. This may be related to the reduction in total HH income and expenditure capacity caused by the lower marginal product of agricultural labour compared to alternative qualified jobs of HH members (Benjamin, 1992).

As shown in Table 4, the positive relationship of distance on nonadopters crop richness and evenness can be explained with the fact that farmers less linked to markets, are those more likely to implement a risk minimization strategy. Through the crop diversification, marginalized farmers also obtain varied diets and can sustain their status of semi-subsistence, which is dependent on their own food production (Ribeiro Palacios et al., 2013, Jones et al., 2014). We also find that both the groups increase crop richness and evenness in response to some of the agricultural stresses and shocks confirming again that diversification is a method of coping with random agricultural shocks. It should also be noted that the two groups show different reactions to temperature increases in terms of crop diversity: while non-adopters increase diversity, adopters reduce it. As expected, in both regimes, the agricultural intensification is a negative determinant of crop diversity. On one hand, a higher use of modern agricultural inputs is coupled with a simplification of agricultural

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

353

Table 4 Estimation results — crop diversity. C) Crop richness

D) Crop evenness

Criterion (7)

Non-adopters (8)

Adopters (9)

Criterion (10)

Non-adopters (11)

Adopters (12)

Variable

Coeff

(SE)

Coeff

(SE)

Coeff

(SE)

Coeff

(SE)

Coeff

(SE)

Coeff

(SE)

HH age age_square edu distance hh_size

−0.505⁎⁎ 0.197⁎ 0.184⁎ −0.098⁎⁎⁎ 0.128⁎⁎

(0.252) (0.115) (0.097) (0.034) (0.065)

0.788⁎⁎⁎ −0.388⁎⁎⁎

−0.112 0.075 0.117 0.088 −0.221⁎⁎⁎

(0.205) (0.098) (0.072) (0.055) (0.072)

−0.499⁎⁎ 0.239⁎

0.055 0.311⁎⁎⁎ −0.075⁎⁎⁎

(0.185) (0.087) (0.046) (0.067) (0.019)

0.142 −0.089⁎⁎⁎ 0.148⁎⁎

(0.251) (0.134) (0.098) (0.032) (0.072)

0.289⁎⁎ −0.141 0.034 0.119⁎⁎⁎

(0.141) (0.089) (0.044) (0.025) (0.035)

−0.118 0.066 0.080 0.075⁎ 0.071⁎⁎

(0.145) (0.071) (0.051) (0.039) (0.036)

(0.095) (0.090) (0.105) (0.089) (0.025) (0.242) (0.311) (0.038) (0.521)

−0.114⁎⁎ 0.097⁎ 0.301⁎⁎⁎ −0.102⁎ −0.044⁎ −0.299⁎ 2.897⁎ −0.410⁎⁎ 0.476⁎⁎⁎

(0.044) (0.055) (0.091) (0.055) (0.026) (0.157) (1.523) (0.161) (0.121)

0.309⁎⁎⁎ 0.401⁎⁎⁎ −0.086⁎⁎⁎ −0.191⁎⁎⁎ 0.019 −0.345 −1.678⁎⁎ 0.161⁎ 0.469⁎⁎⁎

(0.094) (0.082) (0.016) (0.071) (0.031) (0.387) (0.721) (0.094) (0.161)

0.252⁎⁎ 0.234⁎⁎ −0.621⁎⁎⁎ 0.131 0.021 −0.137 −1.118⁎⁎⁎ 0.183⁎⁎ 1.212⁎⁎⁎

(0.101) (0.114) (0.134) (0.112) (0.024) (0.241) (0.404) (0.079) (0.452)

−0.149⁎⁎⁎ 0.114⁎

−0.169⁎⁎ 0.421⁎⁎⁎

−0.052 −0.065⁎ −0.075⁎⁎ 0.202 2.236⁎⁎ −0.242⁎ 0.278⁎⁎⁎

(0.049) (0.062) (0.045) (0.038) (0.035) (0.146) (1.004) (0.128) (0.088)

0.013 −0.115 0.025 −0.329 −1.941⁎⁎ 0.148 0.399⁎⁎⁎

(0.065) (0.116) (0.122) (0.088) (0.041) (0.255) (0.921) (0.091) (0.136)

0.047⁎ −0.016 0.037 0.043⁎⁎ 0.046

(0.025) (0.045) (0.154) (0.019) (0.036)

0.097⁎⁎⁎ −0.075⁎ −0.488⁎⁎⁎ 0.128 0.203⁎

(0.031) (0.042) (0.127) (0.082) (0.116)

−0.026 0.063 0.454⁎⁎ −0.229⁎⁎⁎ 0.081

(0.061) (0.071) (0.178) (0.065) (0.127)

0.078⁎⁎⁎ 0.025 0.024 0.038⁎ 0.057⁎

(0.021) (0.026) (0.074) (0.021) (0.031)

0.055 0.051 −0.228⁎⁎ 0.158⁎ 0.120

(0.041) (0.039) (0.099) (0.088) (0.081)

−0.128 −0.124⁎ −0.058⁎ 0.081⁎⁎⁎ 0.031⁎⁎ −0.151⁎⁎⁎

(0.087) (0.073) (0.034) (0.025) (0.015) (0.021)

−0.215⁎ −0.198⁎ −0.085 0.059⁎⁎⁎ 0.055⁎⁎⁎ −0.101⁎⁎

(0.115) (0.107) (0.062) (0.018) (0.016) (0.051)

0.228⁎⁎ 0.553⁎⁎⁎ 0.182⁎⁎⁎ 0.041 0.028 0.192⁎⁎⁎

(0.115) (0.121) (0.051) (0.036) (0.018) (0.044)

−0.159⁎ −0.115⁎⁎⁎ −0.038⁎⁎ 0.025⁎⁎⁎ 0.022⁎⁎⁎ −0.198⁎⁎⁎

(0.095) (0.038) (0.018) (0.011) (0.005) (0.034)

−0.191⁎ −0.131⁎ −0.028 0.046⁎⁎ 0.028⁎ −0.068

(0.101) (0.070) (0.031) (0.019) (0.015) (0.042)

0.083 0.127 0.061

(0.100) (0.115) (0.039)

0.274⁎ 0.075 −0.084⁎

(0.151) (0.165) (0.045)

(0.071) (0.076) (0.048)

0.191⁎ 0.038 −0.014

(0.110) (0.124) (0.029)

(24.08) (0.087) (0.114)

−33.997 −0.681⁎⁎⁎ −0.655⁎⁎⁎

(41.62) (0.052) (0.016)

(0.118) (0.184) (0.063) (0.062) (0.071) (9.98)

0.041 0.045 0.055

−55.653⁎⁎ −0.818⁎⁎⁎ −0.198⁎

0.187 −0.041 0.148⁎⁎ 0.271⁎⁎⁎ 0.324⁎⁎⁎ 17.502⁎

7.673 −0.450⁎⁎⁎ −0.693⁎⁎⁎

(5.001) (0.079) (0.034)

14.115⁎⁎ −0.714⁎⁎⁎ −0.088⁎

(6.82) 0.031 0.051

Geographic and climatic region_e 0.343⁎⁎⁎ region_n 0.208⁎⁎ region_w −0.650⁎⁎⁎ cool_aez subhumid-cool_aez elevation temperature temperature_square precipitation

0.091 0.017 −0.421⁎ −1.002⁎⁎⁎ 0.129⁎⁎⁎ 1.353⁎⁎

Soil and agro-stresses erosion drought floods soil_poor crop_stress

−0.041 −0.014 0.555⁎⁎⁎ −0.255⁎⁎⁎ −0.203

(0.074) (0.060) (0.181) (0.081) (0.160)

Agricultural inputs maize_share I I_square family_lab organic livestock

0.511⁎⁎ 0.309⁎⁎ 0.157⁎⁎ 0.029 0.033 0.189⁎⁎⁎

(0.209) (0.151) (0.071) (0.024) (0.025) (0.067)

Market constraints and extension services input_constr 0.185 (0.211) output_constr −0.075 (0.199) credit 0.133⁎ (0.054) radio 0.168⁎⁎⁎ (0.057) info 0.148⁎⁎⁎ (0.054) Constant 12.107⁎⁎ (5.61) lnσi ρiε N 2124 ll −2355.614 χ2 Walda 101.31⁎⁎⁎

0.054

2124 −2190.885 48.37⁎⁎⁎

Note: Robust standard errors clustered at the district level in parenthesis and italics. Fixed effects at the district level are included. a LR test for independent equations. ⁎ p b 0.1. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.

Table 5 Average treatment effecta (ATE). Adoption Mean

(SD)

Welfare Adopters Non-adopters

5.391 4.295

(1.43) (0.73)

Crop diversity Adopters Non-adopters

1.566 0.974

(0.71) (0.24)

Non-adoption

ATE

Mean

Δ

(SD)

Adoption %

Mean

(SD)

5.998 5.625

Crop profits (1.02) TT = −0.607⁎⁎⁎ (1.11) TU = −1.330⁎⁎⁎

−11.25 −23.64

5.722 4.764

(0.77) (0.52)

1.967 1.806

Crop richness (0.68) TT = −0.402⁎⁎ (0.54) TU = −0832⁎⁎⁎

−25.67 −46.06

1.841 1.528

(0.68) (0.46)

Non-adoption

ATE

Mean

Δ

(SD)

%

Food pc consumption expenditure 6.419 (0.48) TT = −0.697⁎⁎⁎ 5.305 (0.41) TU = −0.541⁎⁎⁎

−12.18 −10.19

Crop evenness (0.61) TT = −0.270⁎⁎⁎ (0.47) TU = −0.471⁎⁎⁎

−14.66 −23.52

2.111 1.998

a As the outcome variables in the model are log-transformed, the mean predictions are also in logarithmic form. Converting the mean back to the original unit would lead to inaccuracies due to the inequality of arithmetic and geometric means. ⁎ p b 0.1. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.

354

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

systems that reduce the crop diversification, and on the other hand the use of “traditional” inputs such as family labour and organic fertilizers support crop diversity (Tilman et al., 2002). Finally, both the richness and the evenness increase with high inputs prices. The variable is significant only for adopters who are more likely to face price vulnerability being highly dependent on markets. Consequently, as the costs of intensification rises, these farmers disinvest in intensive monocultures and turn to risk minimization behaviours.

4.2. Average Treatment Effects Estimation Table 5 reports the estimated average treatment effects for all the outcomes variables. As pointed out by the ρiε b 0 in Tables 3 and 4, there is evidence of what Fuglie and Bosch (1995) call hierarchical sorting. This means that non-adopters have better than average outcomes in both regimes, but prefer to not adopt, while the adopters face below average outcome values in both regimes, but could be better off by not adopting. As consequence, MVs adoption determines an average negative impact on both HH welfare and crop diversity. In particular, the adoption produces a loss of 11.2% of crop profits and 12.2% of food per capita consumption. The non-adopters, if they could have adopted, would have lost 23.6 and 10.2% of crop profits and food consumption respectively. With regard to crop diversity, significant negative treatment of 25.7 and 14.7% are observable in adopters' crop richness and evenness. The

non-adopters could have reduced their measures for richness by 46.1% and evenness by 23.5% by adopting. Results from Table 5 are disaggregated in order to investigate the variation of treatments by intensification level, AEZ and soil quality. Fig. 1 illustrates the percentage change of the ATE according to three classes of intensification obtained from the tertiles of the I index. The adopters' treatments are illustrated in the quadrants to the left of the vertical axis and the non-adopters' treatments in the quadrants to the right. Three main findings must be highlighted. First, the average ATE (black circles), of crop profits and food consumption expenditure per capita (panel a) shows a concave pattern. The highest average is identifiable at the middle class of intensification for both the outcomes and the adoption status. In the case of adopters, the concave path confirms that, when MVs are unsuitable to the agro-ecological context, a large intensification does not compensate the loss of natural productive benefits related to crop diversity. Second, both indicators of crop diversity (panel b) present a noticeable decreasing pace as the intensification rises: the on-farm crop diversity is increasingly reduced as farmers intensify their agro-ecosystem. Third, the graph also shows the minimum (lowest dash) and maximum (highest dash) value that the ATE assumes in each intensification class. It is worth noting that for crop profits, there is at least one positive value of treatment in the group of adopters. By using this result, we investigate the characteristics of adopters with positive ATE. They represent the 11.8% of the group. Table 6 presents descriptive statistics for this sub-group of HH, and shows the

Fig. 1. (a): Average, minimum and maximum treatment (%) on welfare outcomes, by intensification class. (b): Average, minimum and maximum treatment (%) on crop diversity outcomes, by intensification class.

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

comparisons of their mean values with those of adopters with negative treatment. In this way we can portray a profile of a representative HH that benefits from MV. Important key factors that characterize who benefit from adoption include: an higher educational level, which seems to imply that farmers reduce the gap between expected and real MVs performance by also relying on an easiest access to market and credit; a lower vulnerability to input prices; a larger area cultivated of good soil quality with a higher rain precipitation; living in the eastern and central region and in cool AEZ; a lower exposure to agricultural and climatic related shocks. Finally, Fig. 2 shows the disaggregation of ATE on crop diversity outcomes by AEZ and soil quality. The highest negative treatment is observed in the warm zone for both the regimes, followed by the cool and the sub-humid/cool zones. By soil quality, we see that in the poor soils there is the highest loss of diversity for both the groups and indicators. Combining these findings we conclude that, since in the warm zone there is the larger prevalence of degraded soils (Ayuke et al., 2011), farmers prefer maintaining a diversified agro-ecosystem to manage a fragile land endowment. Hence, HH with poor soils in the warm AEZ conserve the highest stock of crop diversity, and the potential loss of this diversity is at the maximum observable level. 5. Concluding Remarks Analysing the extent to which adoption of new agricultural technology makes marginalized farmers better off in terms of welfare and food security is of fundamental importance to promote effective agricultural policies in rural based economies. The existing empirical evidence suggests contrasting results on the effect of adopting modern varieties in sub-Saharan Africa. The current debate still investigates on the causes of limited performance of improved technologies. Another line of research concentrates on the environmental concerns related to the Table 6 Characteristics of beneficiaries and losers from adoption. Crop profits Positive TT (11.8%) Variable profits_ha consumption_pc crop-richness crop-evenness age edu distance hh_size area region_c region_e region_n region_w warm_aez cool_aez subhumid/cool_aez elevation temperature precipitation erosion drought floods soil_good I crop_stress input_constr output_constr credit ⁎ p b 0.1. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.

Mean 251.101⁎⁎⁎ 334.125⁎ 5.101 6.621 45.013 5.625⁎⁎ 9.124⁎⁎⁎ 6.101 3.343⁎⁎⁎ 0.294⁎⁎ 0.529⁎⁎⁎ 0.176⁎⁎⁎ 0.000⁎ 0.176⁎⁎⁎ 0.823⁎⁎⁎ 0.000⁎⁎⁎ 1191.945⁎ 22.372 1269.018⁎⁎ 0.271⁎⁎⁎ 0.509⁎⁎ 0.029⁎⁎⁎ 0.511⁎⁎⁎ 8.765 0.066 0.027⁎⁎⁎ 0.031 0.485⁎

Negative TT (88.2%) (SD)

Mean

(SD)

(184.201) (200.102) (6.951) (8.020) (13.983) (2.445) (8.018) (2.981) (3.223) (0.469) (0.514) (0.392) (0.000) (0.392) (0.292) (0.000) (246.898) (1.872) (193.898) (0.387) (0.423) (0.283) (0.237) (10.006) (0.241) (0.235) (0.150) (0.401)

233.874 317.201 5.005 6.616 45.305 5.119 13.267 5.838 2.226 0.190 0.340 0.381 0.087 0.634 0.298 0.066 1084.498 22.765 1261.990 0.311 0.566 0.039 0.300 8.758 0.065 0.041 0.032 0.418

(296.101) (215.893) (6.982) (8.360) (13.438) (2.515) (10.545) (2.801) (2.248) (0.393) (0.474) (0.486) (0.282) (0.481) (0.457) (0.250) (199.530) (1.787) (191.345) (0.471) (0.490) (0.191) (0.303) (10.964) (0.240) (0.212) (0.184) (0.498)

355

increasing use of modern varieties, whose diffusion is demonstrated to threaten the conservation of local landraces diversity. This constitutes a valuable genetic stock to preserve for future crop breeding efforts, and to face the potential shocks of climate change. In fact, in the presence of structural market failures and agro-ecological conditions, farmers rely on a rich set of local landraces as part of an agricutural risk minimization strategy. While recent works analysed the effects of adoption on several indicators of farmers' welfare, there is no direct empirical evidence regarding the simultaneous impact on crop diversity conserved on-farm. This paper aims to address this research gap by investigating the magnitude of the overall relative agricultural and environmental performances of the new technology among adopters and non-adopters. We use an endogenous switching regression model on data collected in 2009/2010 in Uganda. This approach allows us to account for endogeneity and self-selection bias potentially occurring in the technology adoption framework. Expectations on the utility provided by adoption of a new variety may be affected by structural endogenous characteristics, while market failures and agro-ecological constraints are likely to influence the suitability of a genetically standardized species for a specific context. All these factors vary between groups. Furthermore, the mutually exclusive nature of farmers' decision is utilized to obtain a counterfactual scenario that provides a measure of the best strategy for farmers' livelihood as well as for diversity conservation. Our results show that, in Uganda, adopters do not have better welfare effect compared to non-adopters. Specifically, both income from crop cultivation and food expenditure per capita would have, on average, been higher if adopters had not adopted. We argue that this could be the consequence of superior resistance of LL to local agroecosystem conditions and agricultural stresses in marginal lands. Farmers who conserve a diversified set of local varieties support natural agricultural processes that substitute external agrochemicals inputs in the soil nutrient management. To maximize their welfare, adopters should supply land with the appropriate combination of different inputs, but vulnerability to market prices and lack of access to credit leave farmers with a sub-optimal level of intensification, and consequently results in overall underperformance compared to nonadopters. We also illustrate that benefits from increased use of external inputs such as agrochemicals are, however, limited. Above a threshold level of intensification, the economic return of adopters is determined by low adaptability of available modern varieties to poor soils conditions, local climate and random agricultural shocks. This stems from the heterogeneous distribution of adoption costs and benefits across the rural population and creates the scope for implementing a best fitting strategy at the farm level. Such a strategy should be put forward to the dissemination of few genetically standardized major crops that farmers, once adopted, are not able to sustain with the required intensification level as far as they are economically vulnerable. In fact, our results show that farmers who benefit from MV adoption are larger owners of land of good soil quality, are better educated, and are less credit and market constrained. Their farms are located in temperate agroecological zones and, thus, are less exposed to agricultural shocks or extreme climate. The effects of adoption on crop richness and evenness conserved onfarm are always negative, though different according to agro-ecological zones and soil quality. The higher the intensification, the stronger is the reduction of local crop biodiversity. In terms of policy interventions, we note that the availability of agricultural extension services, provided by the government or NGO's, to marginalized farmers is one of the main drivers of higher rates of adoption. Though this does not directly affect the farmers' ability to optimally exploit new technology, it has a negative effect on crop diversity conservation. In this context, since our results highlight that both adopters and non-adopters react to agro-climatic shocks and variability of input prices by increasing crop richness and evenness, useful indications for

356

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

Fig. 2. Average treatment (%) on crop diversity, by AEZ and soil quality.

agricultural development and climate change adaptation policies may be learnt and generalized for other sub-Saharan African environmentally and economically fragile countries. While it is necessary to target food security by increasing the availability of high yielding varieties, incentivizing the adoption, irrespective of their adaptability to local agroecosystems, leads to shortcomings incurred by financially vulnerable farmers who, in turn, cannot afford the intensification strategy. Exploiting the stock of local varieties, sustaining marginalized farmers with capacity building in best-fitting agricultural practices, should be not only a transitional strategy until the countries reach modernization in agriculture, but a permanent programme to enhance rural livelihood and facing climate shocks. Further, governments, supporting diversity conservation by those farmers with lower opportunity costs, could focus on capitalizing the adaptability characteristics of local landraces. An R&D public investment on new varieties bioprospecting, genetically based on attractive landraces' traits, should provide countries with a structure for a modern agricultural system founded on a unique crops portfolio suitable for farmers' knowledge, local tradition, and agroclimatic exposure, and thus capable to compete in international food markets. As long as caution is needed in generalizing cross-sectional case studies, the question of the positive welfare impact of crop diversification deserve more research from academics, but also increasing attention by policy makers of developing countries. Acknowledgements The views expressed in this paper are the authors' only and should not be attributed to the institutions they are affiliated with. The authors are grateful to two anonymous referees, Anil Markandya, Alberto Longo, Laura Castellucci, and participants of the 17th ICABR Conference in Ravello (Italy, 2013) for their useful suggestions and comments. References Abdulai, A., Huffman, W., 2014. The adoption and impact of soil and water conservation technology: An endogenous switching regression application. Land Economics 90 (1), 26–43. Alene, A., Manyong, V.M., 2007. The effect of education on agricultural productivity under traditional and improved technology in northern Nigeria: an endogenous switching regression analysis. Empir. Econ. 32, 141–159. Altieri, M.A., 1999. The ecological role of biodiversity in agroecosystems. Agric. Ecosyst. Environ. 74 (1), 19–31.

Altieri, M.A., 2004. Linking ecologists and traditional farmers in the search for sustainable agriculture. Front. Ecol. Environ. 2 (1), 35–42. Altieri, M.A., 2009. Agroecology, small farms and food sovereignty. Mon. Rev. 61, 102–111. Altieri, M.A., Funes-Monzote, F.R., Petersen, P., 2012. Agroecologically efficient agricultural systems for smallholder farmers: contributions to food sovereignty. Agron. Sustain. Dev. 32 (1), 1–13. Asfaw, S., Shiferaw, B., Simtowe, F., Lipper, L., 2012. Impact of modern agricultural technologies on smallholder welfare: evidence from Tanzania and Ethiopia. Food Policy 37 (3), 283–295. Asrat, S., Yesuf, M., Carlsson, F., Wale, E., 2010. Farmers' preferences for crop variety traits: lessons for on-farm conservation and technology adoption. Ecol. Econ. 69 (12), 2394–2401. Auffhammer, M., Hsiang, S.M., Schlenker, W., Sobel, A., 2013. Using weather data and climate model output in economic analyses of climate change. Rev. Environ. Econ. Policy, ret016. Ayuke, F.O., Pulleman, M.M., Vanlauwe, B., de Goede, R.G.M., Six, J., Csuzdi, C., Brussaard, L., 2011. Agricultural management affects earthworm and termite diversity across humid to semi-arid tropical zones. Agric. Ecosyst. Environ. 140 (1), 148–154. Barham, B.L., Chavas, J.P., Fitz, D., Salas, V.R., Schechter, L., 2014. The roles of risk and ambiguity in technology adoption. J. Econ. Behav. Organ. 97, 204–218. Bellon, M.R., 2004. Conceptualizing interventions to support on-farm genetic resource conservation. World Dev. 32 (1), 159–172. Benjamin, D., 1992. Household composition, labor markets, and labor demand: testing for separation in agricultural household models. Econometrica: J. Econ. Soc. 287–322. Benin, S., Nkonya, E., Okecho, G., Randriamamonjy, J., Kato, E., Lubade, G., Kyotalimye, M., 2012. Impact of the National Agricultural Advisory Services (AADS) program of Uganda: considering different levels of likely contamination with the treatment. Am. J. Agric. Econ. 94 (2), 386–392. Block, S.A., 2013. The post-independence decline and rise of crop productivity in subSaharan Africa: measurement and explanations. Oxf. Econ. Pap., gpt010 Blüthgen, N., Dormann, C.F., Prati, D., Klaus, V.H., Kleinebecker, T., Hölzel, N., ... Weisser, W.W., 2012. A quantitative index of land-use intensity in grasslands: integrating mowing, grazing and fertilization. Basic Appl. Ecol. 13 (3), 207–220. Billeter, R., Liira, J., Bailey, D., Bugter, R., Arens, P., Augenstein, I., ... Edwards, P.J., 2008. Indicators for biodiversity in agricultural landscapes: a pan-European study. J. Appl. Ecol. 45 (1), 141–150. Brugnach, M., Dewulf, A.R.P.J., Henriksen, H.J., Van der Keur, P., 2011. More is not always better: coping with ambiguity in natural resources management. J. Environ. Manag. 92 (1), 78–84. Bryan, E., Ringler, C., Okoba, B., Roncoli, C., Silvestri, S., Herrero, M., 2013. Adapting agriculture to climate change in Kenya: household strategies and determinants. J. Environ. Manag. 114, 26–35. Cavatassi, R., Lipper, L., Narloch, U., 2011. Modern variety adoption and risk management in drought prone areas: insights from the sorghum farmers of eastern Ethiopia. Agric. Econ. 42 (3), 279–292. Carpentier, A., Letort, E., 2011. Accounting for heterogeneity in multicrop microeconometric models: implications for variable input demand modeling. Am. J. Agric. Econ., aar132 Chavas, J.P., Di Falco, S., 2012. On the role of risk versus economies of scope in farm diversification with an application to Ethiopian farms. Journal of Agricultural Economics 63 (1), 25–55. Collier, P., Dercon, S., 2013. African agriculture in 50 years: smallholders in a rapidly changing world? World Dev. 63, 92–101.

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358 Conley, T.G., Udry, C.R., 2010. Learning about a new technology: pineapple in Ghana. Am. Econ. Rev. 35-69. Conway, G.R., Barbier, E.B., 2013. After the Green Revolution: Sustainable Agriculture for Development. Routledge. Croppenstedt, A., Demeke, M., Meschi, M.M., 2003. Technology adoption in the presence of constraints: the case of fertilizer demand in Ethiopia. Rev. Dev. Econ. 7 (1), 58–70. Dercon, S., Christiaensen, L., 2011. Consumption risk, technology adoption and poverty traps: evidence from Ethiopia. J. Dev. Econ. 96 (2), 159–173. Dercon, S., Gollin, D., 2014. Agriculture in African development: a review of trends and strategies. Ann. Rev. Resour. Econ. 6 (1). Di Falco, S., Bulte, E., 2013. The impact of kinship networks on the adoption of riskmitigating strategies in Ethiopia. World Dev. 43, 100–110. Di Falco, S., Chavas, J.P., 2009. On crop biodiversity, risk exposure, and food security in the highlands of Ethiopia. Am. J. Agric. Econ. 91 (3), 599–611. Di Falco, S., Perrings, C., 2005. Crop biodiversity, risk management and the implications of agricultural assistance. Ecol. Econ. 55 (4), 459–466. Di Falco, S., Bezabih, M., Yesuf, M., 2010. Seeds for livelihood: crop biodiversity and food production in Ethiopia. Ecol. Econ. 69 (8), 1695–1702. Di Falco, S., Chavas, J.P., Smale, M., 2007. Farmer management of production risk on degraded lands: the role of wheat variety diversity in the Tigray region, Ethiopia. Agric. Econ. 36 (2), 147–156. Di Falco, S., Veronesi, M., 2013. How can African agriculture adapt to climate change? A counterfactual analysis from Ethiopia. Land Econ. 89 (4), 743–766. Di Falco, S., Veronesi, M., Yesuf, M., 2011. Does adaptation to climate change provide food security? A micro-perspective from Ethiopia. Am. J. Agric. Econ. 93 (3), 829–846. Doss, C.R., 2006. Analyzing technology adoption using microstudies: limitations, challenges, and opportunities for improvement. Agric. Econ. 34 (3), 207–219. Duelli, P., Obrist, M.K., 2003. Biodiversity indicators: the choice of values and measures. Agric. Ecosyst. Environ. 98 (1), 87–98. Duflo, E., Kremer, M., Robinson, J., 2008. How high are rates of return to fertilizer? Evidence from field experiments in Kenya. Am. Econ. Rev. 98 (2), 482–488. Evenson, R.E., Gollin, D., 2003. Assessing the impact of the green revolution, 1960 to 2000. Science 300 (5620), 758–762. FAO, 2010. The Second Report on the State of the World's Plant and Genetic Resources for Food and Agriculture, Rome. 2010. FAO, 2012. Undernourishment Around the World in 2012. FAO, Rome. FAO, 1978-81. Reports of the Agro-ecological Zones Project. World Soil Resources Report No. 48. FAO, Rome. FAOSTAT, 2014. Food and Agriculture Organization of the United Nations. Statistics division http://faostat.fao.org/. Feder, G., 1980. Farm size, risk aversion and the adoption of new technology under uncertainty. Oxf. Econ. Pap. 263–283. Flohre, A., et al., 2011. Agricultural intensification and biodiversity partitioning in European landscapes comparing plants, carabids, and birds. Ecol. Appl. 21 (5), 1772–1781. Foley, J.A., Ramankutty, N., Brauman, K.A., Cassidy, E.S., Gerber, J.S., Johnston, M., ... Zaks, D.P., 2011. Solutions for a cultivated planet. Nature 478 (7369), 337–342. Frisvold, G., Ingram, K., 1995. Sources of agricultural productivity growth and stagnation in sub-Saharan Africa. Agricultural Economics 13 (1), 51–61. Fuglie, K.O., Bosch, D.J., 1995. Economic and environmental implications of soil nitrogen testing: a switching-regression analysis. Am. J. Agric. Econ. 77 (4), 891–900. Giller, K.E., Witter, E., Corbeels, M., Tittonell, P., 2009. Conservation agriculture and smallholder farming in Africa: the heretics' view. Field Crop Res. 114, 23–34. Greene, W., 2008. Econometric Analysis (6th Ed.). Prentice Hall, Englewood Cliffs, NJ. Harlan, J.R., 1972. Genetics of disaster. J. Environ. Qual. 1, 212–215. Hartter, J., Stampone, M.D., Ryan, S.J., Kirner, K., Chapman, C.A., Goldman, A., 2012. Patterns and perceptions of climate change in a biodiversity conservation hotspot. PLoS One 7 (2), e32408. Hausman, J.A., 1978. Specification tests in econometrics. Econometrica 46, 1251–1272. Heckman, J., 1979. Sample selection as a specification error. Econometrica 47, 153–161. Hendrickx, F., Maelfait, J.P., Van Wingerden, W., Schweiger, O., Speelmans, M., Aviron, S., ... Bugter, R.O.B., 2007. How landscape structure, land-use intensity and habitat diversity affect components of total arthropod diversity in agricultural landscapes. J. Appl. Ecol. 44 (2), 340–351. Herzog, F., Steiner, B., Bailey, D., Baudry, J., Billeter, R., Bukácek, R., ... Bugter, R., 2006. Assessing the intensity of temperate European agriculture at the landscape scale. Eur. J. Agron. 24 (2), 165–181. Isik, M., Khanna, M., 2003. Stochastic technology, risk preferences, and adoption of sitespecific technologies. Am. J. Agric. Econ. 85 (2), 305–317. Jarvis, D.I., Brown, A.H., Cuong, P.H., Collado-Panduro, L., Latournerie-Moreno, L., Gyawali, S., ... Hodgkin, T., 2008. A global perspective of the richness and evenness of traditional crop-variety diversity maintained by farming communities. Proc. Natl. Acad. Sci. 105 (14), 5326–5331. Johnson, M., Hazell, P., Gulati, A., 2003. The role of intermediate factor markets in Asia's green revolution: lessons for Africa? Am. J. Agric. Econ. 85 (5), 1211–1216. Jones, A.D., Shrinivas, A., Bezner-Kerr, R., 2014. Farm production diversity is associated with greater household dietary diversity in Malawi: findings from nationally representative data. Food Policy 46, 1–12. Kabunga, N.S., Dubois, T., Qaim, M., 2014. Impact of tissue culture banana technology on farm household income and food security in Kenya. Food Policy 45, 25–34. Kassie, M., Shiferaw, B., Muricho, G., 2011. Agricultural technology, crop income, and poverty alleviation in Uganda. World Dev. 39 (10), 1784–1795. Kijima, Y., Otsuka, K., Sserunkuuma, D., 2011. An inquiry into constraints on a green revolution in Sub-Saharan Africa: the case of NERICA rice in Uganda. World Dev. 39 (1), 77–86. Lipper, L., Cooper, D., 2009. Managing plant genetic resources for sustainable use in food and agriculture. Balancing the benefits in the field. In: Kontoleon, A., Pascual, U.,

357

Smale, M. (Eds.), Agrobiodiversity Conservation and Economic Development. Routledge, New York, pp. 27–39. Liverpool, L.S.O., Winter-Nelson, A., 2010. Poverty status and the impact of formal credit on technology use and wellbeing among Ethiopian smallholders. World Dev. 38 (4), 541–554. Lokshin, M., Sajaia, Z., 2004. Maximum likelihood estimation of endogenous switching regression models. Stata J. 4, 282–289. Maddala, G.S., 1983. Limited-dependent and Qualitative Variables in Econometrics (No. 3). Cambridge University Press. Maddala, G.S., Nelson, F.D., 1975. Switching regression models with exogenous and endogenous switching. Proceeding of the American Statistical Association (Business and Economics Section), pp. 423–426. Mäder, P., Fliessbach, A., Dubois, D., Gunst, L., Fried, P., Niggli, U., 2002. Soil fertility and biodiversity in organic farming. Science 296 (5573), 1694–1697. Mercer, K.L., Perales, H.R., 2010. Evolutionary response of landraces to climate change in centers of crop diversity. Evol. Appl. 3 (5–6), 480–493. Ministry of Agriculture, Animal Industry and Fisheries and Ministry of Finance (MAAIF), Planning and Economic Development of Uganda, 2000i. Plan for modernization of agriculture. Eradicating Poverty in Uganda Government Strategy and Operational Framework. Minten, B., Stifel, D., Tamru, S., 2014. Structural transformation of cereal markets in Ethiopia. J. Dev. Stud. 1–19. Mouysset, L., Doyen, L., Jiguet, F., 2012. Different policy scenarios to promote various targets of biodiversity. Ecol. Indic. 14 (1), 209–221. Mundlak, Y., 1978. On the pooling of time series and cross-section data. Econometrica 46 (1), 69–85. NAADS, National Agriculture Advisory Service, 2001. The National Agricultural Advisory Services Act. Narayanan, S., 2014. Profits from participation in high value agriculture: evidence of heterogeneous benefits in contract farming schemes in Southern India. Food Policy 44, 142–157. Narloch, U., Drucker, A.G., Pascual, U., 2011. Payments for agrobiodiversity conservation services for sustained on-farm utilization of plant and animal genetic resources. Ecol. Econ. 70 (11), 1837–1845. Noltze, M., Schwarze, S., Qaim, M., 2013. Impacts of natural resource management technologies on agricultural yield and household income: the system of rice intensification in Timor Leste. Ecol. Econ. 85, 59–68. Omer, A., Pascual, U., Russell, N., 2010. A theoretical model of agrobiodiversity as a supporting service for sustainable agricultural intensification. Ecol. Econ. 69 (10), 1926–1933. Otsuka, K., Larson, D.F., 2013. An African Green Revolution: Finding Ways to Boost Productivity on Small Farms. Springer. Pamuk, H., Bulte, E., Adekunle, A.A., 2014. Do decentralized innovation systems promote agricultural technology adoption? Experimental evidence from Africa. Food Policy 44, 227–236. Pascual, U., Perrings, C., 2007. Developing incentives and economic mechanisms for in situ biodiversity conservation in agricultural landscapes. Agric. Ecosyst. Environ. 121 (3), 256–268. Perrings, C., Jackson, L., Bawa, K., Brussaard, L., Brush, S., Gavin, T., ... De Ruiter, P., 2006. Biodiversity in agricultural landscapes: saving natural capital without losing interest. Conserv. Biol. 20 (2), 263–264. Pingali, P.L., 2012. Green revolution: impacts, limits, and the path ahead. Proc. Natl. Acad. Sci. 109 (31), 12302–12308. Place, F., Barrett, C.B., Freeman, H.A., Ramisch, J.J., Vanlauwe, B., 2003. Prospects for integrated soil fertility management using organic and inorganic inputs: evidence from smallholder African agricultural systems. Food Policy 28 (4), 365–378. Quaas, M.F., Baumgärtner, S., 2008. Natural vs. financial insurance in the management of public-good ecosystems. Ecol. Econ. 65 (2), 397–406. Ribeiro Palacios, M., Huber-Sannwald, E., García Barrios, L., Peña de Paz, F., Carrera Hernández, J., Galindo Mendoza, M.D.G., 2013. Landscape diversity in a rural territory: emerging land use mosaics coupled to livelihood diversification. Land Use Policy 30 (1), 814–824. Richardson, R.B., Fernandez, A., Tschirley, D., Tembo, G., 2012. Wildlife conservation in Zambia: impacts on rural household welfare. World Dev. 40 (5), 1068–1081. Rossi, J.P., Celini, L., Mora, P., Mathieu, J., Lapied, E., Nahmani, J., ... Lavelle, P., 2010. Decreasing fallow duration in tropical slash-and-burn agriculture alters soil macroinvertebrate diversity: a case study in southern French Guiana. Agric. Ecosyst. Environ. 135 (1), 148–154. Smale, M. (Ed.), 2005. Valuing Crop Biodiversity: On-farm Genetic Resources and Economic Change. Cabi. Shiferaw, B., Kassie, M., Jaleta, M., Yirga, C., 2014. Adoption of improved wheat varieties and impacts on household food security in Ethiopia. Food Policy 44, 272–284. Sserunkuuma, D., 2005. The adoption and impact of improved maize and land management technologies in Uganda. Electr. J. Agric. Dev. Econ. 2 (1), 67–84. Suri, T., 2011. Selection and comparative advantage in technology adoption. Econometrica 79 (1), 159–209. Takeshima, H., Pratt, A.N., Diao, X., 2013. Mechanization and agricultural technology evolution, agricultural intensification in Sub-Saharan Africa: typology of agricultural mechanization in Nigeria. Am. J. Agric. Econ., aat045 Teklewold, H., Kassie, M., Shiferaw, B., Köhlin, G., 2013. Cropping system diversification, conservation tillage and modern seed adoption in Ethiopia: impacts on household income, agrochemical use and demand for labor. Ecol. Econ. 93, 85–93. Tilman, D., Cassman, K.G., Matson, P.A., Naylor, R., Polasky, S., 2002. Agricultural sustainability and intensive production practices. Nature 418 (6898), 671–677. Tittonell, P., Giller, K.E., 2013. When yield gaps are poverty traps: the paradigm of ecological intensification in African smallholder agriculture. Field Crop Res. 143, 76–90.

358

M. Coromaldi et al. / Ecological Economics 119 (2015) 346–358

Tripp, R., Rohrbach, D., 2001. Policies for African seed enterprise development. Food Policy 26 (2), 147–161. Tumuhairwe, J.K., Nkwiine, C., Kawangolo, J.B., Tumuhairwe, F.K., 2003. Developing policy and technical recommendations for the conservation of agricultural biodiversity in Uganda. In: Kaihura, F., Stocking, M. (Eds.), Agricultural Biodiversity in Smallholder Farms of East Africa. United Nations Univ. Press.

Weitzman, M.L., 2000. Economic profitability vs. ecological entropy. Q. J. Econ. 115 (1), 237–263. World Bank, 2007. World Development Report 2008. Agriculture for Development, Washington, DC. World Bank Group (Ed.), 2013. World development indicators 2013. World Bank Publications.