A dynamic adoption model with Bayesian learning: an application to U.S. soybean farmers

Adoption of agricultural technology is often sequential, with farmers first adopting a new technology on part of their lands and then adjusting their use of the new technology in later years based on what was learned from the initial partial adoption. Our article explains this experimental behavior by using a dynamic adoption model with Bayesian learning, in which forward‐looking farmers take account of future impacts of their learning from both their own and their neighbors’ experiences with the new technology. We apply the analysis to a panel of U.S. soybean farmers surveyed from 2000 to 2004 to examine their adoption of the genetically modified (GM) seed technology. We compare the results of the forward‐looking model to that of a myopic model, in which farmers maximize current benefits only. Results suggest that the forward‐looking model fits data better than the myopic model does. And potential estimation biases arise when fitting a myopic model to forward‐looking decision makers.     

Spatial dependency and technical efficiency: an application of a Bayesian stochastic frontier model to irrigated and rainfed rice farmers in Bohol,Philippines

We investigated the role of spatial dependency in the technical efficiency estimates of rice farmers using panel data from the Central Visayan island of Bohol in the Philippines. Household‐level data were collected from irrigated and rainfed agro‐ecosystems. In each ecosystem, the geographical information on residential and farm‐plot neighborhood structures was recorded to compare household‐level spatial dependency among four types of neighborhoods. A Bayesian stochastic frontier approach that integrates spatial dependency was used to address the effects of neighborhood structures on farmers’ performance. Incorporating the spatial dimension into the neighborhood structures allowed for identification of the relationships between spatial dependency and technical efficiency through comparison with nonspatial models. The neighborhood structure at the residence and plot levels were defined with a spatial weight matrix where cut‐off distances ranged from 100 to 1,000 m. We found that spatial dependency exists at the residential and plot levels and is stronger for irrigated farms than rainfed farms. We also found that technical inefficiency levels decrease as spatial effects are more taken into account. Because the spatial effects increase with a shorter network distance, the decreasing technical inefficiency implies that the unobserved inefficiencies can be explained better by considering small networks of relatively close farmers over large networks of distant farmers.     

Factors influencing hybrid rice adoption: a Bangladesh case

As rice constitutes the major share in cereal consumption in South and East Asian countries that ranges from as low as 40 per cent in India to 97 per cent in Myanmar, to ensure food security, governments in these countries are encouraging farmers to adopt hybrid rice. This is mainly because hybrid rice provides a yield gain of 15–20 per cent over conventionally bred varieties in general. Yet, despite strenuous government efforts, farmers’ adoption rates have remained low in India, Bangladesh and Vietnam compared with China. Although studies often claim that higher seed costs and inferior grain quality are the major factors limiting hybrid rice adoption, very few studies examine the importance of socio‐economic factors and infrastructure in the adoption of hybrid rice. Using Bangladesh as a case, a comparative analysis has been made on the adoption of hybrid and modern varieties relative to traditional rice varieties and land allocation to these varieties. Econometric results indicate that general land characteristics, loan facilities and general infrastructure, such as roads, irrigation facilities and the availability of government‐approved seed dealers, significantly influence the adoption of hybrid and modern rice varieties and land allocation to these varieties compared with traditional varieties.     

Capturing social network effects in technology adoption: the spatial diffusion of hybrid rice in Bangladesh

In this paper, we demonstrate a method for measuring the effect of spatial interactions on the use of hybrid rice using a unique, nationally representative data set from Bangladesh. In order to circumvent the ‘reflection problem’, we consider an identification and estimation strategy employing a generalised spatial two‐stage least squares procedure with near‐ideal instruments to effectively identify causal influences. Results indicate that neighbour effects are a significant determinant of hybrid rice use. Further, using two specifications of spatial network systems, one based on same‐village membership (irrespective of distance) and the other based on geographical distance (irrespective of village boundary), we demonstrate that a network including nearby hybrid rice adopters is more influential than a network of more distant hybrid rice adopters, and merely having a network with a large number of adopters may be relatively meaningless if they are far away. Furthermore, we show that these network effects are much more important to hybrid cultivation than interactions with agricultural extension officers.     

Taking a new look at empirical models of adoption: average treatment effect estimation of adoption rates and their determinants

This article shows that the observed sample adoption rate does not consistently estimate the population adoption rate even if the sample is random. It is proved that instead the sample adoption rate is a consistent estimate of the population joint exposure and adoption rate, which does not inform about adoption per se. Likewise, it is shown that a model of adoption with observed adoption outcome as a dependent variable and where exposure to the technology is not observed and controlled for cannot yield consistent estimates of the determinants of adoption. The article uses the counterfactual outcomes framework to show that the true population adoption rate corresponds to what is defined in the modern policy evaluation literature as the average treatment effect (ATE), which measures the effect or impact of a "treatment" on a person randomly selected in the population. In the adoption context, a "treatment" corresponds to exposure to the technology. The article uses the ATE estimation framework to derive consistent nonparametric and parametric estimators of population adoption rates and their determinants and applies the results to consistently estimate the population adoption rates and determinants of the NERICA (New Rice for Africa) rice varieties in Côte d'Ivoire. The ATE methodological approach developed in the article has significant policy implications with respect to judging the intrinsic merit of a new technology in terms of its potential demand by the target population independently of issues related to its accessibility and in terms of the decision to invest or not in its wide-scale dissemination.     

Poverty analysis using small area estimation: an application to conservation agriculture in Uganda

This article demonstrates the utility of small area estimation of poverty (SAEp) methods for researchers wishing to conduct a detailed welfare analysis as part of a larger survey of a small geographic area. This study applies SAEp methods as part of an impact assessment of a conservation agriculture production system in Eastern Uganda. Using SAEp, we estimate Foster–Greer–Thorbecke rural poverty indices, estimate the effects of per‐acre farm profit increases to poor households on the indices, and compare the findings to estimates of net returns from a field‐level evaluation of conservation agriculture for maize farmers. Results suggest that increasing the farm profits of the bottom 30% of households by $1.60 per‐acre per‐season would reduce rural poverty incidence by 1 percentage point. Available data on the net returns to conservation agriculture indicate that even these modest increases are achievable for few adopting households.     

Econometric mathematical programming: an application to the estimation of costs and risk preferences at farm level

This study belongs to the barely explored research strand of “Econometric Mathematical Programming” and presents a simultaneous estimation of the cost function and of the farmers’ risk attitude parameter in a programming model setup. Resource and policy constraints of the model are allowed to be not binding. We use crop shares as decision variables to avoid scale bias and we consider price and crop yield variances separately. The model is formulated as a bi‐level programming model and the empirical application concerns three unbalanced panels of specialized arable farms observed for at least three consecutive years in Northern Italy, in the Cologne‐Aachen area in Germany and in the Grandes‐Cultures area in France over the time period 1995–2007. We achieve a quite satisfactory fit in the estimation exercise and find own and cross price elasticities from sensitivity experiments in reasonable ranges. We also propose a novel approach to derive confidence intervals around parameter estimates for Econometric Mathematical Programming.     

Cooperative membership and adoption of green pest control practices: Insights from rice farmers

Reducing the use of chemical pesticide while preserving crop yield is a practical strategy that makes agricultural production economically, socially and environmentally sustainable. Although the adoption of green pest control practices can help achieve such a goal, its adoption rate remains quite low. This study explores whether membership in agricultural cooperatives improves smallholder farmers' adoption of green pest control practices, utilising farm-level data surveyed from rice farmers in China. To enrich our understanding, we also investigate how the adoption of green pest control practices mediates the effects of cooperative membership on chemical pesticide expenditures. An endogenous switching probit model and a bootstrap-based mediation method are employed to achieve these goals. The empirical results show that cooperative membership significantly increases the probability of adopting physical pest control practices (e.g. pest-killing lamps or sticky plate traps) and biological pest control practices (e.g. biopesticides) by 6% and 19%, respectively. Cooperative membership significantly reduces chemical pesticide expenditures through its mediation effect on improving the probability of adopting biological pest control practices. There is also a complementary relationship between physical and biological pest control practices in pest management of rice production. The adoption of physical pest control practices significantly increases rice yield, while the adoption of biological pest control practices does not.     

Role of education and extension in the adoption of technology: a study of upland rice and soybean farmers in Central-West Brazil

This paper explores reduced form determinants of the adoption of certain technologies by upland rice and soybean farmers in the Center-West region of Brazil. We merge community level data on the availability and quality of publicly provided infrastructure, principally extension, to the farm level data containing information on farmer human capital as well as land quantity and quality. By using community level measures of availability and quality of extension, we avoid problems of endogeneity of farm level measures of extension use. We find positive impacts of farmer education on the diffusion process, in accordance with other studies. We also isolate effects of the quality in regional extension investment as measured by the average experience of technical extension staff. These results indicate that investments in human capital of extension workers does have a payoff in terms of farmer adoption of improved cultivation practices.     

Joint estimation using revealed and stated preference data: An application using a national forest

We combine contingent behaviour with travel cost data to estimate the change in the recreational use value of a National Forest due to quality and price changes. Instead of the usual improvement scenario, a hypothetical deterioration in the conditions of the forest due to a fire is considered. A dataset containing five observations for each respondent enabled the estimation of three models for which the number of scenarios differed. The results show that visitors are sensitive to price and quality changes and that in the forest fire scenario the intended number of trips would be reduced and that respondents would experience a welfare loss. Signs of inconsistency between preferences expressed by revealed and intended behaviour were found. This research also provides some indications that strategic bias affects answers to price changes.     

The determinants of technology adoption by UK farmers using Bayesian model averaging: the cases of organic production and computer usage

We review and implement a reversible jump approach to Bayesian model averaging for the Probit model with uncertain regressors. Two applications are investigated. The first is the adoption of organic systems in UK farming, and the second is the influence of farm and farmer characteristics on the use of a computer on the farm. While there is a correspondence between the conclusions we would obtain with and without model averaging results, we find important differences, particularly in smaller samples. Concerning the adoption of an organic system, we find that attitudes to the sustainability of the current system along with the ability of organic farms alone to satisfy society’s needs for food are influential. Additionally, the source of management information used by the farmer has a significant impact. Regarding the adoption of computers, we confirm the findings of previous work that the level of education affects uptake and that age is a factor determining adoption. We also find that dairy and organic farms are more likely to use a computer. The physical size of the farm is positively associated with the probability of computer use while net farm income has a limited impact.     

Location and uptake: integrated household and GIS analysis of technology adoption and land use, with application to smallholder dairy farms in Kenya

CIS‐derived measures of location and space have increasingly been used in models of land use and ecology. However, they have made few inroads into the literature on technology adoption in developing countries, which continues to rely mainly on survey‐derived information. Location, with all its dimensions of market access, demographics and agro‐climate, nevertheless remains key to understanding potential for technology use. The measures of location typically used in the adoption literature, such as locational dummy variables that proxy a range of locational factors, now appear relatively crude given the increased availability of more explicit GIs‐derived measures. This paper attempts to demonstrate the usefulness of integrating CIS‐measures into analysis of technology uptake, for better differentiating and understanding locational effects. A set of GIs‐derived measures of market access and agro‐climate are included in a standard household model of technology uptake, applied to smallholder dairy farms in Kenya, using a sample of 3330 geo‐referenced farm households. The three technologies examined are keeping of dairy cattle, planting of specialised fodder, and use of concentrate feed. Logit estimations are conducted that significantly differentiate effects of individual household characteristics from those related to location. The predicted values of the locational variables are then used to make spatial predictions of technology potential. Comparisons are made with estimations based only on survey data, which demonstrate that while overall explanatory power may not improve with CIS‐derived variables, the latter yield more practical interpretations, which is further demonstrated through predictions of technology uptake change with a shift in infrastructure policy. Although requiring large geo‐referenced data sets and high resolution GIS layers, the methodology demonstrates the potential to better unravel the multiple effects of location on farmer decisions on technology and land use.