Search results

Bayesian additive regression trees (BART) is a fully Bayesian approach to modeling with ensembles of trees. BART can uncover complex regression functions with high-dimensional regressors in a fairly automatic way and provide Bayesian quantification of the uncertainty through the posterior. However, BART assumes independent and identical distributed (i.i.d) normal errors. This strong parametric assumption can lead to misleading inference and uncertainty quantification. In this chapter we use the classic Dirichlet process mixture (DPM) mechanism to nonparametrically model the error distribution. A key strength of BART is that default prior settings work reasonably well in a variety of problems. The challenge in extending BART is to choose the parameters of the DPM so that the strengths of the standard BART approach is not lost when the errors are close to normal, but the DPM has the ability to adapt to non-normal errors.

The chapter contributes to on-going debates about the inclusion of smallholders in export value chains for high-value agricultural products. Specifically, it investigates the factors driving the procurement practices of exporter of fresh fruits and vegetables in sub-Saharan Africa, and specifically sourcing from smallholders. A survey is undertaken of exporters of fresh fruit and vegetables in sub-Saharan Africa. The resulting data are used to estimate econometrically the propensity of exporters to source from smallholders, and the intensity of sourcing among those exporters who do procure from smallholders. Explanatory variables include firm and market characteristics, supply chain costs, type of product, availability of alternative sources of supply, and judgments regarding the performance of smallholders and other sources of supply.The propensity to procure from smallholders is found to be negatively associated with being a small exporter and the performance of medium- and large-scale producers. Exporters are more likely to source from smallholders if they have their own production capacity and smallholders are judged to perform well. The requirement of customers to comply with private food safety standards is found to have no significant effect on the propensity to procure from smallholders. Conversely, compliance with private standards has a strong influence on the intensity of sourcing from smallholders. Exporters judging smallholders to perform well are more likely to source intensively from smallholders, but to source less if they judge their own production to perform well. High fixed costs tend to be associated with lower intensity of sourcing from smallholders. The results suggest that compliance with private food safety standards does not drive the exclusion of smallholders from export value chains; indeed, conversely, the requirement to comply with such standards is associated with greater intensity of sourcing from smallholders. Smallholders evidently play a key role in the defrayment of risk by exporters in that many exporters combine their own production with smallholder procurement. Costs of procurement from smallholders, however, remain an issue. Evidently, the fixed costs of smallholder supply chains increase appreciably with the intensity of sourcing. The research reported here provides a new perspective on the inclusion of smallholders in export value chains for horticultural products. The incorporation of smallholders into these value chains is seen as the outcome of the procurement decisions of exporters. Contrary to much of the discourse in this area, the results suggest that smallholders can and do compete in export value chains for horticultural products even in the context of exacting food safety standards.

Machine learning (ML) has recently gained momentum as a method for measurement in strategy research. Yet, little guidance exists regarding how to appropriately apply the method for this purpose in our discipline. We address this by offering a guide to the application of ML in strategy research, with a particular emphasis on data handling practices that should improve our ability to accurately measure our constructs of interest using ML techniques. We offer a brief overview of ML methodologies that can be used for measurement before describing key challenges that exist when applying those methods for this purpose in strategy research (i.e., sample sizes, data noise, and construct complexity). We then outline a theory-driven approach to help scholars overcome these challenges and improve data handling and the subsequent application of ML techniques in strategy research. We demonstrate the efficacy of our approach by applying it to create a linguistic measure of CEOs' motivational needs in a sample of S&P 500 firms. We conclude by describing steps scholars can take after creating ML-based measures to continue to improve the application of ML in strategy research.

Risk preference theory posits that females are more religious than males because they are more risk averse and are thus more motivated by the threat of afterlife punishment. We evaluate the theory formally and empirically. Formally, we show that the rational choice reasoning implied by the theory leads to unexpected conclusions if one considers belief in eternal rewards as well as eternal punishment. Empirically, we examine cross-cultural data and find that, across many populations, sex differences in religiosity are no smaller among those who do not believe in hell. We conclude by arguing that psychological characteristics are almost certainly crucial to understanding the difference, just not risk preference.

We investigate whether giving workers autonomy through delegation of contract choice intrinsically motivates effort. In a novel laboratory experiment that controls for contract preferences and outcomes, principals can either choose the contract under which agents work on a real-effort task, or delegate the contract choice to the agents. We evaluate whether agents exert higher effort when they are allowed to choose the contract versus when the contract is imposed on them. We find no difference between the two conditions, even after controlling for baseline ability and for locus of control. Because our design excludes the possibility that preferences play a role, and because workers engaged in a real-effort task, this result casts doubt on an intrinsic link between the autonomy granted through delegation and the motivation of employees in the workplace. Our results do not deny, however, the possible instrumental benefits of autonomy (which did not play a role in our design) and their potentially powerful impact on motivation.

Internet of Things (IoT) and artificial intelligence are two leading technologies that bought revolution to each and every field of humans using in daily life by making everything smarter than ever. IoT leads to a network of things which creates a self-configuring network. Improving farm productivity is essential to meet the rapidly growing demand for food. In this chapter, the authors have introduced a smart greenhouse by integration of two leading technologies in the market (i.e., Machine Learning and IoT). In proposed model, several sensors are used for data collection and managing the environment of greenhouse. The idea is to propose an IoT and Machine Learning based smart nursery that helps in healthy growing and monitoring of the seed. The structure will be a dome-like structure for observation and isolation of an egg with various sensors like pressure, humidity, temperature, light, moisture, conductivity, air quality, etc. to monitor the nursery internal environment and maintain the control and flow of water and other minerals inside the nursery. The nursery will have a solar panel from which it stores the electricity generated from the sun, a small fan to control the flow of air and pressure. A camera will also be equipped inside the nursery that will use computer vision technology to monitor the health of the plant and will be trained on the past data to notify the user if the plant is diseased or need attention.

The chapter studies strategic default using an experimental approach.

The experiment considers a stochastic asset process and a loan with no down-payment. The treatments are two asset volatilities (high and low) and the absence and presence of social interactions via a direct effect on the subject's payoff.

I demonstrate that (i) people appear to follow the prediction of the strategic default model quite closely in the high asset volatility treatment, and that (ii) incorporating social interactions delays the strategic default beyond what is considered optimal.

The study tests adequately the strategic default using a novel experimental design and analyzes the neighbor's effect on that decision.