Sample Selection Bias Research Articles

Uplift modeling and its implications for appointment date prediction in attended home delivery

Successful attended home delivery (AHD) is the most important aspect of e-commerce order fulfillment. Prior literature focuses on incentive scheme development for customers' choices of delivery windows and predictive analytics for delivery results, but it is not clear whether the effect of AHD on the appointment date set by customers increases the success rate of AHD. Therefore, we developed an uplift modeling method, PSM-NDML, as a relevant prescriptive analytic tool for AHD on an appointment date, which aims to estimate the causal effect of the by-appointment delivery on the delivery result. PSM-NDML integrates propensity score matching and double machine learning, effectively addressing sample selection bias, low predictive performance, and poor interpretability. Applied to a real-world product delivery dataset of a Chinese logistics company, PSM-NDML achieves superior performance relative to ten other state-of-the-art uplift models in terms of cumulative gain and the Qini coefficient. The predicted responses gained from PSM-NDML are also visually interpreted at the global and local levels, which reveals various managerial insights. In practice, the findings expand managers' understanding of the heterogeneous effects of AHD on appointment dates and provide decision support for logistics companies in the development of home delivery plans.

Linguistic steganalysis via multi-task with crossing generative-natural domain

In the real world, the sources of text often transcend the boundaries of generative and natural domains, which introduces challenges for existing linguistic steganalysis methods. The typical problem arises from the sample selection bias caused by training solely on a single-source domain, rendering the model incapable of inferencing across the entire generative-natural (GN) space. Moreover, there exists the problem of overlooking the sensitive discrepancies between generative text and natural text, which hampers model fitting. In this paper, we model steganalysis in a brand-new perspective by employing multi-task learning to build the main task and auxiliary tasks in the cross GN domain. The proposed Cross Generative-Natural Domain Multi-task Model (CG-NDMM) can concurrently address the two aforementioned issues through i) modeling steganalysis across the entire GN space, incorporating two auxiliary tasks alongside a main task, and ii) utilizing a feature representation transfer learning strategy to harmonize two sub-networks. Furthermore, we employ diverse steganography algorithms to construct the datasets, which comprise four types of texts (generative-cover, generative-steganographic, natural-cover, and natural-steganographic) derived from two public datasets, Movie and Twitter. The experiments on these datasets demonstrate the effectiveness of the proposed approach, showcasing its substantially superior performance over the comparative baseline methods.

Climate governance and carbon risk in the global energy sector: Insights into corporate environmental initiatives

This study examines the influence of sustainable governance, specifically climate governance, on carbon risk within the global energy sector. Additionally, we investigate the role of eco-innovation as a mediating factor in this relationship. By analyzing a dataset comprising 13,376 publicly listed energy companies from 91 different countries and employing Baron and Kenny's (1986) four-step mediation model, our research shows that improved climate governance mechanisms result in decreased carbon emissions from energy firms. This reduction can be primarily attributed to their increased participation in eco-innovation initiatives. Furthermore, these main findings are more pronounced in companies with robust environmental, social, and governance (ESG) practices. Our results also reveal various firm-level and country-level characteristics that moderate our identified relationship. Moreover, our results remain consistent even after addressing potential concerns related to endogeneity and sample selection bias. This research provides valuable insights for policymakers and managers who seek to mitigate carbon emissions within the global energy sector while fostering environmentally responsible practices to combat the impacts of climate change.

R&D Subsidies in Spain: are they Really Useful?

AbstractThis article is aimed at estimating the impact of R&D subsidies on the demand for researchers, using information provided by a panel of Spanish firms belonging to the industrial and services sectors followed up during the 2004–2016 period. Estimates include corrections for the endogeneity bias generated by wages and by a set of variables measuring R&D public financing, as well as for the sample selection bias coming from the fact that the only wage data available are those from firms that are internal R&D performers and hire researchers. Estimate results show that public R&D financing has a positive and relevant impact on the demand for researchers. This result allows supporting the implementation of policies to promote public research funding in Spain, regardless of the origin of the funds received.

PERCEPTIONS OF CLIMATE CHANGE AND THEIR DETERMINANTS AMONG THE SMALLHOLDER FARMERS IN MACHAKOS COUNTY

Climate change is expected to affect the livelihoods of smallholder farmers in Machakos County due to their overdependence on rain-fed agriculture. Farmers perception of climate change and variability influences their understanding and management of climate change-induced risks and uncertainties, especially in rain-fed agriculture. The study was carried out to evaluate smallholder farmers awareness and perception of climate change and the factors influencing the perception of climate change in Machakos County. Data was collected from 400 farmers from six locations sampled through a multistage and simple random sampling procedure. The Heckman probit model was fitted to the data to avoid sample selection bias since not every farmer who may perceive climate change responds by adapting. The analysis revealed that 96% of smallholder farmers in Machakos County were aware of climate change, 87.3% perceived that temperature was rising, and 96.8% perceived that rainfall was decreasing. In this regard, the age of the household head, Gender, education, household size, group membership, distance to the nearest input sellers, distance to the closest market access to insurance, land size, off-farm income, access to information on climate change, access to extension services, access to credit, changes in temperature and rainfall were found to have a significant influence on the probability of farmers to perceive and/or adapt to climate change.

Target firm's integrity culture and M&A performance

AbstractThis study investigates whether the attribute of integrity culture (derived from target firms’ annual reports) influences merger and acquisition (M&A) performance. We find that a target firm's integrity culture, measured from its 10‐K reports, has a positive and significant effect on market reaction to the bidder firm's M&A announcement. Our study's analysis is found to be robust to sample selection bias by utilising the entropy balancing technique and to endogeneity concerns by employing an instrumental variable approach. Our results are also robust to alternative measures of integrity culture and when controlling for a target firm's religiosity and corporate social responsibility, an acquirer firm's fixed effects, chief executive officer's fixed effects, governance for corporate control and advisor quality. We identify the retention of a target firm's directors and of its customers as channels that underlie our main findings. Furthermore, we find that acquisition synergies improve, with decreased time taken to complete the deal, for acquisitions of target firms with a higher integrity culture.

Estimating response propensities in nonprobability surveys using machine learning weighted models

Propensity Score Adjustment (PSA) is a widely accepted method to reduce selection bias in nonprobability samples. In this approach, the (unknown) response probability of each individual is estimated in a nonprobability sample, using a reference probability sample. This, the researcher obtains a representation of the target population, reflecting the differences (for a set of auxiliary variables) between the population and the nonprobability sample, from which response probabilities can be estimated.Auxiliary probability samples are usually produced by surveys with complex sampling designs, meaning that the use of design weights is crucial to accurately calculate response probabilities. When a linear model is used for this task, maximising a pseudo log-likelihood function which involves design weights provides consistent estimates for the inverse probability weighting estimator. However, little is known about how design weights may benefit the estimates when techniques such as machine learning classifiers are used.This study aims to investigate the behaviour of Propensity Score Adjustment with machine learning classifiers, subject to the use of weights in the modelling step. A theoretical approximation to the problem is presented, together with a simulation study highlighting the properties of estimators using different types of weights in the propensity modelling step.

The gap that Matilda will bridge: a look at the Colombian case.

To determine gender gaps in Emeritus researchers in Colombia. Oaxaca-Blinder-Kitakagwa decomposition model, correcting the sample selection bias with the inclusion of Mills' inverse ratio (Heckman's Lambda) through an ordered probit model. Data: Information available in the ScienTI Platform - Colombia during the period 2015-2021. The results show that the gender gap between female and male researchers is 5.8%. To achieve Emeritus status, one must be over 65 years old, and the possibility of achieving Emeritus status is 5.1% higher for female researchers than for their male counterparts. These differences can be explained by the time constraints that female researchers face in being productive, as they spend more time than male researchers on caregiving responsibilities, either due to motherhood or the care of other dependent family members. The results obtained allow us to affirm that there is a gender gap in scientific research in Colombia in the Emeritus research category in the calls for proposals for the period 2015-2021. Moreover, the existing gap cannot be explained by factors associated with attributes of education and academic productivity that are part of the regulatory requirements, insofar as not being explained by them, it evidences the existence of discrimination against women researchers to access the highest research category.

Effects of mobile Internet use on the health of middle-aged and older adults: evidences from China health and retirement longitudinal study

The rapid development of digital technology has radically changed people’s lives. Simultaneously, as the population is rapidly aging, academic research is focusing on the use of Internet technology to improve middle-aged and older people’s health, particularly owing to the popularity of mobile networks, which has further increased the population’s accessibility to the Internet. However, related studies have not yet reached a consensus. Herein, empirical analysis of the influence of mobile Internet use on the subjective health and chronic disease status of individuals in their Middle Ages and above was conducted utilizing ordered logit, propensity score matching (PSM), and ordered probit models with data from the 2020 China Health and Retirement Longitudinal Study. The study aimed to provide a theoretical basis and reference for exploring technological advances to empower the development of a healthy Chinese population and to advance the process of healthy aging. The health of middle-aged and older adults mobile Internet users was greatly improved, according to our findings. Further, the use of mobile Internet by these persons resulted in improvements to both their self-assessed health and the state of their chronic diseases. As per the findings of the heterogeneity analysis, the impact of mobile Internet use was shown to be more pronounced on the well-being of middle-aged persons aged 45–60 years compared to those aged ≥ 60 years. Further, the endogeneity test revealed that the PSM model could better eliminate bias in sample selection. The results suggest that the estimates are more robust after eliminating endogeneity, and that failure to disentangle sample selectivity bias would overestimate not only the facilitating effect of mobile Internet use on the self-assessed health impacts of middle-aged and older adults, but also the ameliorating effect of mobile Internet use on the chronic diseases of middle-aged and older adults. The results of the mechanistic analysis suggest that social engagement is an important mediating mechanism between mobile Internet use and the health of middle-aged and older adults. This implies that mobile Internet use increases opportunities for social participation among middle-aged and older adults, thereby improving their health.

Job seekers’ pay expectations: The effect of voluntary disclosure in online résumés

Pay expectations are not fully explained, especially when they are presented in online résumés. There is also limited research that has investigated when job seekers are more likely to open up about pay expectations in résumés. However, this knowledge may help recruiters and company leaders to make prudent decisions about communication strategy with candidates, especially those belonging to stigmatized groups. The purpose of this study is to explore pay expectation disclosure through individual determinants available in online résumés. Drawn from both signaling perspectives and the selection, optimization, and compensation (SOC) theory, we examine human capital-related traits, demographics, and signals related to career development and analyze their relevance when job seekers indicate or hide pay expectations. The study explored the anonymous online résumés of 26,594 job-seeking applicants, using regression analysis with the Heckman correction to overcome the sample selection bias of pay indication. The analysis revealed a tendency among applicants, who were less educated and less experienced but with a higher number of job-related skills, to signal pay expectations compared to more educated and more experienced job seekers. Women and more mature job seekers were also inclined to provide disclosed-pay résumés. Our findings showed that the impact of the individual determinants on applicants’ pay expectations coincided with the impact of the same determinants on real market pay found in other studies. These findings provide insights into the signaling role of pay expectation disclosure in online résumés to support employers in building effective communication with job seekers.

The impact of the implementation of hierarchical medical system on population health: evidence from China

IntroductionThe implementation of a hierarchical medical system holds significant practical importance in advancing the Healthy China strategy and elevating the overall health status of the population of China.MethodsThis article empirically examines the influence of the implementation of a hierarchical medical system on the health of the population using the latest 2020 China Family Panel Studies database. Furthermore, it investigates the variability of this impact across distinct health insurance participation statuses and literacy groups.ResultsThe findings of the study demonstrate that the implementation of the hierarchical medical system substantially enhances the health of the population, not only in terms of self-assessed health but also through a notable positive effect on alleviating chronic diseases. These results maintain their validity even after conducting robustness tests utilizing a replacement estimation model. Heterogeneity analysis reveals that the impact of the hierarchical medical system on the population’s health status exhibits significant variation concerning health insurance participation and literacy. Specifically, regarding health insurance participation, the hierarchical medical system effectively improves both self-assessed health and chronic disease status among the insured population. However, for those not enrolled in health insurance, the hierarchical medical system only demonstrates improvement in chronic disease status, with insignificant results observed in enhancing self-assessed health status. Moreover, propensity score matching (PSM) was also used to address endogeneity problems resulting from sample selectivity bias. The findings demonstrate that endogeneity issues can be suitably addressed by the PSM model. Additionally, they point out that an overestimation of the impact of the hierarchical medical system on the population’s self-assessed health state would result from failing to take sample selectivity bias into account. On the other hand, it will lead to the underestimation of the effect of the hierarchical medical system on the status of chronic diseases.DiscussionMoving forward, steadfast efforts should be directed toward further enhancing the implementation of the hierarchical medical system. This includes the comprehensive promotion and using the pivotal role of the hierarchical medical system in improving the health of the population.

Does childhood maltreatment affect hikikomori via traits of modern-type depression?

BackgroundIn the 1990s, the concepts of hikikomori and modern-type depression (MTD) emerged in Japan. Hikikomori is a condition of social avoidance or isolation, characterized by staying at home and being physically isolated for at least six months. MTD is characterized by depressive symptoms—mainly in stressful work or school situations during adolescence and early adulthood—which tend to rapidly reduce or disappear after leaving the stressful situation. We hypothesized that childhood maltreatment can form MTD traits that lead to hikikomori. MethodsAs a first step, we conducted a multigroup path analysis between childhood maltreatment, MTD traits, and physical isolation in the hikikomori group. This study utilized the nine-item Patient Health Questionnaire (PHQ-9), Home Environment Questionnaire (HEQ), 22-item Tarumi Modern-Type Depressive Trait Scale (TACS-22), 25-item Hikikomori Questionnaire (HQ-25), and Hamilton Depression Rating Scale (HDRS). The HQ-25 contains three factors: physical isolation, lack of socialization, and lack of emotional support. ResultsThe hikikomori group included 92 patients and the control group comprised 137 healthy individuals. All total and subscale scores of PHQ-9, HEQ, TACS-22, HQ-25, and HDRS were significantly higher in the hikikomori group than in the control group. The risk model of childhood maltreatment for physical isolation via MTD traits obtained good fit with a goodness-of-fit index of.982. LimitationsThe study's limitations were its sample selection bias, cross-sectional design, and use of self-report scales. ConclusionsOur findings support the hypothesis that childhood maltreatment is an important risk factor for hikikomori via MTD traits.

Making the most of AI and machine learning in organizations and strategy research: Supervised machine learning, causal inference, and matching models

AbstractResearch SummaryWe spotlight the use of machine learning in two‐stage matching models to deal with sample selection bias. Recent advances in machine learning have unlocked new empirical possibilities for inductive theorizing. In contrast, the opportunities to use machine learning in regression studies involving large‐scale data with many covariates and a causal claim are still less well understood. Our core contribution is to guide researchers in the use of machine learning approaches to choosing matching variables for enhanced causal inference in propensity score matching models. We use an analysis of real‐world technology invention data of public–private relationships to demonstrate the method and find that machine learning can provide an alternative approach to ad hoc matching. However, as with any method, it is also important to understand its limitations.Managerial SummaryThis article explores the use of machine learning to enhance decision‐making, particularly in addressing sample selection bias in large‐scale datasets. The rapid development of AI and machine learning offers new, powerful tools especially for digital ecosystems where complex data and causal relationships are complex to analyze. We offer managers and stakeholders insight into the effective integration of machine learning for selecting critical variables in propensity score matching models. Through a detailed examination of real‐world data on technology inventions within public–private relationships, we demonstrate the effectiveness of machine learning as a robust alternative to traditional matching methods.

Acceptance of offshore wind farm in Southwest Sardinia in Italy. Do regional energy policies matter?

PurposeThe social acceptance of wind energy is increasingly conditioning the Italian Government and regions to authorize the construction of these plants. The proposal of offshore wind farm in the south-west of Sardinia has raised many perplexities both from the marine environmental point of view and from concern about increasing the electricity production in a region that already exports electricity to the peninsula. The purpose of this study is to evaluate what are the factors that most affect the coastal residents’ acceptability of an offshore wind farm.Design/methodology/approachThe data is based on a Sardinia-wide in-person survey with about 512 participants in the period between May and June 2023. Respondents were selected randomly from five different locations in Carloforte region. Multiple regression analysis investigates the factors that influenced acceptability to construction and development of offshore wind power plant. Four independent variables were selected in the regression models such as (i) interest towards wind energy, (ii) attitude towards renewable energy production, (iii) perceived regional energy policy and (iv) attitude towards offshore wind farm. The dependent variables are the rates of coastal local residents’ acceptance of offshore wind farm for respondents in five coastal towns.FindingsFishermen and summer tourism operators’ respondents expressed significant reluctance to accept the new project because they perceive a strong risk of environmental impact on both tuna fishing and marine recreational activities in the vicinity. The distance between the turbines at sea and local residents along the coast and surrounding small islands does not have an influence to accept wind energy. The energy policy of the Region of Sardinia, which is perceived by respondents as an expansive energy policy in the production of electricity that exceeds their needs, not being accompanied by a scenario of reducing that from fossil fuels, is the variable that led about 70% of respondents to express a negative opinion towards this project.Research limitations/implicationsA potential problem associated with the survey is that local response produced sample selection bias. The proportion of respondents with secondary education and no school certificate is quite high. It is possible that respondents who provided valid questionnaires were more low-educated and therefore, the sample may be biased towards lower-educated people. While this does not invalidate the results of this study, it is important to note that the sample was on average less educated than the overall population in Italy. In addition, people with lower incomes were more likely to complete the questionnaires; the results are more representative of a portion of the population with incomes below the national average.Practical implicationsResults showed how coastal communities are in favour of wind energy within their island, but not within a marine protected area that is considered unique for its beauty and marine ecosystem. Wind energy developers should, before submitting a proposal for an offshore wind farm project, consider among other things the long approval time and understand the activities carried out by local communities and their attachment to those places. Politicians and developers should develop a coherent energy transition policy based on a long-term vision of zero emissions, because according to the findings of this study, it was the regional energy policy that is the most controversial reason for residents to reject the project.Originality/valueThe contribution of this study is to fill a part of the research gap linking to ongoing energy transitions. In particular, this study analysed for the first time in Italy the acceptability level of an offshore renewable energy project considering the environmental implications and risks in the fragile marine ecosystem of Carloforte waters. This study also made it possible to analyse the impact of the regional energy policy on the acceptability of residents towards the development of offshore wind farms, which is scarcely discussed in the literature.

Female foragers sometimes hunt, yet gendered divisions of labor are real: a comment on Anderson et al. (2023) The Myth of Man the Hunter

Gendered divisions of labor are a feature of every known contemporary hunter-gatherer (forager) society. While gender roles are certainly flexible, and prominent and well-studied cases of female hunting do exist, it is more often men who hunt. A new study (Anderson et al., 2023) surveyed ethnographically known foragers and found that women hunt in 79% of foraging societies, with big-game hunting occurring in 33%. Based on this single type of labor, which is one among dozens performed in foraging societies, the authors question the existence of gendered division of labor altogether. As a diverse group of hunter-gatherer experts, we find that claims that foraging societies lack or have weak gendered divisions of labor are contradicted by empirical evidence. We conducted an in-depth examination of the data and methods of Anderson et al. (2023), finding evidence of sample selection bias and numerous coding errors undermining the paper's conclusions. Anderson et al. (2023) have started a useful dialogue to ameliorate the potential misconception that women never hunt. However, their analysis does not contradict the wide body of empirical evidence for gendered divisions of labor in foraging societies. Furthermore, a myopic focus on hunting diminishes the value of contributions that take different forms and downplays the trade-offs foragers of both sexes routinely face. We caution against ethnographic revisionism that projects Westernized conceptions of labor and its value onto foraging societies.

Least-cost decarbonization pathways for electricity generation in Finland: A convex quantile regression approach

This study investigates the least-cost decarbonization pathways in the Finnish electricity generation industry in order to achieve the national carbon neutrality goal by 2035. Various abatement measures, such as downscaling production, capital investment, and increasing labor and intermediate inputs, are considered. The marginal abatement costs (MACs) of greenhouse gas emissions are estimated using the convex quantile regression method and applied to unique register-based firm-level greenhouse gas emission data merged with financial statement data. We adjust the MAC estimates for the sample selection bias caused by zero-emission firms by applying the two-stage Heckman correction. Our empirical findings reveal that the median MAC ranges from 0.1 to 3.5 euros per tonne of CO2 equivalent. The projected economic cost of a 90% reduction in emissions is 62 million euros, while the estimated cost of achieving zero emissions is 83 million euros.

The impact of lobbying on managerial short-term resource adjustment decisions

This study examines the effect of political lobbying on firms' short-term resource adjustment decisions. Controlling for a wide range of known determinants of managerial cost behaviour, our results suggest that U.S. lobbying firms exhibit significantly less cost stickiness than non-lobbying firms. Lobbying reduces managers' "wait-and-see games" as they obtain preferential access to information on political and legislative processes. With early knowledge of impending (political and regulatory) threats and long-lasting downturns, managers can adjust unutilised capacity more swiftly. This effect is more pronounced for firms without alternative information channels like permanent exchange with regulators. Our findings hold for several robustness checks, for instance, controlling for potential sample selection bias or employing alternative measures of lobbying as well as (political) uncertainty and alternative samples.

Reject inference in credit scoring based on cost-sensitive learning and joint distribution adaptation method

As traditional credit evaluation methods generally only use accepted sample modeling, the rejected data is omitted, which means the model's prediction of new customers is biased. However, reject inference can be used to solve this credit evaluation sample selection bias. This paper proposes a new reject inference method based on joint distribution adaptation (JDA) and cost-sensitive semi-supervised support vector machines (CS4VM). First, this method uses both accepted (labeled) samples and rejected (unlabeled) samples modeling, which overcomes the deviations in traditional credit evaluation methods. Second, as the accepted sample and the rejected sample distributions are different, this method reduces the distribution differences between the accepted and rejected sample sets, which ensures that the sample data conforms to the basic assumptions in the semi-supervised model, and improves the performance of the classification model. Third, this method reduces the overall cost in the actual credit business by considering both the traditional misclassification costs when mining the default samples and the different decision weights for the accepted and rejected samples. Finally, an empirical study verifies the excellent predictive performance of the proposed method and effectively reduces the total credit costs.

The Effects of Antenatal Care Utilisation on Birthweight: Evidence from Ghana

Abstract In 2015, about 20.5 million babies were born with low birthweight (LBW), with the majority from low- and middle-income countries. The use of antenatal care (ANC) is a recommended health policy measure to reduce the risk of LBW. However, there is limited evidence for its effectiveness in reducing LBW in sub-Saharan Africa, although the region accounts for almost one-quarter of LBW babies, second only to Asia. Moreover, while the number of LBW in other regions is reducing, the number of LBW babies in sub-Saharan Africa is increasing. We therefore use data from Ghana (a sub-Saharan African country with a high incidence rate of LBW) to investigate the effect of using ANC on birthweight. The number of ANC visits and the month of ANC initiation were used as indicators for ANC utilisation. We applied an extended linear regression technique to deal with the potential sample selection bias and endogeneity associated with the use of ANC. The results from the study indicate that early ANC initiation and having more ANC visits are both associated with an increase in birthweight; however, the estimated effect of an early ANC visit is larger than having an additional visit. Therefore, we recommend that policies aimed at tackling the problem of LBW should mostly be geared towards promoting ANC utilisation, especially early ANC initiation. We also find that estimating the effect of ANC on birthweight without accounting for sample selection bias and endogeneity of ANC underestimates the effect of ANC on birthweight.

The Sustainability committee and environmental disclosure: International evidence

This paper contributes to the growing debate about the role of board-level sustainability committees, focusing on whether they leverage sustainability expertise for impactful environmental initiatives or purely serve as symbolic actions aimed at greenwashing. Using a large set of firms from 35 countries over the 2010–2017 period, we find that the presence of a sustainability committee is positively associated with higher-quality GHG emissions disclosure. This finding is robust to endogeneity and sample selection bias concerns. The sustainability committee effect is more pronounced when external environmental institutions are too weak to properly monitor corporate environmental disclosure. Our findings suggest that sustainability committees are not a symbolic management tool, but play a crucial role in enhancing corporate environmental disclosure.