Causal Interpretation Research Articles

History-restricted marginal structural model and latent class growth analysis of treatment trajectories for a time-dependent outcome.

In previous work, we introduced a framework that combines latent class growth analysis (LCGA) with marginal structural models (LCGA-MSM). LCGA-MSM first summarizes the numerous time-varying treatment patterns into a few trajectory groups and then allows for a population-level causal interpretation of the group differences. However, the LCGA-MSM framework is not suitable when the outcome is time-dependent. In this study, we propose combining a nonparametric history-restricted marginal structural model (HRMSM) with LCGA. HRMSMs can be seen as an application of standard MSMs on multiple time intervals. To the best of our knowledge, we also present the first application of HRMSMs with a time-to-event outcome. It was previously noted that HRMSMs could pose interpretation problems in survival analysis when either targeting a hazard ratio or a survival curve. We propose a causal parameter that bypasses these interpretation challenges. We consider three different estimators of the parameters: inverse probability of treatment weighting (IPTW), g-computation, and a pooled longitudinal targeted maximum likelihood estimator (pooled LTMLE). We conduct simulation studies to measure the performance of the proposed LCGA-HRMSM. For all scenarios, we obtain unbiased estimates when using either g-computation or pooled LTMLE. IPTW produced estimates with slightly larger bias in some scenarios. Overall, all approaches have good coverage of the 95 % confidence interval. We applied our approach to a population of older Quebecers composed of 57,211 statin initiators and found that a greater adherence to statins was associated with a lower combined risk of cardiovascular disease or all-cause mortality.

CAUSAL LANGUAGE AND STATISTICS INSTRUCTION: EVIDENCE FROM A RANDOMIZED EXPERIMENT

Most current statistics courses include some instruction relevant to causal inference. Whether this instruction is incorporated as material on randomized experiments or as an interpretation of associations measured by correlation or regression coefficients, the way in which this material is presented may have important implications for understanding causal inference fundamentals. Although the connection between study design and the ability to infer causality is often described well, the link between the language used to describe study results and causal attribution typically is not well defined. The current study investigates this relationship experimentally using a sample of students in a statistics course at a large western university in the United States. It also provides (non-experimental) evidence about the association between statistics instruction and the ability to understand appropriate causal attribution. The results from our experimental vignette study suggest that the wording of study findings impacts causal attribution by the reader, and, perhaps more surprisingly, that this variation in level of causal attribution across different wording conditions seems to pale in comparison to the variation across study contexts. More research, however, is needed to better understand how to tailor statistics instruction to make students sufficiently wary of unwarranted causal interpretation.

Biochemical and haematological effects of serum PFOA, ADV and cC6O4 in workers of a chemical company producing fluoropolymers, Italy, 2013–2022

IntroductionPerfluoroalkyl and polyfluoroalkyl substances (PFAS) are widely used in the manufacture of fluoropolymers. We evaluated biochemical and haematological effects of three PFAS, serum perfluorooctanoic acid (PFOA), ADV, and cC6O4 in workers of a fluoropolymer company. MethodsUsing data (2013–2022), we fitted random intercept regression models adjusted for several covariates and reciprocal adjustment between the three PFAS. ResultsWe analysed data of 814 workers (698 men, 116 women), 607 from the chemical plant, 207 from the research centre, for a total of 4912 blood samples (2065 with all three PFAS measured). Median levels of PFOA and ADV were 21.3 and 120 μg/L. Most (65.5%) cC6O4 measurements were below the limits of quantification (which varied over time from 5 to 0.1 μg/L). For PFOA, we observed positive associations with total cholesterol (+1.1% increase per ln(PFOA) increase) and apolipoprotein B (+1.4%) and negative associations with alkaline phosphatase (−1.5%); suggestive associations were also found with RBC (−0.4%), IgA (−1.5%), IgM (−1.4%). ADV was positively associated with total and LDL cholesterol (+1.0% and +1.6% per ln(ADV) increase), apolipoprotein B (+1.0%), GGT (+2.1%), IgM (+1.4%), and WBC (+1.5%) and negatively associated with direct bilirubin (−2.3%) and alpha-2-globulins (−0.7%); suggestive associations were found for indirect bilirubin (−2.0%), oestradiol (−2.1%), ad CRP (+6.0%). For samples with detectable cC6O4 levels we observed higher values of ALP (+2.3%), proteins (+0.5%), IgG (+0.7%) and platelets (+1.6%) and suggestively increased total bilirubin (+3.9%), RBC (+0.6%), and oestradiol (+5.8%). Some associations (total cholesterol, apolipoprotein B, WBC, total bilirubin, and alkaline phosphatase showed reverse time trends in parallel with the strong decrease of serum PFOA and ADV over the study period. DiscussionWe found associations of serum PFOA and ADV with lipid metabolism, liver function, and immunoglobulins. The reverse time trends of some endpoints in parallel with decrease of serum PFOA and ADV reinforce causal interpretation of results. cC6O4 showed a different pattern of associations.

Investigating the influence of oral contraceptive pill use on multiple sclerosis risk using UK Biobank data

ObjectiveTo investigate the association between oral contraceptive (OC) pill use and the risk of developing Multiple Sclerosis (MS), attempting to address the limitations present in previous studies which produced conflicting results. DesignPopulation-based cohort study using data from the UK Biobank. SubjectsThe study included 181,058 women of white ethnicity born in England between 1937 and 1970, among which 1131 had an MS diagnosis. ExposureOC use, considering the self-reported age of initiation and discontinuation. The exposures of interest include: i) ever use, ii) current use, iii) duration of current use in years, and iv) age and year at initiation. Main Outcome MeasuresMS diagnosis (ICD-10: G35) was used as outcome of interest, and the associations with the exposures of interest were investigated using marginal structural models with a time-to-event approach. To adjust for confounding, we included in the models several variables including MS polygenic risk score (MS-PRS), education level, parity, smoking, fertility problems, obesity, and mononucleosis. We further aimed to evaluate the influence of parity using a mediation analysis. ResultsThe association of both ever and current OC use did not result in a statistically significant MS hazard increase (ever vs never users, HR=1.30 [95%CI:0.93,1.82, p=0.12]; current vs never users, HR=1.35 [95%CI:0.81,2.25, p=0.25]). However, we highlighted parity as an effect modifier for this association. In nulliparous women, ever and current use resulted in a significant two-fold and three-fold MS hazard increase (HR=2.08 [95%CI:1.04,4.17, p=0.04] and HR=3.15 [95%CI:1.43,6.92, p=0.004]). These associations were supported by significant MS hazard increases for a higher duration of current use and for an earlier age at initiation. We further highlighted genetic MS susceptibility as another effect modifier, as a stronger OC-MS hazard association was found in women with a low MS-PRS. ConclusionOur findings highlighted how the association between OC use and MS vary based on individual characteristics such as parity and genetic MS susceptibility. Importantly, current use in nulliparous women was found to be associated with a three-fold increase in MS hazard. We acknowledge the need for cautious causal interpretation and further research to validate these findings across diverse populations and OC types.

Average treatment effects on binary outcomes with stochastic covariates.

When evaluating the effect of psychological treatments on a dichotomous outcome variable in a randomized controlled trial (RCT), covariate adjustment using logistic regression models is often applied. In the presence of covariates, average marginal effects (AMEs) are often preferred over odds ratios, as AMEs yield a clearer substantive and causal interpretation. However, standard error computation of AMEs neglects sampling-based uncertainty (i.e., covariate values are assumed to be fixed over repeated sampling), which leads to underestimation of AME standard errors in other generalized linear models (e.g., Poisson regression). In this paper, we present and compare approaches allowing for stochastic (i.e., randomly sampled) covariates in models for binary outcomes. In a simulation study, we investigated the quality of the AME and stochastic-covariate approaches focusing on statistical inference in finite samples. Our results indicate that the fixed-covariate approach provides reliable results only if there is no heterogeneity in interindividual treatment effects (i.e., presence of treatment-covariate interactions), while the stochastic-covariate approaches are preferable in all other simulated conditions. We provide an illustrative example from clinical psychology investigating the effect of a cognitive bias modification training on post-traumatic stress disorder while accounting for patients' anxiety using an RCT.

Supplemental Nutrition Assistance Program and depressive symptoms among adults with low family income in the U.S.: The National Health and Nutrition Examination Survey 2011–2018

BackgroundThe Supplemental Nutrition Assistance Program (SNAP) has been established to reduce food insecurity. Limited evidence is available on SNAP participation status over time and depressive symptoms. We aimed to examine the associations of SNAP status over time among low-income individuals, with depressive symptoms in the U.S. MethodsNHANES participants aged ≥20 years of low family income from 2011 to 2018 with information available on depressive symptoms and SNAP use were included in analysis. Depressive symptoms were assessed using 9-item Patient Health Questionnaire (PHQ-9), and PHQ-9 score ≥ 10 is indicative of significant depressive symptoms. Multivariable linear and logistic regressions models were conducted to examine the associations of SNAP participation status over time (never receiving SNAP, receiving SNAP prior to >12 months ago, current receiving SNAP, receiving SNAP in the last 12 months but not currently) with depressive symptoms and significant depressive symptoms. ResultsCurrently receiving SNAP (beta (β) = 0.17, 95 % CI: 0.10, 0.25; odds ratio (OR) = 1.52, 95 % confidence interval (CI): 1.16, 2.00) and receiving SNAP in the last 12 months but not currently (β = 0.24, 95 % CI: 0.04, 0.43; OR = 1.83, 95 % CI: 1.16, 2.89) were associated with higher depressive symptoms and higher prevalence of significant depressive symptoms. LimitationsThe cross-sectional design precludes causal interpretation, and key variables were measured with self-report. ConclusionReceiving SNAP in the last 12 months was associated with higher levels of depressive symptoms among individuals with low family income. Improvement on diet quality may be important for reducing depressive symptoms among SNAP users.

(Re)Emerging disease and conflict risk in Africa, 1997-2019.

While the number of infectious zoonotic disease outbreaks has been rising, their impact on civil war and social conflict is poorly understood. This study addresses this fundamental limitation using a geolocated monthly dataset on 22 zoonotic diseases in Africa. Zoonotic disease is a key driver of new epidemics, making such pathogens a useful test case. Results suggest that over the January 1997 to December 2019 period, zoonotic disease was negatively associated with state initiation of civil conflict and positively associated with social conflict involving identity militias. Additional analyses find that the effect for identity militias is consistent with a causal interpretation. Rebel violence is not significantly associated with outbreaks. The results are robust to endogeneity concerns and additional sensitivity analyses.

Psychological distress mediates the association between COVID-19-related discrimination and subsequent PTSD symptom severity in healthcare workers: a two-year follow-up study

BackgroundPast research has suggested a cross-sectional association between COVID-19-related discrimination and PTSD symptom severity. However, no cohort study has examined the longitudinal association that better supports causal interpretation. Also, even if such an association genuinely exists, the specific pathway remains unclear.MethodsWe conducted a two-year follow-up study, obtaining data from healthcare workers in a hospital setting. We first evaluated how COVID-19-related discrimination in 2021 was associated with subsequent PTSD symptom severity in 2023. Thereafter, we conducted causal mediation analysis to examine how this association was mediated by psychological distress in 2022, accounting for exposure-mediator interaction. Missing data were handled using random forest imputation.ResultsA total of 660 hospital staff were included. The fully adjusted model showed greater PTSD symptom severity in individuals who experienced any COVID-19-related discrimination compared with those without such experiences (β, 0.44; 95% CI, 0.04–0.90). Regarding each type of discrimination, perceived discrimination was associated with greater PTSD symptom severity (β, 0.52; 95% CI, 0.08–0.96), whereas verbal discrimination did not reach statistical significance. Psychological distress mediated 28.1%–38.8% of the observed associations.ConclusionsCOVID-19-related discrimination is associated with subsequent PTSD symptom severity in healthcare workers. Psychological distress may serve as an important mediator, underscoring the potential need for interventions targeting this factor.

Work-Related Psychosocial Factors and Their Effects on Mental Workload Perception and Body Postures.

The perception of work is closely linked to body reactions that facilitate task performance. Previous studies have shown that psychosocial work factors significantly impact employee health on both psychological and physical levels, though their cross-sectional designs limit causal interpretations. In this study, participants performed sitting and standing tasks under four different levels of mental workload. The NASA-Task Load Index (NASA-TLX) assessed mental workload perception across six dimensions, while Rapid Entire Body Assessment (REBA) and Rapid Upper Limb Assessment (RULA) scores evaluated body postures for standing and sitting tasks, respectively. This study examined the effects of alarms, distractions, and time constraints-common psychosocial factors in healthcare environments-on human behavior. We compared NASA-TLX scores with corresponding REBA/RULA scores to evaluate how perceived mental workload affects body postures. One-way ANOVA assessed the impact of experimental conditions on response variables, and Pearson correlation analyses explored the relationships between psychosocial factors and these variables. Results indicated that alarm conditions most negatively impacted mental workload perception and body postures. Temporal demand and effort scores were particularly affected by psychosocial factors in both tasks. Gender influenced physical demand and performance scores (higher in females) for the standing task but did not affect REBA and RULA scores. These findings suggest that organizational and psychosocial factors significantly influence healthcare workers' behavior, health, and patient safety. Further research is needed to evaluate the specific effects of psychosocial factors on both physical and mental workload to understand the relationship between overall task workload and occupational disorders.

Does smoking cessation reduce other substance use, psychiatric symptoms, and pain symptoms? Results from an emulated hypothetical randomized trial of US veterans.

Quitting smoking may lead to improvement in substance use, psychiatric symptoms, and pain, especially among high-risk populations who are more likely to experience comorbid conditions. However, causal inferences regarding smoking cessation and its subsequent benefits have been limited. We emulated a hypothetical open-label randomized control trial of smoking cessation using longitudinal observational data of HIV-positive and HIV-negative US veterans from 2003-2015 in the Veterans Aging Cohort Study. We followed individuals from the first time they self-reported current cigarette smoking (baseline). We categorized participants as quitters or non-quitters at the first follow-up visit (approximately 1 year after baseline). Using inverse probability weighting to adjust for confounding and selection bias, we estimated odds ratios for improvement of co-occurring conditions (unhealthy alcohol use, cannabis use, illicit opioid use, cocaine use, depressive symptoms, anxiety symptoms, and pain symptoms) at second follow-up (approximately 2 years after baseline) for those who quit smoking compared to those who did not, among individuals who had the condition at baseline. Of 4,165 eligible individuals (i.e., current smokers at baseline), 419 reported no current smoking and 2,330 reported current smoking at the first follow-up. Adjusted odds ratios (95% confidence intervals) for associations between quitting smoking and improvement of each condition at second follow-up were: 2.10 (1.01, 4.35) for unhealthy alcohol use, 1.75 (1.00, 3.06) for cannabis use, 1.10 (0.58, 2.08) for illicit opioid use, and 2.25 (1.20, 4.24) for cocaine use, 0.78 (0.44, 1.38) for depressive symptoms, 0.93 (0.58, 1.49) for anxiety symptoms, and 1.31 (0.84, 2.06) for pain symptoms. While a causal interpretation of our findings may not be warranted, we found evidence for decreased substance use among veterans who quit cigarette smoking but none for the resolution of psychiatric conditions or pain symptoms. Findings suggest the need for additional resources combined with smoking cessation to reduce psychiatric and pain symptoms for high-risk populations.

P-776 School-age outcomes among IVF and ICSI -conceived children: a causal inference analysis using linked population-wide data

Abstract Study question What is the causal effect of method of oocyte insemination; standard in-vitro fertilisation (IVF) or intracytoplasmic sperm injection (ICSI), on childhood development at school entry? Summary answer The developmental outcomes at school entry for children conceived by ICSI, without severe male factor infertility, are equivalent to those conceived via standard IVF. What is known already Over 12 million children have been conceived globally with the assistance of assisted reproductive technology (ART). Within ART, ICSI continues to increase as the most common mode of oocyte insemination, even when not clearly indicated (i.e. without male factor infertility). Large cohort studies have suggested that ART, and ICSI in particular, may be associated with an increase in congenital abnormalities, autism spectrum disorder, developmental delay and intellectual disability. However, the studies are heterogenous and the evidence conflicting. The association between the ICSI technique itself and long-term childhood development remains unclear. Study design, size, duration Causal inference methods were used to analyse observational data in a way that emulates a target randomised clinical trial. For example, patients where standard IVF was not an option, (e.g., due to severe male factor infertility), were excluded. The state-wide cohort included linked maternal and childhood administrative data from Victoria, Australia. To eliminate potential confounding associated with IVF, comparison was made with children conceived via standard IVF (rather than conception without assistance). Participants/materials, setting, methods The study included singleton infants conceived via ART and born between 2005-2013. The primary exposure was conception via ICSI, with those born following conventional IVF insemination as controls. The outcome examined was the Australian Early Development Census (AEDC), a broad assessment of development and well-being performed in Australian schools, every triennium, at school entry (age 4-6). We combined multiple imputation and inverse probability weighting with regression adjustment to estimate population average causal effects. Main results and the role of chance The final cohort included 4,035 IVF and ICSI conceived children across the two outcome cohorts. Linked records were available for 2,468 ICSI children and 1,567 IVF-conceived controls. Maternal characteristics between the two treatment cohorts were similar for age, parity, education, language background, and number of eggs collected during treatment cycle. The ICSI conceived children were generally from residential areas of lower socio-economic status and conception was more likely to involve donated gametes. The unadjusted rate of the primary outcome; AEDC-defined “developmental vulnerability”, was 10.83% among the IVF group and 10.14% among the ICSI group. Given the use of observational data, there were missing data and inherent differences in the covariate profile between exposure cohorts. Multiple imputation, bootstrapping and doubly robust inverse probability weighted regression adjustment modelling was utilised to allow a causal interpretation of results. There was no evidence of a causal effect of ICSI-conception on the risk of AEDC-defined developmental vulnerability at school entry compared with spontaneously conceived children; adjusted risk difference -0.8% (95% CI -2.7% to 1.1%) and adjusted risk ratio 0.92 (95% CI 0.76 to 1.12). The adjustment variable set was determined a priori via a modified Delphi procedure. Limitations, reasons for caution Children who did not attend school due to severe disability were excluded, leading to potential selection bias. Within our study cohort of 2,468 ICSI children, no increased risk of developmental vulnerability was identified. It is possible that unmeasured confounders could have led to bias in estimating the treatment effects. Wider implications of the findings This study, in contrast to previous evidence, suggests that empiric use of ICSI compared with conventional IVF insemination does not affect early childhood developmental outcomes. These findings provide important reassurance for current and prospective parents, and clinicians alike. Trial registration number not applicable

Heterogenous causal effects: Potentials and pitfalls as illustrated with fatherhood and earnings

AbstractObjectiveTo discuss how methods to estimate heterogenous causal effects can be applied in Family Science and to supply empirical examples using the case of fatherhood and earnings.BackgroundMany questions important to family scientists do not focus on one‐size‐fits‐all average effects but rather on whether and how effects differ across groups. Recent methodological advances can assist this latter focus, offering new insights for theory and policy.MethodUsing Danish administrative data on all men who entered fatherhood 2005–2016 and on men of comparable age who did not, we focus on two types of heterogeneity in effects. First, effect heterogeneity across observed and unobserved covariates; second, treatment effect heterogeneity across the distribution of outcome variables.ResultsThe fatherhood premium on annual labor income is, in fact, a fatherhood penalty on average and across most margins of heterogeneity. Substantial heterogeneity exists across observed and unobserved characteristics and across the distribution of labor market earnings, with results indicating larger penalties for lower earners and those least likely to become fathers.ConclusionsEffect heterogeneity in Family Science holds great potential to inform policy and theory. However, causal interpretations always require assumptions, and researchers must be vigilant that the assumptions they make are warranted for each specific application.

Characterizing dehydration in short-term spaceflight using evidence from Project Mercury

Short-term spaceflight is commonly perceived as posing minimal risk to human health and performance. However, despite their duration, short-term flights potentially induce acute physiological changes that create risk to crews. One such change is dehydration (primarily body water loss) due to a heat-stressed environment. Such loss, if severe and prolonged, can lead to decrements in performance as well as increase the risk of more serious medical conditions. Though the general mechanisms of dehydration are broadly understood, the rate and extent of dehydration in short-term spaceflight has not been characterized. Combining data from the six spaceflights of the US Mercury program with a causal diagram illustrating the mechanisms of dehydration, we fit a path model to estimate the causal effects for all pathways in the causal model. Results demonstrate that Mercury astronauts experienced some degree of dehydration across the range of suited time and that the relationship between suited time and dehydration appears to be logarithmic. We discuss causal interpretations of the results and how the results from this and similar analyses can inform countermeasure development for short-term spaceflight.

Matrix Metalloproteinase-2 and CKD Progression: The Chronic Renal Insufficiency Cohort (CRIC) Study

Matrix metalloproteinase 2 (MMP-2) plays an important role in the development of fibrosis, the final common pathway of chronic kidney disease (CKD). This study aimed to assess the relationship between repeated measures of MMP-2 and CKD progression in a large, diverse prospective cohort. In a prospective cohort of Chronic Renal Insufficiency Cohort (CRIC) participants (N=3,827), MMP-2 was measured at baseline. In a case-cohort design, MMP-2 was additionally measured at year 2 in a randomly selected subcohort and cases of estimated glomerular filtration rate (eGFR) halving or kidney replacement therapy (KRT) (N= 1,439). CRIC is a multicenter prospective cohort of adults with CKD. MMP-2 measured in plasma at baseline and at year2. A composite kidney endpoint (KRT/eGFR halving). Weighted Cox proportional hazards models for case-cohort participants. Participants were followed for a median of 4.6 years from year 2 and 6.9 years from the baseline. Persistently elevated MMP-2 (≥300ng/mL at both baseline and year 2) increased the hazard of the composite kidney endpoint (HR, 1.61; 95% CI, 1.07-2.42; P=0.09) after adjusting for covariates. The relationship of persistently elevated MMP-2 was modified by levels of inflammation, with a 2.6 times higher rate of the composite kidney endpoint in those with high-sensitivity C-reactive protein<2.5g/dL at study entry. Heterogeneity of effect was found with proteinuria, with a baseline MMP-2 level of≥300ng/mL associated with an increased risk of the composite kidney endpoint (HR, 1.30; 95% CI, 1.09-1.54) only with proteinuria≥442mg/g. The observational study design limits causal interpretation. Elevated MMP-2 is associated with CKD progression, particularly among those with low inflammation and those with proteinuria. Future investigations are warranted to confirm the reduction in risk of CKD progression among these subgroups of patients with CKD.

Do professionals show a bias specific to treatment for people who have sexually offended in their interpretation of ambiguous evidence?

ABSTRACT The purpose of the current study was to examine whether professionals who work with people who have sexually offended are biased towards accepting ambiguous evidence as supportive of the effectiveness of treatment for these clients. In an online survey, professionals who work with people who have sexually offended (N = 58) were randomly assigned to receive a brief summary of ambiguous correlational evidence about either treatment for sexual offending or treatment for people who have cancer. Participants were then asked to select from causal and alternate interpretations of the evidence, whether they would recommend implementation of the treatment, and the proportion of funding they would allocate to implementation of the treatment. More than half of the participants incorrectly drew causal inferences and recommended implementing the treatment. However, there was limited evidence that this bias was specific to treatment for people who have sexually offended and there was no evidence that this bias was greater for participants who were more involved versus less involved in treatment for people who have sexually offended. Overall, our results most clearly suggest the operation of the pervasive, general critical thinking error of inferring causation from correlation rather than a self-serving or otherwise motivated bias. PRACTICE IMPACT STATEMENT The current study raises awareness of a common critical thinking error that can lead to the implementation of ineffective or even harmful practices and policies. We make recommendations to help reduce this error, which would facilitate more effective practice and policy.

Trajectories of Young People’s Cannabis Use During the COVID-19 Pandemic in Germany

Abstract: Objectives: This analysis aims to examine different trajectories of cannabis use within adolescents and young adults in Germany before and during the first two years of the COVID-19 pandemic. Methods: A cross-sectional online survey was conducted between January and Mid-February 2022 with 16,765 German youths aged 14 to 21 years (51.4 % female). Group-based trajectory modelling based on retrospective data on cannabis use before and during different stages of the COVID-19 pandemic was calculated with subsequent multinomial analysis predicting trajectory group membership by sociodemographic variables, sensation seeking and experienced pandemic related burdens. Results: Five different trajectory groups of cannabis users were observed: rare cannabis users (53 %), increasing cannabis users, rare pre-pandemic use (22 %), increasing cannabis users, moderate pre-pandemic use (12 %), decreasing cannabis users (8 %), and (nearly) daily cannabis users (6 %). Individuals with older age, male gender, higher sensation seeking, lower education, and a higher pandemic-related burden were significantly more likely to report a more frequent/heavy cannabis use during the COVID-19 pandemic. Conclusion: The large proportion of youths with an increased cannabis use during the COVID-19 pandemic shows need for low threshold help offers, and cannabis prevention programs. Due to the study design, one should be careful with causal interpretations.

Causal inference for treatment effects in partially nested designs.

artially nested designs (PNDs) are common in intervention studies in psychology and other social sciences. With this design, participants are assigned to treatment and control groups on an individual basis, but clustering occurs in some but not all groups (e.g., the treatment group). In recent years, there has been substantial development of methods for analyzing data from PNDs. However, little research has been done on causal inference for PNDs, especially for PNDs with nonrandomized treatment assignments. To reduce the research gap, in the current study, we used the expanded potential outcomes framework to define and identify the average causal treatment effects in PNDs. Based on the identification results, we formulated the outcome models that could produce treatment effect estimates with causal interpretation and evaluated how alternative model specifications affect the causal interpretation. We also developed an inverse propensity weighted (IPW) estimation approach and proposed a sandwich-type standard error estimator for the IPW-based estimate. Our simulation studies demonstrated that both the outcome modeling and the IPW methods specified following the identification results can yield satisfactory estimates and inferences of the average causal treatment effects. We applied the proposed approaches to data from a real-life pilot study of the Pregnant Moms' Empowerment Program for illustration. The current study provides guidance and insights on causal inference for PNDs and adds to researchers' toolbox of treatment effect estimation with PNDs. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Director networks, accounting conservatism and director reputation: Evidence after financial reporting failure

This study examines whether connected boards of directors restore their reputation via conservative accounting after financial misstatements. Using a sample of restating firms from 2004 to 2020, we find central boards of directors are positively related to accounting conservatism in the post-restatement period. More importantly, we find accounting conservatism has positive effects on the reputation restoration of directors in central boards, leading to more prestigious board seats, additional outside directorships, and higher compensation following restatements. The reputation recovery of directors in central boards is more prominent among restating firms that seek external financing where potential benefits of director reputation are greater. The findings suggest that directors in central boards successfully recover their damaged reputation via conservative accounting after restatements and minimize adverse consequences in the director market. Our robustness tests support the causal interpretation of our evidence. Collectively, our study offers novel insights into the monitoring incentives resulted from the reputation restoration of directors in central boards via conservative accounting after financial reporting failures.

Gendered choices of labour market integration programmes: evidence from the United States

PurposeWe investigate whether women are more likely than men to choose to pursue a competency-based labour market integration programme, rather than the time-based labour market integration programme. We further investigate whether women with existing but uncertified skills are even more likely to pursue a competency-based labour market integration programme.Design/methodology/approachWe test our hypotheses using ordinary least squares applied to linear probability models. We discuss the relative advantages of this methodology. We show the robustness of our results through multiple specifications and estimation methods. Finally, we discuss the reasons preventing us from granting our results a causal interpretation and discuss how they are surmountable in future research.FindingsWomen are significantly more likely to enrol into competency-based programmes, relative to time-based. Women with existing but uncertified skills are significantly more likely to enrol into competency-based programmes, whereas women without skills or with college degrees are not significantly different from the baseline. Our findings are robust to various specifications, and we include a comprehensive set of fixed-effect vectors, addressing industrial, occupational and time-varying state specificities.Research limitations/implicationsFirst, our empirical test of hypothesis H2 is hindered by the construction of the “some college or associate’s degree” variable in RAPIDS data. “Some college” is very different from an associate’s degree. Second we had to choose between omitted variable bias and selection bias. Because of the demonstrated importance of the occupation and industry variables in existing literature, we included those variables at the risk of selection bias. Occupation and industry fixed effects reduce, but do not eliminate, omitted variable bias. Finally, the third limitation of this paper is external validity. Registered Apprenticeship programmes are quite idiosyncratic to the United States.Social implicationsThe rollout and expansion of CBRA may thus be an avenue through policymakers may reduce the gender training gap. This may in turn give more women access to the labour market and allow more women to benefit from the “wage premia” of Registered Apprenticeship completion on the labour market (Lou and Hawley, 2019).Originality/valueThis article is the first that applies econometric methods to investigate women’s choices of labour market integration programmes, using Registered Apprenticeship as a case study. We discuss the implications of our findings, highlighting how competency-based programmes may be an approach to better serving more diverse populations in Registered Apprenticeship.

Semiparametric distribution regression with instruments and monotonicity

This paper proposes IV-based estimators for the semiparametric distribution regression model in the presence of an endogenous regressor, which are based on an extension of IV probit estimators and the idea of control functions. We discuss the causal interpretation of the estimators and two methods (monotone rearrangement and isotonic regression) to ensure a monotonically increasing distribution function. Asymptotic properties and simulation evidence are provided. An application to income equations with German SOEP data reveals statistically significant and heterogeneous differences to the inconsistent non-IV-based estimator.