Dynamic Treatment Effects Research Articles

Assessing the effectiveness of energy efficiency measures in the residential sector gas consumption through dynamic treatment effects: Evidence from England and Wales

Improving energy efficiency (EE) is vital to ensure a sustainable, affordable, and secure energy system. The residential sector represents, on average, 18.6% of the total final energy consumption in the OECD countries in 2018, reaching 29.5% in the UK (IEA, 2020a). Using a staggered differences-in-differences approach with dynamic treatment effects, we analyse changes in residential gas consumption five years before and after the adoption of energy efficiency measures. The analysis includes energy efficiency interventions involving the installation of new heating-related insulation equipment—i.e., of loft insulation and cavity walls, supported by energy efficiency programmes in England and Wales between 2005 and 2017—using a panel of 55,154 households from the National Energy Efficiency Data-Framework (NEED). We control for, among other factors, energy prices and the extent to which gas consumption changes are dependent on household characteristics and variations in weather conditions. Our results indicate that the adoption of EE measures is associated with significant reductions in household residential gas consumption one year after their implementation. However, the effect does not last in the long run and energy savings disappear four years after the retrofitting of cavity wall insulation measures and after two years following the installation of loft insulation. The disappearance of energy savings in the longer run could be explained by the energy performance gap, the rebound effect and/or by concurrent residential construction projects and renovations associated with increases in energy consumption. Notably, for households in deprived areas, the installation of these efficiency measures does not deliver energy savings. These results confirm the existence of effects that reduce the energy savings from the adoption of these efficiency technologies over time and indicates that, for some groups, these net savings do not seem to materialize.

The effect of cigarette and e-cigarette taxes on prescriptions for smoking cessation medications.

To test the effect of cigarette and e-cigarette taxes on prescriptions for smoking cessation medications. Symphony Health, IDV all-payer prescription claims data for the United States over the period 2009-2017. Prescription fills for smoking cessation products were provided at the patient's age, patient's sex, brand/generic, payment type, year, and quarter levels. We study the effect of state-level cigarette and e-cigarette tax rates on prescriptions for smoking cessation medications using two-way fixed effect modified difference-in-differences regressions. We also use a multiperiod difference-in-differences estimator robust to bias from dynamic and heterogeneous treatment effects with a staggered policy rollout. We use fills for Chantix, Zyban, and their generics, as well as Food and Drug Administration-approved nicotine replacement therapies that are paid for by insurance. We observe no statistically significant change in prescription fills following an increase in the e-cigarette tax rate, though we are unable to rule out potentially large effects. However, following a $1.00 increase in the cigarette tax rate, we observe a 1052 increase in prescription fills per 100,000 adults (95% CI: 57, 2046; 4.2% increase). The effect of cigarette taxes on prescription fills was particularly large for 18-34 year-olds. Our findings suggest that, during a period when e-cigarettes are widely available, cigarette tax increases remain effective in increasing use of these medications, but e-cigarette taxes do not increase use of these medications.

Examining evidence of time-dependent treatment effects: an illustration using regression methods

BackgroundFor the design and analysis of clinical trials with time-to-event outcomes, the Cox proportional hazards model and the logrank test have been the cornerstone methods for many decades. Increasingly, the key assumption of proportionality—or time-fixed effects—that underpins these methods has been called into question. The availability of novel therapies with new mechanisms of action and clinical trials of longer duration mean that non-proportional hazards are now more frequently encountered.MethodsWe compared several regression-based methods to model time-dependent treatment effects. For illustration purposes, we used selected endpoints from a large, community-based clinical trial of low dose daily aspirin in older persons. Relative and absolute estimands were defined, and analyses were conducted in all participants. Additional exploratory analyses were undertaken by selected subgroups of interest using interaction terms in the regression models.DiscussionIn the trial with median 4.7 years follow-up, we found evidence for non-proportionality and a time-dependent treatment effect of aspirin on cancer mortality not previously reported in trial findings. We also found some evidence of time-dependence to an aspirin by age interaction for major adverse cardiovascular events. For other endpoints, time-fixed treatment effect estimates were confirmed as appropriate.ConclusionsThe consideration of treatment effects using both absolute and relative estimands enhanced clinical insights into potential dynamic treatment effects. We recommend these analytical approaches as an adjunct to primary analyses to fully explore findings from clinical trials.

Disentangled representation for sequential treatment effect estimation

Background and objectiveTreatment effect estimation, as a fundamental problem in causal inference, focuses on estimating the outcome difference between different treatments. However, in clinical observational data, some patient covariates (such as gender, age) not only affect the outcomes but also affect the treatment assignment. Such covariates, named as confounders, produce distribution discrepancies between different treatment groups, thereby introducing the selection bias for the estimation of treatment effects. The situation is even more complicated in longitudinal data, because the confounders are time-varying that are subject to patient history and meanwhile affect the future outcomes and treatment assignments. Existing methods mainly work on cross-sectional data obtained at a specific time point, but cannot process the time-varying confounders hidden in the longitudinal data. MethodsIn this study, we address this problem for the first time by disentangled representation learning, which considers the observational data as consisting of three components, including outcome-specific factors, treatment-specific factors, and time-varying confounders. Based on this, the proposed approach adopts a recurrent neural network-based framework to process sequential information and learn the disentangled representations of the components from longitudinal observational sequences, captures the posterior distributions of latent factors by multi-task learning strategy. Moreover, mutual information-based regularization is adopted to eliminate the time-varying confounders. In this way, the association between patient history and treatment assignment is removed and the estimation can be effectively conducted. ResultsWe evaluate our model in a realistic set-up using a model of tumor growth. The proposed model achieves the best performance over benchmark models for both one-step ahead prediction (0.70% vs 0.74% for the-state-of-the-art model, when γ = 3. Measured by normalized root mean square error, the lower the better) and five-step ahead prediction (1.47% vs 1.83%) in most cases. By increasing the effect of confounders, our proposed model always shows superiority against the state-of-the-art model. In addition, we adopted T-SNE to visualize the disentangled representations and present the effectiveness of disentanglement explicitly and intuitively. ConclusionsThe experimental results indicate the powerful capacity of our model in learning disentangled representations from longitudinal observational data and dealing with the time-varying confounders, and demonstrate the surpassing performance achieved by our proposed model on dynamic treatment effect estimation.

Uncovering the dynamic effects of DEX treatment on lung cancer by integrating bioinformatic inference and multiscale modeling of scRNA-seq and proteomics data

Lung cancer is one of the leading causes of cancer-related death, with a five-year survival rate of 18%. It is a priority for us to understand the underlying mechanisms affecting lung cancer therapeutics' implementation and effectiveness. In this study, we combine the power of Bioinformatics and Systems Biology to comprehensively uncover functional and signaling pathways of drug treatment using bioinformatics inference and multiscale modeling of both scRNA-seq data and proteomics data. Based on a time series of lung adenocarcinoma derived A549 cells after DEX treatment, we first identified the differentially expressed genes (DEGs) in those lung cancer cells. Through the interrogation of regulatory network of those DEGs, we identified key hub genes including TGFβ, MYC, and SMAD3 varied underlie DEX treatment. Further gene set enrichment analysis revealed the TGFβ signaling pathway as the top enriched term. Those genes involved in the TGFβ pathway and their crosstalk with the ERBB pathway presented a strong survival prognosis in clinical lung cancer samples. With the basis of biological validation and literature-based curation, a multiscale model of tumor regulation centered on both TGFβ-induced and ERBB-amplified signaling pathways was developed to characterize the dynamic effects of DEX therapy on lung cancer cells. Our simulation results were well matched to available data of SMAD2, FOXO3, TGFβ1, and TGFβR1 over the time course. Moreover, we provided predictions of different doses to illustrate the trend and therapeutic potential of DEX treatment. The innovative and cross-disciplinary approach can be further applied to other computational studies in tumorigenesis and oncotherapy. We released the approach as a user-friendly tool named BIMM (Bioinformatic Inference and Multiscale Modeling), with all the key features available at https://github.com/chenm19/BIMM.

A Practical Guide to Counterfactual Estimators for Causal Inference with Time‐Series Cross‐Sectional Data

AbstractThis paper introduces a simple framework of counterfactual estimation for causal inference with time‐series cross‐sectional data, in which we estimate the average treatment effect on the treated by directly imputing counterfactual outcomes for treated observations. We discuss several novel estimators under this framework, including the fixed effects counterfactual estimator, interactive fixed effects counterfactual estimator and matrix completion estimator. They provide more reliable causal estimates than conventional two‐way fixed effects models when treatment effects are heterogeneous or unobserved time‐varying confounders exist. Moreover, we propose a new dynamic treatment effects plot, along with several diagnostic tests, to help researchers gauge the validity of the identifying assumptions. We illustrate these methods with two political economy examples and develop an open‐source package, fect, in both R and Stata to facilitate implementation.

Evaluating (weighted) dynamic treatment effects by double machine learning

Summary We consider evaluating the causal effects of dynamic treatments, i.e., of mul-tiple treatment sequences in various periods, based on double machine learning to control for observed, time-varying covariates in a data-driven way under a selection-on-observables assumption. To this end, we make use of so-called Neyman-orthogonal score functions, which imply the robustness of treatment effect estimation to moderate (local) misspecifications of the dynamic outcome and treatment models. This robustness property permits approximating outcome and treatment models by double machine learning even under high-dimensional covariates. In addition to effect estimation for the total population, we consider weighted estimation that permits assessing dynamic treatment effects in specific subgroups, e.g., among those treated in the first treatment period. We demonstrate that the estimators are asymptotically normal and $\sqrt{n}$-consistent under specific regularity conditions and investigate their finite sample properties in a simulation study. Finally, we apply the methods to the Job Corps study.

The effects of occupational mobility on wages of rehabilitated workers in Brazil

This work investigates how occupational mobility, after a professional rehabilitation process, affects the hourly wages of workers who are inserted in the formal job market in the commerce, services and industry sectors of Brazil. For that purpose, a data panel was used with individual information from the Annual List of Social Information (Relação Anual de Informações Sociais- RAIS) for the period of 2007 to 2016. The methodology used was the Staggered Difference in Differences and dynamic treatment effects, through an event study design. The results show a positive effect of occupational mobility on hourly wages in the sectors analyzed, however, this effect is decreasing in the period after the professional change.

Dynamic experiments of acid mine drainage with Rhodopseudomonas spheroides activated lignite immobilized sulfate-reducing bacteria particles treatment

Aiming at the problem that the treatment of acid mine drainage (AMD) by sulfate-reducing bacteria (SRB) biological method is susceptible to pH, metal ions, sulfate and carbon source. Lignite immobilized SRB particles (SRB-LP) and Rhodopseudomonas spheroides (R. spheroides) activated lignite immobilized SRB particles (R-SRB-LP) were prepared using microbial immobilization technology with SRB, R. spheroides and lignite as the main substrates. The dynamic experimental columns 1# and 2# were constructed with SRB-LP and R-SRB-LP as fillers, respectively, to investigate the dynamic repair effect of SRB-LP and R-SRB-LP on AMD. The mechanism of AMD treated with R-L-SRB particles was analyzed by scanning electron microscopy (SEM), fourier transform infrared (FTIR) spectrometer and low-temperature nitrogen adsorption. The result showed that the combination of R. spheroides and lignite could continuously provide carbon source for SRB, so that the highest removal rates of SO42−, Cu2+ and Zn2+ in AMD by R-SRB-LP were 93.97%, 98.52% and 94.42%, respectively, and the highest pH value was 7.60. The dynamic repair effect of R-SRB-LP on AMD was significantly better than that of SRB-LP. The characterization results indicated that after R-SRB-LP reaction, the functional groups of −OH and large benzene ring structure in lignite were broken, the lignite structure was destroyed, and the specific surface area was 1.58 times larger than before reaction. It illustrated that R. spheroides provided carbon source for SRB by degrading lignite. The strong SRB activity in R-SRB-LP, SRB can co-treat AMD with lignite, so that the dynamic treatment effect of R-SRB-LP on AMD is significantly better than that of SRB-LP.

Factor-augmented Bayesian treatment effects models for panel outcomes

A new, flexible model for inference of the effect of a binary treatment on a continuous outcome observed over subsequent time periods is proposed. The model allows to separate the associations due to endogeneity under treatment selection and additional longitudinal association of the outcomes, thus yielding unbiased estimates of dynamic treatment effects if both sources of association are present. The performance of the proposed method is investigated on simulated data and employed to re-analyze data on the longitudinal effects of a long maternity leave on mothers’ earnings after their return to the labour market.

Assessing the Effectiveness of Energy Efficiency Measures in the Residential Sector Gas Consumption Through Dynamic Treatment Effects: Evidence from England and Wales

Assessing the Effectiveness of Energy Efficiency Measures in the Residential Sector Gas Consumption Through Dynamic Treatment Effects: Evidence from England and Wales

Bayesian interpretation of immunotherapy trials with dynamic treatment effects

IntroductionThe mechanism of action of immune checkpoints inhibitors hinders the writing of rational statistical analysis plans for phase III randomised clinical trials (RCTs) because of their unpredictable dynamic effects. The purpose is to illustrate the advantages of Bayesian reporting of treatment efficacy analysis in immunotherapy RCTs, in contrast to frequentist reporting. MethodFourteen RCTs (one with two pairwise comparisons) that failed to achieve their primary objective (overall survival, OS) were selected. These RCTs were reanalysed using Bayesian Cox models with dynamic covariate coefficients and time-invariant models. ResultsThe RCTs that met inclusion criteria were 7 lung cancer trials, various other tumours, with antiPD1, antiPDL1 or antiCTLA4 therapies. The minimum detectable effect (δS) was superior to the true benefit observed in all cases, in conditions of non-proportional hazards. Schoenfeld tests indicated the existence of PH assumption violations (p<0.05) in 6/15 cases.The Bayesian Cox models revealed a probability of benefit >79% in all the RCTs, with the therapeutic equivalence hypothesis unlikely. The OS curves diverged after a median of 9.1 months. Since the divergency, no non-proportionality was evinced in 13/15, while the Wald tests achieved p<0.05 in 12/15 datasets. In all cases, the Bayesian Cox models with dynamic coefficients detected fluctuations of the hazard ratio, and increased 2-year OS was the most likely hypothesis. ConclusionWe recommend progressively implementing Bayesian and dynamic analyses in all RCTs of immunotherapy to interpret and assess the credibility of frequentist results.

Dynamic treatment effects of job training

SummaryThis paper estimates the dynamic returns to job training. We posit a model of sequential training participation, where decisions and outcomes depend on observed and unobserved characteristics. We analyze different treatment effects, including policy relevant parameters, and link them to continuation values and latent skills. The empirical analysis exploits administrative data combining job training records, matched employee‐employer information, and pre‐labor market ability measures from Chile. Although the average returns to training are small, these vary across the unobserved ability distribution and previous training choices. In fact, among young workers, the returns to training are lower when followed by additional training, providing evidence of dynamic substitutability. Policy experiments illustrate how increasing the local availability of training programs may affect earnings heterogeneously across dynamic responses.

Did the administrative approval reform in China affect the productivity of energy firms? – A quasi-natural experimental approach

This paper examines the impacts of China's administrative approval reform on the pro- ductivity of firms in the energy industry, using the establishment of the Administrative Approval Bureau (AAB) as a quasi-natural experiment. We apply event study strategy to exploit the dynamic treatment effects as the setting time of the AAB varies across cities. The findings show that the administrative approval reform has significantly induced an increase in the productivity of energy firms in China, by around 1% across various model specifications based on the data of listed energy firms from 2006 to 2019. This finding remains robust to alternative measurements of productivity and variations related to time trend, self-selection, outlier values and endogeneity. The effect of the reform is heteroge- neous across different industries, regions, and firms' ownership. Further mechanism anal- ysis suggests that the reform impacted firms' productivity through impeding technological innovation and promoting resource allocation, with the latter effect exceeding the former one. These results provide important policy implications for increasing the productivity of firms in both the energy sector and other energy dependent sectors.

Heterogeneity in returns to agricultural technologies with incomplete diffusion: Evidence from Ghana

AbstractIn this study, we employ a dynamic treatment effect approach to analyze heterogeneity in returns to farmers at different stages of adoption of a newly introduced inoculant technology, using a recent survey data of 600 soybean farmers from northern Ghana. Although farmers differ in their returns to adoption of new technologies, many empirical studies often fail to account for this heterogeneity. The empirical results reveal that farmers who are at advanced stages of adoption appear to, on average, more than double their yields and farm net returns, suggesting that the inoculant technology may be a game changer in the fight against extreme poverty in the region, where poverty is endemic and crop yields are persistently below the average potential yield target. Our findings further reveal that extension services as well as efficient input and output markets are key to the adoption process, by influencing knowledge acquisition, adoption, and continued adoption. Our findings also show significant impact heterogeneity at each adoption stage, with the long‐term benefits of the inoculant technology outweighing its short‐term benefits.

Fighting for not-So-Religious souls: The role of religious competition in secular conflicts.

Many countries embroiled in non-religious civil conflicts have experienced a dramatic increase in religious competition in recent years. This study examines whether increasing competition between religions affects violence in non-religious or secular conflicts. The study focuses on Colombia, a deeply Catholic country that has suffered one of the world’s longest-running internal conflicts and, in the last few decades, has witnessed an intense increase in religious competition between the Catholic Church and new non-Catholic churches. The estimation of a dynamic treatment effect model shows that establishing the first non-Catholic church in a municipality substantially increases the probability of conflict-related violence. The effect is larger for violence by guerrilla groups, and is concentrated on municipalities where the establishment of the first non-Catholic church leads to more intense religious competition. Further analysis suggests that the increase in guerrilla violence is associated with an expectation among guerrilla groups that their membership will decline as a consequence of more intense competition with religious groups for followers.

Sector heterogeneity and dynamic effects of innovation subsidies: Evidence from Horizon 2020

We evaluate the effect of a pan-European innovation funding program on firm growth and innovative output. Using a difference-in-differences estimation on a sample of matched firms, we find that subsidized firms are able to invest more in tangible and intangible assets, achieve higher growth of turnover and employment, and file more patent applications. We then analyze the dynamic treatment effect and find that the effects of subsidization tend to get stronger over time. Moreover, our findings indicate that the effect of subsidization is highly heterogeneous across sectors with different R&D or knowledge intensity and level of competition. Finally, we explore some economic channels to explain how subsidies generate strong effects on firm performance. We show that subsidized firms are able to generate more internal financing and attract more long-term borrowing after receiving the subsidy, yet we find no evidence that subsidized firms are able to attract more external equity financing than similar unsubsidized firms.

The treatment effect of the Shanxi Comprehensive Reform Area policy on PM2.5 concentrations: A study based on a quasi-experiment.

Based on panel data on 248 prefecture-level cities in China from 2003 to 2018, this study first estimates the treatment effect of the Shanxi Comprehensive Reform Area policy on PM2.5 concentrations using a PSM-DID method within the counterfactual framework. The average treatment effect shows that contrary to the naive before-after analysis, on average, the Shanxi Comprehensive Reform Area policy significantly increased the PM2.5 concentrations of prefecture-level cities in Shanxi Province by 0.211% annually, and the place-based placebo test shows that the treatment effect obtained above is robust. Second, the dynamic treatment effects show a continuous decrease in incremental effects during 2011-2018, gradually decreasing from a significant positive increment during 2011-2015 to a zero or even a negative increment during 2016-2018, indicating that the Shanxi Comprehensive Reform Area policy gradually increased in environmental friendliness. Third, the mediating effects estimated by the causal steps procedure show that the Shanxi Comprehensive Reform Area policy influenced PM2.5 concentrations by increasing the intensity of resource exploitation and decreasing the intensity of environmental regulations, but the capacity of scientific and technological innovations had no mediating effect on the relationship between the policy and PM2.5 concentrations. Therefore, the government should further reduce the intensity of resource exploitation, strengthen the intensity of environmental regulations, and promote environmentally focused scientific and technological innovations to reduce PM2.5 concentrations in Shanxi Province.

Dynamic evaluation of job search assistance

SummaryThis paper evaluates a job search assistance program for unemployed teachers where the assignment to the program is dynamic. We discuss the methodology of estimating dynamic treatment effects and identification conditions. In the empirical analysis, we use administrative data from a unique institutional environment in which we observe all variables determining assignment to the job search assistance program. This allows us to compare results from a dynamic discrete‐time evaluation model and a continuous‐time duration model. All approaches show that participation in the job search assistance program reduces exit rates from unemployment, in particular when starting the program early during the spell of unemployment. The discrete‐time approach makes less strict parametric assumptions, but the results are sensitive to the choice of control group and the unit of time.

Spousal bereavement and the cognitive health of older adults in the US: New insights on channels, single items, and subjective evidence

This study provides novel insights into older adults’ cognitive functioning before and after widowhood onset and possible effect channels. It further examines gender heterogeneity in the adaptation to (anticipated or actual) spousal bereavement, comparing objective evidence with subjective evidence of cognitive abilities. We used longitudinal data of up to 26,584 participants of the Health and Retirement Study, aged over 50 at recruitment, assessed biennially between 1998 and 2016. Two-way fixed effects with dynamic treatment effects were estimated for various cognitive measures, including six aggregated indices and six single item scales. After adjusting for effect channels including depression, social vulnerability, and stress, there remained significant widowhood effects on older adults’ cognitive health. Using single item scales, we established the adverse contemporaneous and adaptation effects on bereaved older females’ short-term memory, semantic memory, and numeracy. For bereaved older males, working memory and focus-of-attention deteriorated after widowhood onset. Meanwhile, subjective memory rating remained intact, contrary to objective evidence. We conclude that cognitive transitions to and from widowhood can exhibit distinctive patterns across objective and subjective cognitive domains. With the effect channels in mind, cognitive intervention for widowed older adults should be tailored to the temporal distance to spousal loss, gender, and task.