Santiago Acerenza, Otávio Bartalotti, Federico Veneri
This paper develops sharp testable implications for Tobit and IV-Tobit models' identifying assumptions: linear index specification, (joint) normality of latent errors, and treatment (instrument) exogeneity and relevance. The new sharp testable equalities can detect all possible observable violations of the identifying conditions. We propose a testing procedure for the model's validity using existing inference methods for intersection bounds. Simulation results suggests proper size for large samples and that the test is powerful to detect large violation of the exogeneity assumption and violations in the error structure. Finally, we review and propose new alternative paths to partially identify the parameters of interest under less restrictive assumptions.
{"title":"Testing identifying assumptions in Tobit Models","authors":"Santiago Acerenza, Otávio Bartalotti, Federico Veneri","doi":"arxiv-2408.02573","DOIUrl":"https://doi.org/arxiv-2408.02573","url":null,"abstract":"This paper develops sharp testable implications for Tobit and IV-Tobit\u0000models' identifying assumptions: linear index specification, (joint) normality\u0000of latent errors, and treatment (instrument) exogeneity and relevance. The new\u0000sharp testable equalities can detect all possible observable violations of the\u0000identifying conditions. We propose a testing procedure for the model's validity\u0000using existing inference methods for intersection bounds. Simulation results\u0000suggests proper size for large samples and that the test is powerful to detect\u0000large violation of the exogeneity assumption and violations in the error\u0000structure. Finally, we review and propose new alternative paths to partially\u0000identify the parameters of interest under less restrictive assumptions.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"453 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141947523","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ramon de Punder, Timo Dimitriadis, Rutger-Jan Lange
Score-driven models have been applied in some 400 published articles over the last decade. Much of this literature cites the optimality result in Blasques et al. (2015), which, roughly, states that sufficiently small score-driven updates are unique in locally reducing the Kullback-Leibler (KL) divergence relative to the true density for every observation. This is at odds with other well-known optimality results; the Kalman filter, for example, is optimal in a mean squared error sense, but may move in the wrong direction for atypical observations. We show that score-driven filters are, similarly, not guaranteed to improve the localized KL divergence at every observation. The seemingly stronger result in Blasques et al. (2015) is due to their use of an improper (localized) scoring rule. Even as a guaranteed improvement for every observation is unattainable, we prove that sufficiently small score-driven updates are unique in reducing the KL divergence relative to the true density in expectation. This positive$-$albeit weaker$-$result justifies the continued use of score-driven models and places their information-theoretic properties on solid footing.
{"title":"Kullback-Leibler-based characterizations of score-driven updates","authors":"Ramon de Punder, Timo Dimitriadis, Rutger-Jan Lange","doi":"arxiv-2408.02391","DOIUrl":"https://doi.org/arxiv-2408.02391","url":null,"abstract":"Score-driven models have been applied in some 400 published articles over the\u0000last decade. Much of this literature cites the optimality result in Blasques et\u0000al. (2015), which, roughly, states that sufficiently small score-driven updates\u0000are unique in locally reducing the Kullback-Leibler (KL) divergence relative to\u0000the true density for every observation. This is at odds with other well-known\u0000optimality results; the Kalman filter, for example, is optimal in a mean\u0000squared error sense, but may move in the wrong direction for atypical\u0000observations. We show that score-driven filters are, similarly, not guaranteed\u0000to improve the localized KL divergence at every observation. The seemingly\u0000stronger result in Blasques et al. (2015) is due to their use of an improper\u0000(localized) scoring rule. Even as a guaranteed improvement for every\u0000observation is unattainable, we prove that sufficiently small score-driven\u0000updates are unique in reducing the KL divergence relative to the true density\u0000in expectation. This positive$-$albeit weaker$-$result justifies the continued\u0000use of score-driven models and places their information-theoretic properties on\u0000solid footing.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"90 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141947445","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The realization of FDI and DDI from January to December 2022 reached Rp1,207.2 trillion. The largest FDI investment realization by sector was led by the Basic Metal, Metal Goods, Non-Machinery, and Equipment Industry sector, followed by the Mining sector and the Electricity, Gas, and Water sector. The uneven amount of FDI investment realization in each industry and the impact of the COVID-19 pandemic in Indonesia are the main issues addressed in this study. This study aims to identify the factors that influence the entry of FDI into industries in Indonesia and measure the extent of these factors' influence on the entry of FDI. In this study, classical assumption tests and hypothesis tests are conducted to investigate whether the research model is robust enough to provide strategic options nationally. Moreover, this study uses the ordinary least squares (OLS) method. The results show that the electricity factor does not influence FDI inflows in the three industries. The Human Development Index (HDI) factor has a significant negative effect on FDI in the Mining Industry and a significant positive effect on FDI in the Basic Metal, Metal Goods, Non-Machinery, and Equipment Industries. However, HDI does not influence FDI in the Electricity, Gas, and Water Industries in Indonesia.
{"title":"Analysis of Factors Affecting the Entry of Foreign Direct Investment into Indonesia (Case Study of Three Industrial Sectors in Indonesia)","authors":"Tracy Patricia Nindry Abigail Rolnmuch, Yuhana Astuti","doi":"arxiv-2408.01985","DOIUrl":"https://doi.org/arxiv-2408.01985","url":null,"abstract":"The realization of FDI and DDI from January to December 2022 reached\u0000Rp1,207.2 trillion. The largest FDI investment realization by sector was led by\u0000the Basic Metal, Metal Goods, Non-Machinery, and Equipment Industry sector,\u0000followed by the Mining sector and the Electricity, Gas, and Water sector. The\u0000uneven amount of FDI investment realization in each industry and the impact of\u0000the COVID-19 pandemic in Indonesia are the main issues addressed in this study.\u0000This study aims to identify the factors that influence the entry of FDI into\u0000industries in Indonesia and measure the extent of these factors' influence on\u0000the entry of FDI. In this study, classical assumption tests and hypothesis\u0000tests are conducted to investigate whether the research model is robust enough\u0000to provide strategic options nationally. Moreover, this study uses the ordinary\u0000least squares (OLS) method. The results show that the electricity factor does\u0000not influence FDI inflows in the three industries. The Human Development Index\u0000(HDI) factor has a significant negative effect on FDI in the Mining Industry\u0000and a significant positive effect on FDI in the Basic Metal, Metal Goods,\u0000Non-Machinery, and Equipment Industries. However, HDI does not influence FDI in\u0000the Electricity, Gas, and Water Industries in Indonesia.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"10 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141947557","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Researchers are often interested in evaluating the impact of a policy on the entire (or specific parts of the) distribution of the outcome of interest. In this paper, I provide a practical toolkit to recover the whole counterfactual distribution of the untreated potential outcome for the treated group in non-experimental settings with staggered treatment adoption by generalizing the existing quantile treatment effects on the treated (QTT) estimator proposed by Callaway and Li (2019). Besides the QTT, I consider different approaches that anonymously summarize the quantiles of the distribution of the outcome of interest (such as tests for stochastic dominance rankings) without relying on rank invariance assumptions. The finite-sample properties of the estimator proposed are analyzed via different Monte Carlo simulations. Despite being slightly biased for relatively small sample sizes, the proposed method's performance increases substantially when the sample size increases.
{"title":"Distributional Difference-in-Differences Models with Multiple Time Periods: A Monte Carlo Analysis","authors":"Andrea Ciaccio","doi":"arxiv-2408.01208","DOIUrl":"https://doi.org/arxiv-2408.01208","url":null,"abstract":"Researchers are often interested in evaluating the impact of a policy on the\u0000entire (or specific parts of the) distribution of the outcome of interest. In\u0000this paper, I provide a practical toolkit to recover the whole counterfactual\u0000distribution of the untreated potential outcome for the treated group in\u0000non-experimental settings with staggered treatment adoption by generalizing the\u0000existing quantile treatment effects on the treated (QTT) estimator proposed by\u0000Callaway and Li (2019). Besides the QTT, I consider different approaches that\u0000anonymously summarize the quantiles of the distribution of the outcome of\u0000interest (such as tests for stochastic dominance rankings) without relying on\u0000rank invariance assumptions. The finite-sample properties of the estimator\u0000proposed are analyzed via different Monte Carlo simulations. Despite being\u0000slightly biased for relatively small sample sizes, the proposed method's\u0000performance increases substantially when the sample size increases.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141947558","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Machine learning methods for estimating treatment effect heterogeneity promise greater flexibility than existing methods that test a few pre-specified hypotheses. However, one problem these methods can have is that it can be challenging to extract insights from complicated machine learning models. A high-dimensional distribution of conditional average treatment effects may give accurate, individual-level estimates, but it can be hard to understand the underlying patterns; hard to know what the implications of the analysis are. This paper proposes the Distilled Causal Tree, a method for distilling a single, interpretable causal tree from a causal forest. This compares well to existing methods of extracting a single tree, particularly in noisy data or high-dimensional data where there are many correlated features. Here it even outperforms the base causal forest in most simulations. Its estimates are doubly robust and asymptotically normal just as those of the causal forest are.
{"title":"Distilling interpretable causal trees from causal forests","authors":"Patrick Rehill","doi":"arxiv-2408.01023","DOIUrl":"https://doi.org/arxiv-2408.01023","url":null,"abstract":"Machine learning methods for estimating treatment effect heterogeneity\u0000promise greater flexibility than existing methods that test a few pre-specified\u0000hypotheses. However, one problem these methods can have is that it can be\u0000challenging to extract insights from complicated machine learning models. A\u0000high-dimensional distribution of conditional average treatment effects may give\u0000accurate, individual-level estimates, but it can be hard to understand the\u0000underlying patterns; hard to know what the implications of the analysis are.\u0000This paper proposes the Distilled Causal Tree, a method for distilling a\u0000single, interpretable causal tree from a causal forest. This compares well to\u0000existing methods of extracting a single tree, particularly in noisy data or\u0000high-dimensional data where there are many correlated features. Here it even\u0000outperforms the base causal forest in most simulations. Its estimates are\u0000doubly robust and asymptotically normal just as those of the causal forest are.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"61 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141947447","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Mingyang Li, Maoqin Yuan, Han Pengsihua, Yuan Yuan, Zejun Wang
This study investigates the potential impact of "LK-99," a novel material developed by a Korean research team, on the power equipment industry. Using evolutionary game theory, the interactions between governmental subsidies and technology adoption by power companies are modeled. A key innovation of this research is the introduction of sensitivity analyses concerning time delays and initial subsidy amounts, which significantly influence the strategic decisions of both government and corporate entities. The findings indicate that these factors are critical in determining the rate of technology adoption and the efficiency of the market as a whole. Due to existing data limitations, the study offers a broad overview of likely trends and recommends the inclusion of real-world data for more precise modeling once the material demonstrates room-temperature superconducting characteristics. The research contributes foundational insights valuable for future policy design and has significant implications for advancing the understanding of technology adoption and market dynamics.
{"title":"Application of Superconducting Technology in the Electricity Industry: A Game-Theoretic Analysis of Government Subsidy Policies and Power Company Equipment Upgrade Decisions","authors":"Mingyang Li, Maoqin Yuan, Han Pengsihua, Yuan Yuan, Zejun Wang","doi":"arxiv-2408.01017","DOIUrl":"https://doi.org/arxiv-2408.01017","url":null,"abstract":"This study investigates the potential impact of \"LK-99,\" a novel material\u0000developed by a Korean research team, on the power equipment industry. Using\u0000evolutionary game theory, the interactions between governmental subsidies and\u0000technology adoption by power companies are modeled. A key innovation of this\u0000research is the introduction of sensitivity analyses concerning time delays and\u0000initial subsidy amounts, which significantly influence the strategic decisions\u0000of both government and corporate entities. The findings indicate that these\u0000factors are critical in determining the rate of technology adoption and the\u0000efficiency of the market as a whole. Due to existing data limitations, the\u0000study offers a broad overview of likely trends and recommends the inclusion of\u0000real-world data for more precise modeling once the material demonstrates\u0000room-temperature superconducting characteristics. The research contributes\u0000foundational insights valuable for future policy design and has significant\u0000implications for advancing the understanding of technology adoption and market\u0000dynamics.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"5 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141947446","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The synthetic control method (SCM) is widely used for causal inference with panel data, particularly when there are few treated units. SCM assumes the stable unit treatment value assumption (SUTVA), which posits that potential outcomes are unaffected by the treatment status of other units. However, interventions often impact not only treated units but also untreated units, known as spillover effects. This study introduces a novel panel data method that extends SCM to allow for spillover effects and estimate both treatment and spillover effects. This method leverages a spatial autoregressive panel data model to account for spillover effects. We also propose Bayesian inference methods using Bayesian horseshoe priors for regularization. We apply the proposed method to two empirical studies: evaluating the effect of the California tobacco tax on consumption and estimating the economic impact of the 2011 division of Sudan on GDP per capita.
合成控制法(SCM)被广泛应用于面板数据的因果推断,尤其是在处理单位较少时。SCM 假定稳定单位处理值假设(SUTVA),即潜在结果不受其他单位处理状态的影响。然而,干预措施往往不仅影响治疗单位,也影响未治疗单位,这就是所谓的溢出效应。本研究介绍了一种新颖的面板数据方法,该方法扩展了单因素模型,以考虑溢出效应,并同时估计治疗效应和溢出效应。该方法利用空间自回归面板数据模型来考虑溢出效应。我们还提出了使用贝叶斯马蹄先验进行正则化的贝叶斯推断方法。我们将提出的方法应用于两项实证研究:评估加利福尼亚烟草税对消费的影响,以及估算 2011 年苏丹分裂对人均 GDP 的经济影响。
{"title":"Bayesian Synthetic Control Methods with Spillover Effects: Estimating the Economic Cost of the 2011 Sudan Split","authors":"Shosei Sakaguchi, Hayato Tagawa","doi":"arxiv-2408.00291","DOIUrl":"https://doi.org/arxiv-2408.00291","url":null,"abstract":"The synthetic control method (SCM) is widely used for causal inference with\u0000panel data, particularly when there are few treated units. SCM assumes the\u0000stable unit treatment value assumption (SUTVA), which posits that potential\u0000outcomes are unaffected by the treatment status of other units. However,\u0000interventions often impact not only treated units but also untreated units,\u0000known as spillover effects. This study introduces a novel panel data method\u0000that extends SCM to allow for spillover effects and estimate both treatment and\u0000spillover effects. This method leverages a spatial autoregressive panel data\u0000model to account for spillover effects. We also propose Bayesian inference\u0000methods using Bayesian horseshoe priors for regularization. We apply the\u0000proposed method to two empirical studies: evaluating the effect of the\u0000California tobacco tax on consumption and estimating the economic impact of the\u00002011 division of Sudan on GDP per capita.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"184 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141884646","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper serves as a literature review of methodology concerning the (modern) causal inference methods to address the causal estimand with observational/survey data that have been or will be used in social science research. Mainly, this paper is divided into two parts: inference from statistical estimand for the causal estimand, in which we reviewed the assumptions for causal identification and the methodological strategies addressing the problems if some of the assumptions are violated. We also discuss the asymptotical analysis concerning the measure from the observational data to the theoretical measure and replicate the deduction of the efficient/doubly robust average treatment effect estimator, which is commonly used in current social science analysis.
{"title":"Methodological Foundations of Modern Causal Inference in Social Science Research","authors":"Guanghui Pan","doi":"arxiv-2408.00032","DOIUrl":"https://doi.org/arxiv-2408.00032","url":null,"abstract":"This paper serves as a literature review of methodology concerning the\u0000(modern) causal inference methods to address the causal estimand with\u0000observational/survey data that have been or will be used in social science\u0000research. Mainly, this paper is divided into two parts: inference from\u0000statistical estimand for the causal estimand, in which we reviewed the\u0000assumptions for causal identification and the methodological strategies\u0000addressing the problems if some of the assumptions are violated. We also\u0000discuss the asymptotical analysis concerning the measure from the observational\u0000data to the theoretical measure and replicate the deduction of the\u0000efficient/doubly robust average treatment effect estimator, which is commonly\u0000used in current social science analysis.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"37 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141884647","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
When do linear regressions estimate causal effects in quasi-experiments? This paper provides a generic diagnostic that assesses whether a given linear regression specification on a given dataset admits a design-based interpretation. To do so, we define a notion of potential weights, which encode counterfactual decisions a given regression makes to unobserved potential outcomes. If the specification does admit such an interpretation, this diagnostic can find a vector of unit-level treatment assignment probabilities -- which we call an implicit design -- under which the regression estimates a causal effect. This diagnostic also finds the implicit causal effect estimand. Knowing the implicit design and estimand adds transparency, leads to further sanity checks, and opens the door to design-based statistical inference. When applied to regression specifications studied in the causal inference literature, our framework recovers and extends existing theoretical results. When applied to widely-used specifications not covered by existing causal inference literature, our framework generates new theoretical insights.
{"title":"Potential weights and implicit causal designs in linear regression","authors":"Jiafeng Chen","doi":"arxiv-2407.21119","DOIUrl":"https://doi.org/arxiv-2407.21119","url":null,"abstract":"When do linear regressions estimate causal effects in quasi-experiments? This\u0000paper provides a generic diagnostic that assesses whether a given linear\u0000regression specification on a given dataset admits a design-based\u0000interpretation. To do so, we define a notion of potential weights, which encode\u0000counterfactual decisions a given regression makes to unobserved potential\u0000outcomes. If the specification does admit such an interpretation, this\u0000diagnostic can find a vector of unit-level treatment assignment probabilities\u0000-- which we call an implicit design -- under which the regression estimates a\u0000causal effect. This diagnostic also finds the implicit causal effect estimand.\u0000Knowing the implicit design and estimand adds transparency, leads to further\u0000sanity checks, and opens the door to design-based statistical inference. When\u0000applied to regression specifications studied in the causal inference\u0000literature, our framework recovers and extends existing theoretical results.\u0000When applied to widely-used specifications not covered by existing causal\u0000inference literature, our framework generates new theoretical insights.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"24 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141864943","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
As service systems grow increasingly complex and dynamic, many interventions become localized, available and taking effect only in specific states. This paper investigates experiments with local treatments on a widely-used class of dynamic models, Markov Decision Processes (MDPs). Particularly, we focus on utilizing the local structure to improve the inference efficiency of the average treatment effect. We begin by demonstrating the efficiency of classical inference methods, including model-based estimation and temporal difference learning under a fixed policy, as well as classical A/B testing with general treatments. We then introduce a variance reduction technique that exploits the local treatment structure by sharing information for states unaffected by the treatment policy. Our new estimator effectively overcomes the variance lower bound for general treatments while matching the more stringent lower bound incorporating the local treatment structure. Furthermore, our estimator can optimally achieve a linear reduction with the number of test arms for a major part of the variance. Finally, we explore scenarios with perfect knowledge of the control arm and design estimators that further improve inference efficiency.
{"title":"Experimenting on Markov Decision Processes with Local Treatments","authors":"Shuze Chen, David Simchi-Levi, Chonghuan Wang","doi":"arxiv-2407.19618","DOIUrl":"https://doi.org/arxiv-2407.19618","url":null,"abstract":"As service systems grow increasingly complex and dynamic, many interventions\u0000become localized, available and taking effect only in specific states. This\u0000paper investigates experiments with local treatments on a widely-used class of\u0000dynamic models, Markov Decision Processes (MDPs). Particularly, we focus on\u0000utilizing the local structure to improve the inference efficiency of the\u0000average treatment effect. We begin by demonstrating the efficiency of classical\u0000inference methods, including model-based estimation and temporal difference\u0000learning under a fixed policy, as well as classical A/B testing with general\u0000treatments. We then introduce a variance reduction technique that exploits the\u0000local treatment structure by sharing information for states unaffected by the\u0000treatment policy. Our new estimator effectively overcomes the variance lower\u0000bound for general treatments while matching the more stringent lower bound\u0000incorporating the local treatment structure. Furthermore, our estimator can\u0000optimally achieve a linear reduction with the number of test arms for a major\u0000part of the variance. Finally, we explore scenarios with perfect knowledge of\u0000the control arm and design estimators that further improve inference\u0000efficiency.","PeriodicalId":501293,"journal":{"name":"arXiv - ECON - Econometrics","volume":"73 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141865026","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}