Bayesian Synthetic Control Methods

Kim, Sungjin; Lee, Clarence; Gupta, Sachin

doi:10.1177/0022243720936230

Cited by 41 publications

(6 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The synthetic control method (SCM), developed by Abadie and Gardeazabal (2003) and Abadie et al (2010Abadie et al ( , 2015, uses a weighted average of outcomes of the controls to construct the counterfactual of treated units. Since its inception, a number of counterfactual models has been developed to extend it to multiple treated units (Ben-Michael et al, 2021a), staggered treatment (Gobillon and Magnac, 2016;Xu, 2017;Ben-Michael et al, 2021b), treatment reversal (Liu et al, 2021), nonparametric estimation of weights (Cerulli, 2019), spatial spillovers (Cao and Dowd, 2019;Grossi et al, 2020), and the Bayesian realm (Kim et al, 2020;Pang et al, 2022).…”

Section: Augmented Synthetic Control Methodsmentioning

confidence: 99%

Effects of Covid-19 lockdowns on social distancing in Turkey

Bilgel

2022

The Econometrics Journal

View full text Add to dashboard Cite

This paper elucidates the causal effect of lockdowns on social distancing behaviour in Turkey by adopting an augmented synthetic control and a factor-augmented model approach for imputing counterfactuals. By constructing a synthetic control group that reproduces pre-lockdown trajectory of mobility of the treated provinces and that accommodates staggered adoption, the difference between the counterfactual and actual mobility of treated provinces is assessed in the post-lockdown period. The analysis shows that in the short-run following the onset of lockdowns, outdoor mobility would have been about 17 to 53 percentage points higher on average in the absence of lockdowns, depending on social distancing measure. On the other hand, residential mobility would have been about 12 percentage points lower in the absence of lockdowns. The findings are corroborated using interactive fixed-effects and matrix completion counterfactuals that accommodate staggered adoption and treatment reversals.

show abstract

Section: Augmented Synthetic Control Methodsmentioning

confidence: 99%

Effects of Covid-19 lockdowns on social distancing in Turkey

Bilgel

2022

The Econometrics Journal

View full text Add to dashboard Cite

show abstract

“…Most studies document a sizable pass-through, between 59% and 103%, resulting in significant price increases for consumers (Cawley et al 2021) and significant drops in sales of taxed soda. Seiler, Tuchman, and Yao (2021) report a 46% reduction in volumes sold in Philadelphia; Powell and Leider (2020) find a 22% reduction in Seattle; Kim, Lee, and Gupta (2020) document a 6% reduction for Washington's short-lived soda tax; and mixed results emerge for Berkeley, including decreases ranging from 20% (Silver et al 2017) to 9% (Taylor et al 2019), though Rojas and Wang (2021) find no evidence of any effect. Extant research considers single locations and confounds regular price and promotional changes by assessing the price paid only, disregarding that consumers react differently to regular price and promotional price (Datta, Ailawadi, and Van Heerde 2017).…”

Section: Conceptual Foundationmentioning

confidence: 99%

“…The former includes tax introductions (e.g., soda taxes; e.g.,Seiler et al 2021;Kim et al 2020), the state of the economy (which includes raw material cost changes and its effect on marketing spending;Lamey et al 2012), and popular events (e.g., the Olympics;Keller, Deleersnyder, and Gedenk 2019). The latter captures events such as national-level boycotts (e.g.,Sun et al 2021) or industry-wide product-harm crises (e.g., Cleeren, Dekimpe, and Van Heerde 2017).6 It is important to recognize that we do not model retailer and soda brands as distinct strategic actors.…”

mentioning

confidence: 99%

Soda Taxes and Marketing Conduct

Keller¹,

Guyt²,

Grewal³

2023

Journal of Marketing Research

View full text Add to dashboard Cite

Soda taxes are an increasingly popular policy tool that discourages purchases of sugar-sweetened beverages. This study analyzes how emphasis on marketing conduct elements and their effectiveness might change after soda tax introductions. Prior studies on the effect of soda taxes focus on price increases but neglect other, relevant marketing conduct tools, i.e., promotional frequency, promotional discount depth, and feature promotion frequency. This study documents the changes to marketing conduct elements and their effectiveness due to the introduction of soda tax across more than 200 retail stores in five markets. Findings related to price changes are consistent with prior literature; in addition, the study reveals a substantial, hitherto overlooked decrease in promotional frequency (-2%), promotional depth (-12%), and feature promotion frequency (-14%) compared with matched control markets, exacerbating the tax’s negative sales effect. Introducing a soda tax also considerably influences the marketing conduct’s effectiveness, such that consumers become less sensitive to changes in regular price, feature promotions, and the depth of the promotional discount but respond more to the presence of promotions. Importantly, marketing conduct and effectiveness changes do not align (e.g., while consumers become more sensitive to promotion frequency, managers often reduce them), a relevant insight for policymaking.

show abstract

“…A number of marketing studies make use of quasi-experimental data to estimate causal effects. Examples include studies of the effect of bookstore openings on sales at Amazon (Forman, Ghose, and Goldfarb 2009), the effect of TV advertising on online shopping (Liaukonyte, Teixeira, and Wilbur 2015), the effect of offline stores on online sales (Wang and Goldfarb 2017), and the effect of soda taxes on price and consumption (Kim, Lee, and Gupta 2020).…”

mentioning

confidence: 99%

“…Hahn and Shi (2017) show that the validity of any permutation-based approach requires a strong symmetry assumption that is unlikely to hold in quasi-experimental data and the violation of which leads to biased confidence intervals. Recently, some authors have suggested Bayesian alternatives to SC for regularization approaches (Kim, Lee, and Gupta 2020; Pang, Liu, and Xu 2022). The Bayesian approach solves the inference problem by allowing for the construction of credible intervals (of any quantity of interest) via simulation draws from the posterior distribution of model parameters.…”

mentioning

confidence: 99%

Statistical Inference for the Factor Model Approach to Estimate Causal Effects in Quasi-Experimental Settings

Li¹,

Sonnier²

2023

Journal of Marketing Research

View full text Add to dashboard Cite

Causal inference using quasi-experimental data is of great interest to marketers. The factor model approach (FM) to estimate treatment effects accommodates a large number of control units and can easily handle a large number of treatment units while flexibly allowing for cases where the treatment is outside the range of the control units. However, the factor model method lacks formal inference theory, instead relying on bootstrap or permutation procedures with strong assumptions. Specifically, the extant Xu (2017) bootstrap procedure requires that the treatment and control error variance are equal. We establish that when this assumption is violated, the bootstrap procedure results in biased coverage intervals. We develop formal inference theory for the factor model approach to estimate the average treatment effect on the treated (ATT). Our approach allows researchers to formally quantify uncertainty through hypothesis testing and confidence intervals. Our inference method is applicable to both stationary and non-stationary data. More importantly, our inference theory accommodates treatment and control unit outcomes with different distributions, which includes different error variances as a special case. We show the performance of our inference theory in simulated data. Finally, we apply our method to empirically quantify the uncertainty in the effect of legalizing recreational marijuana on the beer market and the sales effect of a digitally native online brand opening a physical showroom.

show abstract

Bayesian Synthetic Control Methods

Cited by 41 publications

References 56 publications

Effects of Covid-19 lockdowns on social distancing in Turkey

Effects of Covid-19 lockdowns on social distancing in Turkey

Soda Taxes and Marketing Conduct

Statistical Inference for the Factor Model Approach to Estimate Causal Effects in Quasi-Experimental Settings

Contact Info

Product

Resources

About