x

NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
The Welfare Effects of Persuasion and Taxation: Theory and Evidence from the Field
Last registered on August 13, 2020

Pre-Trial

Trial Information
General Information
Title
The Welfare Effects of Persuasion and Taxation: Theory and Evidence from the Field
RCT ID
AEARCTR-0002814
Initial registration date
March 28, 2018
Last updated
August 13, 2020 5:36 AM EDT
Location(s)
Region
Primary Investigator
Affiliation
University of Muenster
Other Primary Investigator(s)
PI Affiliation
University of Münster
Additional Trial Information
Status
Completed
Start date
2018-03-29
End date
2019-08-31
Secondary IDs
Abstract
A small body of the recent literature in economics analyzes how the negative welfare effects of behavioral biases can be countered by taxes. While there is an ongoing debate about whether governments should address behavioral biases by taxes or nudges (e.g. sugar taxes vs. food labels), little is known about how these policy tools perform relative to each other. One challenge for such a comparison is the limited evidence of how nudges affect choices and welfare.

We intent to contribute to this question by analyzing the welfare effects of information nudges. A particular feature about most of the information nudges employed in the real world is that they only coarsely inform consumers about the relevant topic. We model information nudges as noisy signals that coarsely inform consumers about the uninternalized consequences of a choice. Our model provides sufficient statistics to analyze the welfare effect of a coarse nudge in comparison to a more specific nudge and an internality tax.

The theoretical model guides our design of a large-scale natural field experiment in which we look at the case where the internality distortions are caused by biased beliefs about the benefits of energy efficiency. We construct two nudges that differ in their degree of coarseness and compare its effect on consumer behavior. By combining these treatments with exogenous price variation for different lighting technologies, we are able to quantify the welfare effects of nudges with different levels of coarseness and compare it to a subsidy on energy-efficient light bulbs.

Our experimental design further incorporates a novel approach to control for bias heterogeneity post-treatment in a between-subject design. This approach circumvents the need for within-subject designs increasingly used in behavioral welfare economics.
External Link(s)
Registration Citation
Citation
Rodemeier, Matthias and Andreas Löschel. 2020. "The Welfare Effects of Persuasion and Taxation: Theory and Evidence from the Field." AEA RCT Registry. August 13. https://doi.org/10.1257/rct.2814-6.0.
Former Citation
Rodemeier, Matthias and Andreas Löschel. 2020. "The Welfare Effects of Persuasion and Taxation: Theory and Evidence from the Field." AEA RCT Registry. August 13. http://www.socialscienceregistry.org/trials/2814/history/73854.
Experimental Details
Interventions
Intervention(s)
Intervention Start Date
2018-03-29
Intervention End Date
2018-05-31
Primary Outcomes
Primary Outcomes (end points)
The number of LED, CFL, halogen and incandescent light bulbs purchased.

The average energy efficiency level (wattage) of the products purchased.

The price elasticity for all four light bulb technologies (incandescent, halogen, CFL, LED).

We also estimate whether our interventions affected the extensive margin, i.e. the probability of purchasing a light bulb.
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
Visitors of an online shop are randomly assigned to different groups in a 3x5 between-subject design.

We vary both the prices of different lighting technologies and the information provided on energy efficiency. This experimental design facilitates the estimation of price elasticities for different lighting technologies under different sets of information provision.

After customers went through the purchasing decision they receive an invitation to participate in a survey. The survey elicits variables such as energy and financial literacy, income, patience, environmental attitudes and other consumer-specific characteristics.
Experimental Design Details
The experiment will be implemented in the online shop of a German appliance retailer. Potential customers who visit the website of the appliance retailer are randomly assigned to one of 15 groups with equal probability. We use a 3x5 design in which customers get randomized into 3 different informational groups (groups 1 to 3) and 5 different price discount groups (groups A to E). Visitors receive either 1) no information 2) coarse information or 3) specific information on the financial benefits of energy efficiency. In addition, every visitor is randomly assigned to a price discount on A) incandescent bulbs B) halogen bulbs C) CFL bulbs D) LED bulbs or E) does not receive a price discount. Group 1 Visitors in group 1 receive no additional information on the financial benefits of energy efficiency, other than the information already provided by the online retailer (e.g. wattage, energy efficiency label, etc.). Group 2 For subjects in group 2, a banner is shown at the top of the browser and contains information on the annual electricity savings of three lighting technologies (halogen, CFL, LED) in comparison to a traditional 40W incandescent light bulb. We visualize these savings by using a bar chart. In particular, subjects in group 2 are only informed about the savings of these light bulbs in percentage (e.g. a 4 LED saves 90% compared to the 40W incandescent). We do not inform subjects explicitly about how these relative savings in electricity costs translate into monetary savings. This information screen serves as our coarse-information nudge as it provides subjects with potentially useful information but leaves room for interpretation (e.g. it may be unclear to the consumer whether 90% savings translate into 2 or 200 Euros per year). Group 3 Subjects in group 3 receive almost the same banner as subject in group 2 but are also informed about the annual savings of the different lighting technologies in Euro. That is, besides receiving information on relative savings (in percent) we also tell them the absolute savings (in Euro). We explicitly tell the subject which electricity price and utilization of the light bulb we have assumed for calculating the monetary savings. The informational intervention in group 3 serves as our nudge with more specific information as it informs the consumer about both relative and absolute savings of different lighting technologies. Groups A-E Each of the three groups is divided into 5 subgroups (A-E) in which we offer subjects a 20% price discount on one of the four lighting technologies or no discount. Thus, we have 15 in groups in total: 1.A, 1.B, 1.C, 2.A, …, 3.E. The price discount is shown directly next to the informational intervention so that every subject wo has seen the information should also have seen the individual discount. The randomly assigned price discounts facilitate the estimation of price elasticities which are needed for our welfare analysis. After subjects made their purchase, they are invited to participate in a survey. Participation is incentivized through a lottery. The survey includes questions on the level of energy literacy, financial literacy, environmental attitudes, patience, education, age, gender and the subject’s zip-code. We use this survey to deal with the challenge of bias heterogeneity. The recent literature has used within-subject designs to elicit the bias (under-/ overvaluation of energy efficiency) for each subject individually. Since this is generally not possible in a natural field experiment with a between-subject design, we collect information on variables that have predictive power about the individual-specific treatment effect but are not affected by the treatment. This should allow us to identify the “bias type” of the consumer (e.g. high or low bias). We then interact our treatments with these variables and estimate heterogeneous treatment effects (of both the informational and the price interventions). To identify the bias type, we intent to use questions on the level of financial literacy, income, patience and some additional questions on energy literacy that are unlikely to be influenced by our intervention. As an alternative approach to using answers from the survey, we match the customers zip-code with available data on income and other relevant variables that potentially identify the bias-type of the consumer (e.g. green party vote share). Our sufficient statistic approach allows us to estimate the effect of our intervention on consumer surplus under a set of assumptions. For a more complete welfare analysis, we also take into account how our intervention could have affected environmental externalities (CO2 emissions) if subjects bought more or less efficient lighting technologies. We plan to take estimates on the social value of a ton of CO2 from the established literature.
Randomization Method
This experiment is between-subject. The retailer will use a professional A/B-testing tool from a third-party provider to randomly assign each visitor into one of 15 experimental groups with equal probability.
Randomization Unit
The randomization unit is the HTTP-Cookie of the visitor.
Was the treatment clustered?
No
Experiment Characteristics
Sample size: planned number of clusters
The treatments are not clustered.
Sample size: planned number of observations
Around 1.4 million observations.
Sample size (or number of clusters) by treatment arms
Around 93,333 individuals per treatment.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
University of Münster
IRB Approval Date
2018-02-12
IRB Approval Number
N/A
Post-Trial
Post Trial Information
Study Withdrawal
Intervention
Is the intervention completed?
No
Is data collection complete?
Yes
Data Collection Completion Date
Final Sample Size: Number of Clusters (Unit of Randomization)
Was attrition correlated with treatment status?
Final Sample Size: Total Number of Observations
Final Sample Size (or Number of Clusters) by Treatment Arms
Data Publication
Data Publication
Is public data available?
No
Program Files
Program Files
Reports, Papers & Other Materials
Relevant Paper(s)
Abstract
How much information should governments reveal to consumers if consumption choices have uninternalized consequences to society? How does an alternative tax policy compare to information disclosure? We develop a price theoretic model of information design that allows empiricists to identify the welfare effects of any arbitrary information policy. Based on this model, we run a natural field experiment in cooperation with a large European appliance retailer and randomize information regarding the financial benefits of energy-efficient household lighting among more than 640,000 subjects. We find that more informative signals strongly decrease demand for energy efficiency, while less informative signals increase demand. More information reduces social welfare because the increase in consumer surplus is outweighed by the rise in environmental externalities. By randomizing product prices, we identify the optimal tax vector as an alternative policy and show that sizable taxes on energy-inefficient products yield larger welfare gains than any information policy.
Citation
Rodemeier, Matthias, and Andreas L¨oschel. 2020. “The Welfare Effects of Persuasion and Taxation: Theory and Evidence from the Field.” CESifo Working Paper No. 8259.
REPORTS & OTHER MATERIALS