Encouraging Healthy Nutrition Purchases in an Online Grocery Setting using Experimental Economics: Implicit versus Explicit Taxes

Last registered on November 21, 2018

Pre-Trial

General Information

Title

RCT ID

AEARCTR-0003176

Initial registration date

July 26, 2018

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published

July 27, 2018, 1:26 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated

November 21, 2018, 1:52 AM EST

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Country

Singapore

Region

Primary Investigator

Name

Eric Finkelstein

Affiliation

Duke-NUS Medical School

Contact Primary Investigator

Other Primary Investigator(s)

Additional Trial Information

Status

In development

Start date

2018-08-03

End date

2018-11-30

Keywords

Health

Additional Keywords

Taxes, obesity, salience, implicit, explicit, messaging, front-of-pack, labeling, food purchasing, nutrition, groceries, online, diet

JEL code(s)

Secondary IDs

Abstract

The link between poor diet and rising rates of obesity and chronic disease is well documented. In efforts to improve diet quality, targeted sin taxes on sugar sweetened beverages and/or other unhealthy products have been levied in locales worldwide.

Research has revealed that the way taxes are implemented can influence consumer behavior. Sin taxes on selected ingredients usually raise market prices of products containing those ingredients. However, the taxes may not be salient and/or clearly identifiable to consumers. Similarly, taxes that are only added at the time of payment, as opposed to being added to price tags on shelves, may be less effective.

Since a highly salient and clearly delineated sin tax provides a signal to consumers that the government considers the taxed foods to be less healthy, sin taxes may be effective in reducing consumption even if suppliers were to absorb the tax and not raise prices. However, consumers who are not supportive of the tax may be least likely to moderate their purchases and could even increase purchases as a form of protest.

In light of the above, we propose to conduct a four arm randomized controlled trial via novel web-based grocery store and a 20% tax to test the following hypotheses:

1. Targeted sin taxes that result in higher prices will reduce demand for taxed products.
2. The effectiveness of a tax is greater if the tax is salient and clearly delineated.
3. A salient and clearly delineated sin tax will reduce demand for the taxed products even if there is no price increase.
4. Demand responses for salient and clearly delineated taxes will be moderated by the level of support for the tax, with those who actively support (oppose) the tax showing greater (smaller) reductions in purchases than when the tax is not prominently displayed.

External Link(s)

Registration Citation

Citation

Finkelstein, Eric . 2018. "Encouraging Healthy Nutrition Purchases in an Online Grocery Setting using Experimental Economics: Implicit versus Explicit Taxes." AEA RCT Registry. November 21. https://doi.org/10.1257/rct.3176-2.0

Former Citation

Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Experimental Details

Interventions

Intervention(s)

This randomized controlled trial relies on a fully functional and customizable web grocery store with over 4,000 products. We will tax the 20% of products (that we assume is fully passed through) within each food category (excluding fresh fruits and vegetables which will go untaxed) that are highest in terms of calories per serving. The four arms are:
1) No taxes (Control),
2) Implicit taxes (IT) showing only the post-tax price on the taxed products,
3) Explicit taxes (ET) showing the post-tax price and a label that says “Price includes a 20% ‘high in calories’ tax”, and
4) Fake taxes (FT) showing the same label as in ET but using Control prices for all products.

Intervention Start Date

2018-08-03

Intervention End Date

2018-11-30

Primary Outcomes

Primary Outcomes (end points)

Proportion of taxed products purchased (or targeted for tax if in control arm).

Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)

Secondary Outcomes:
1. Calories per serving (kCal per serving)
2. Calories per dollar (kcal/$)
3. Total Spend ($)
4. Expenditure on taxed products (or targeted for tax if in control arm) ($)
5. Diet quality per shopping trip, as measured by the Alternative Healthy Eating Index.

Secondary Outcomes (explanation)

Secondary Outcomes:
1. Calories per serving (kCal per serving). It will be calculated as the weighted average of number of servings per product purchased.
2. Calories per dollar (kcal/$) will be calculated since labels may simultaneously influence diet quality, cost and purchase volume.
3. Total Spend ($) will be calculated as the sum cost of all products in the basket.
4. Expenditure on taxed products (or targeted for tax if in control arm) ($) will be calculated as the sum cost of products targeted for the tax
5. Diet quality per shopping trip, as measured by the Alternative Healthy Eating Index. We propose to use the AHEI-2010 model.

Experimental Design

For this four-arm RCT, the NUSMart Online Grocery Store will be used (see Preliminary studies) for data collection. Participants will only be recruited if they are Singapore residents aged 21 years or older. This will be a once-off shopping experience with each participant exposed to one of the four shopping conditions (arms).

The 4 arms are:
1) No taxes (Control),
2) Implicit taxes (IT) showing only the post-tax price on the taxed products,
3) Explicit taxes (ET) showing the post-tax price and a label that says “Price includes a 20% ’high in calories’ tax” the price has been increased by 20% as a result of a sin tax aimed to control rising rates of obesity as shown in Figure 3, and
4) Fake taxes (FT) showing the same label as in ET but using Control prices for all products.

We will tax the 20% of products within each food category (excluding fresh fruits and vegetables which will go untaxed) that are highest in terms of calories per serving. Although our primary outcome will focus on the proportion of products purchased that are the target of the tax, secondary outcomes will focus on measures of overall diet quality, including calories per serving given that this is the ultimate objective of such a tax, and other measures of diet quality to test for unintended consequences.

While participants will be made aware that there are different versions of NUSMart, the exact details of the arms will not be revealed to ensure validity of results. It will be deleterious to the study to reveal the differences between study arms as that would likely confound participant responses.

Participants will be requested to shop as they would normally shop in a typical grocery shopping experience and spend a minimum of 50 Singapore dollars and maximum of 100 Singapore dollars. This may take them between 10 – 30 minutes to complete. The consent form will state that each shopping trip has up to a 50% chance of requiring the shopper to purchase the groceries (which is technically accurate). Participants will be told that if they are required to purchase, they will be directed to the NUS Online Collection Facility to pay with their credit card. The study team will then repurchase the foods and contact participants to confirm delivery of the groceries to their homes. We do this to ensure participants shop under the assumption that they may have to purchase the foods selected. This lends credibility to the results. Participants will also be told that if they complete the shop and take a brief survey, they will then receive a Lazada e-Voucher worth $20 that can be redeemed for any product on Lazada. This approach will allow us to quickly and cost effectively gather the necessary information to test our hypotheses.

Those who consent will then take a brief baseline questionnaire (see Baseline Questionnaire), be sequentially randomized to one of the four arms and directed to the corresponding version of the store. After completing the shopping experience they will be asked to checkout and spin a digital wheel on NUSMart to determine whether the purchase is to be made. The wheel will land on ‘No Purchase’, following which participants will be asked to participate in a brief survey about their shopping experience to conclude their participation. The survey should take approximately 5 minutes to complete.

Outcome measures
All outcome measures will be calculated using the sales orders submitted by participants via the on-line system.

Experimental Design Details

Randomization Method

Randomization done in office by a computer program.

Randomization Unit

Individual

Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters

940 individuals who will each shop once.

Sample size: planned number of observations

940 observations.

Sample size (or number of clusters) by treatment arms

235 per arm.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Although our primary outcome is the proportion of products purchased that are the target of the tax, because the ultimate objective is to quantify differences in calories per serving across arms, we powered the trial to ensure our ability to test hypotheses using this outcome, which requires a larger sample size than the test of proportions given the greater variation in this variable. We will test for differences in calories per serving using an alpha of 0.05, power of 0.8, and effect size of 0.30, a relatively small effect. We will also limit post hoc comparisons to 4 (Control vs. Fake Tax, Explicit Tax vs. Fake Tax, Implicit Tax vs. Fake Tax, and Explicit Tax vs. Implicit Tax). Our prior studies showed that the mean and standard deviation in kCal/serving is 139 and 65 respectively. This means that we are powering the study to detect a difference of roughly 20 kCal/serving across arms (20/65 = approx. 0.30). Based on these values, we require 235 participants per arm or 940 in total.

Supporting Documents and Materials

IRB