The Impact of Restaurant Menu Calorie Labeling on Food Choice

The experiment will be conducted at two sit-down restaurants. Parties of patrons will be randomized into the treatment group or control group. The treatment group will receive menus with calorie labels whereas the control group will receive the same menus without such calorie counts.

Data on each patron’s order will be taken by the server and recorded in the Point of Sales (POS) system already in place. Research assistants will record whether the table in question was in the treatment group (received the menus with the calorie labels) or the control group (received the menus without calorie labels). After the meal, patrons will be surveyed in order to collect information on their age, education, interest in health, and the frequency and extent of their participation in exercise. Patrons will also be asked whether their menus contained calorie information, whether they used that information, and whether they approve or disapprove of including calorie counts on menus. Research assistants will also collect information from servers on whether the patrons asked about or discussed the calorie labels.

The patron survey will also record whether the patron has dined at the restaurant since the experiment began. Such repeat customers will be dropped from the analysis sample.

2015-11-09

2017-09-30

The hypotheses we will test are as follows:
H1: Provision of calorie labels will result in patrons ordering fewer calories. We will test this hypothesis overall as well as within each course. We will also examine whether provision of information reduces the probability that patrons order any appetizer, or any dessert (i.e. the extensive margin of each course).

H2: Provision of calorie labels will result in patrons being more likely to choose the daily special; the reason is that daily specials are exempted from the calorie label requirement. The logic is that some consumers may exhibit optimism bias and hope that the daily special has fewer calories than the listed items, some consumers may be risk loving and gamble that the special has fewer calories, and some consumers may find the calorie labels disagreeable and deliberately choose not to know how many calories they are consuming. Only one of the two restaurants examined offers specials, however, and not even every day, so this will have limited power.

H3: Patrons will report seeing, using, and appreciating the presence of the calorie labels. This information will be collected in the patron survey.

H4: Provision of calorie labels will lower revenue and "profit" for the restaurant. For our measure of "profit", only food ingredient costs, not labor costs, can be considered.

This experiment takes place at a public restaurant. When patrons approach the maitre d', the entire party is randomized into treatment or control using a smartphone app. The treatment group receives menus with calorie counts, and the control group receives the usual menu without calorie counts. Orders are recorded by the server, and after the meal patrons are asked to take a survey.

smartphone app (Randomizer)

The randomization unit is the entire party (group of patrons dining together).

Yes

Approximately 2,000 "parties" (groups of restaurant patrons).

5,000

Roughly half treatment, and half control (exact proportions depend on randomization).

We assume that the control group orders an average of 1,400 calories at dinner, with a SD of 500. We assume that the treatment effect of providing calorie information is to reduce calories ordered by 5.8% (this is the effect found by Bollinger et al., (2011). Setting a 5% threshold for statistical significance, to have 80% power we need a sample size of 1,196.