Jayson Jia and Kunal Patel

Jayson Jia and Kunal Patel

Buffet vs. Retail Menu Constraints: A Study of Preferences Under Constraint

Experiment Summary

Introduction

This study sets about to examine whether the imposition of budget constraints can sometimes result in higher levels of consumption than if there were no constraints in the first place. Overall, we were curious as to how the modification of a consumption constraint changes a consumer’s consumption patterns.

The intuition of this study is based on our observations of real life economic phenomenon at the Yale Law school cafeteria as opposed to the Yale College dining halls. In the buffet-style College dining halls, students ate as much as they felt like and seemed to be maximizing individual and differentiated utility functions (people took everything from a waffle to multiple plates). However, in the retail-style Law School, consumption was almost uniform as all students tried to reach the consumption level set by the budget constraint. For the people who don’t usually eat much in the dining hall, this means that the imposition of a constraint in the Law School caused them to consume more than usual! Our class experiment tried to hold this situation as a guiding beacon and examine the forces that caused by this phenomenon in Yale’s natural experiment. Although there were some biases (which we discuss later in our discussion section), we tried to keep the situation as similar as possible – albeit with reasonable simplifications – in order to fully capture the dynamics of consumer behavior.

Theoretical Background

Our initial intuition was that the effect would be caused by the psychological impact of loss aversion or some kind of information effect from the presence of a budget constraint. Based on the predictions of prospect theory (Kahneman and Tversky 1979, Tversky and Kahneman 1991), we believe that changing the budget constraint changed consumer’s reference point. We would like to note that the budget constraint itself is not a reference point. Rather, consistent with Kahneman and Tversky 1991, it is the best package available for consumption that acts as the reference point. In our experiment, the best package for consumption became better with higher budget constraints, and worse with smaller budget constraints. This causes a reference effect that causes feelings of loss aversion in situations which are framed as losses (so the shift in reference point leaves current consumption in the loss frame). Since customers are thus at a lower point in the value function, they experience disutility. This in turn may motivate them consume more.

Our experiment also touches on issues of mental accounting (Thaler 1985, 1990, 1999). In particular, people’s transaction utility may be affected by the different budget constraints. Some budget constraints were essentially better deals than others. We expected that the worse deals would elicit greater consumption as a response. We tried to take the predictions of mental accounting in account by limiting the prediction to one non-fungible mental account: dining out. However, it is also possible that people started splitting their budgets into different accounts. For example, one of the subjects pointed out during our presentation that he had created a sort of ‘sushi account’ that was particularly insatiable.

Experiment

The experiment was a simple excel-based menu exercise in which people were asked to simulate dining activity in different scenarios. We assumed an i.i.d. population, non durable goods (consume or throw away), the existence of one venue and one time period for each treatment.

The experiment was divided into two separate items: the pure budget effect treatment and the information treatment. The former examines the effects of imposing budget constraints on consumption levels. The latter examines the same thing except with the full knowledge of other people’s consumption levels. Two types of information were released: 1) efficiency of other consumers in spending their budget and 2) level of absolute spending. We were hoping that this treatment would provide an information effect. If there were changes in the level of consumption, this would have shown that the information effect was palpable.

Basic Treatments:

·  Scenario 0: buffet – tell people to ‘eat’ as much as they would normally like– don’t tell people how much they paid – see how much they would like to consume – use this to calculate x

·  Scenario 1: buffet where people are told they paid $x

·  Scenario 2: impose a constraint on consumption at $x, for which you pay $x

·  Scenario 3: pay $x for a constraint y, where y<x (gain frame)

·  Scenario 4: pay $x for a constraint z, where z>x (loss frame)

Scenario 0 was intended to see how much people would like to eat if no constraints existed. Essentially, it asked people to maximize their utility without the presence of price or budget.

Scenario 1 is different from scenario 0 in that it has a cost attached to it. We split scenario 1 into two sub-treatments: one in which the market price of food is unknown and one in which it was known.

Scenarios 2-4 were the retail situations in which the costs of dining are equivalent to the buffet. The cost is determined by how much was consumed in Scenario 0. Scenario 2 was supposed to be equivalent to Scenario 0 in that you can consume as much as you want in an utopian situation (albeit at a cost). Scenarios 3 and 4 were meant to be ‘good’ and ‘bad’ deals respectively. In Scenario 3, you receive a higher budget (+$2 and +$4 in our experiment) for the same price, while in Scenario 4, you receive a smaller budget (-10% and -20% in our experiment) for the same price. The increase and decreases in budget were set up in such a way that your budget would change substantially even if you only had a bottle of water in scenario 0 (from which your budget is determined).

Results

Overall results showed that people tried to consume at the budget constraint whenever possible. Consumption relative to budget available ranged between 97% and 99% for all cases.[1] Although spending in some constrained situations were higher than that in Scenario 0 (the ideal world in which you can eat as much as you want for no cost), buffet consumption was much higher than retail consumption. Regardless, it seemed that people started consuming a lot more once they had to pay for something.

Due to our small sample size, buffet results were heavily skewed by a couple of outlier results. Indeed, it may be more informative to examine the behavior of the few ‘underconsumers’ (those who didn’t have much in the buffet situations but markedly increased consumption once constraints were imposed). Consistent with intuition, the people were planned to eat a lot at buffets would not have been able to consume as much when constraints were imposed. However, those who ate around the same in buffet situations as they did in the ideal world did increase their consumption in the constrained retail situation. Thus, some people at more in a constrained situation than they did for buffets and ideal-world situations, which is exactly the type of behavior we were looking for.

Discussion

Although the experiment was not large enough in scope for us to make claims of great scientific validity, we did isolate the kind of behavior that we were looking for. Namely, there are cases where more constraints lead to higher consumption for some people.

Our results can also be used to make policy recommendations for the Yale dining hall. Overall, consumption was lowest for the ‘ideal world’ situation (scenario 0). Some people ate more in the buffet situation, and others ate more in the retail situation. We do not have enough data on Yale students to say which type predominates. However, it seems that more liberalized dining policies probably result in more socially considerate levels of consumption.

Since people currently pay for their meals with swipes from an electronic card (which may not be psychologically framed as payment since it’s prepaid by parents – although this needs further investigation), scenario 0 may be the best operationalization of a Yale College dining hall. It is informative that people consumed only the amount they needed in a situation without any constraints whatsoever. We would also hypothesize that satisfaction is higher for the unconstrained situations compared to the constrained situations (although this would require more studies).

Limitations and Future Improvements

There were several biases and unforeseen experimental difficulties in our project. Most importantly, the effects of each treatment were impossible to separate due to the huge change in consumption after the first round. The impact of setting a price (in scenario 1) was much stronger than we anticipated: so much so that we could no longer distinguish the effects of different budget constraints. To solve this problem, we could increase the changes in budget from around ±10% and ±20% to ±25% and ±50%.

One bias of the experiment was the allowance of diversity-seeking behavior. Since there was a large selection, people found it difficult to be satiated since they could just buy more of something they had not tried before. While we considered running a one-commodity experiment (e.g. tokens, emission credits, etc), we decided against it in order to better capture the natural experiment we saw taking place in the Yale law school cafeteria. In order to control for other unforeseen effects, we could have switched the order of the experiment. Indeed, the order may have an effect by setting different kinds of consumption policies as a point of reference.

To make the experiment more interesting, we could also have measured overall utility for each treatment. This would have been more informative in suggesting policy changes. For example, although the lowest constraints obviously limited consumption the most, the fall in consumer satisfaction would probably have been so high that dining hall management would not want to enact such policies.

Appendix A

Scenario / Consumption ($) / Consumption relative to price/ budget (%)
Ideal World / 12.80 / n/a
Buffet - No Price / 21.73 / 1.82
Buffet - Price / 22.35 / 1.85
Retail x+$0 / 12.65 / 0.99
Retail x+$2 / 14.55 / 0.99
Retail x+$4 / 16.65 / 0.99
Retail x-10% / 9.93 / 0.97
Retail x-20% / 7.45 / 0.97

Table 1

Total Consumption ($)
Scenario / Situation 2: Revealed Percentage Spending / Situation 3: Revealed Absolute Spending
Retail x+$0 / 12.70 / 12.72
Retail x+$2 / 14.67 / 14.68
Retail x+$4 / 16.47 / 16.47
Retail x-10% / 10.02 / 9.98
Retail x-20% / 7.45 / 7.42

Table 2

[1] See Appendix A for detailed results.