Christmas Holidays and Gift Giving

As it is Christmas Day, today’s topic will be Christmas-themed. Although the question might seem light-hearted, it shows how economics can be used to inform us on any social question.

Holiday gift-giving is usually only relevant for economists and policy makers in the context of what impact it has on economy-wide spending, as this impacts the macroeconomy. However, we can also focus on the micro-economic scale, as gift-giving is a transfer between two people. A question that is often asked is whether you should give a gift or give cash to your gift-recipient, as they might not value the gift they receive. From a purely theoretical perspective, one can assume that the gift-giver aims to replicate what the gift-receiver would do with the money that is being spent on the gift. The gift-giver can sometimes choose an item that the gift-receiver values more than the cost of acquiring, due to imperfect information (the gift-giver knows something the receiver doesn’t). Sometimes, however, the value of the gift to the recipient can be below the price paid by the gift-giver due to a preference mismatch, generating what is called in economics deadweight loss (the value to the recipient is below the actual price of the gift). 

This question started off a chain of research and responses to the research attempting to determine whether gift-giving creates or destroys value. 

First Experiments

Joel Waldfogel studied this question in his 1993 paper “The Deadweight Loss of Christmas”.  Waldfogel conducted two experiments on his undergraduate economics students at Yale. He asked the first group what gift did they receive, from whom relationship-wise, and what is the maximum they would pay for the gift. In the other group, only the last question was changed to ask what is the minimal amount of money they would accept in lieu of the gift.

The result he found suggested significant deadweight loss – gift receivers value the gift between 10% to 33% below the amount that is paid for them. The “yield” of the gift, measured as value over cost, was between 66% and 90%. These are large numbers, especially when we take into account that just in the US, individuals expect to spend around $870 on gifts this Christmas. 

Waldfogel also looked into the ‘success’ rate of gifts based on the relationships between the gift-givers and gift-recipients. He found that the proximity of relationship matters – gifts from grandparents had the lowest value (a 62.9% yield), aunts and uncles (a 64.4% yield), while gifts from friends, siblings, and significant others had the highest value (a yield between 86 to 99%). This also, in part, explains why cash gifts in Waldfogel’s sample were more popular from grandparents (42.9% of the gifts were cash) and aunts and uncles (14.3% were cash), while cash gifts from friends, siblings, and significant others only accounted for 6% of their gifts.

The results of this paper stirred the economics community and quite a few follow-up experiments were conducted.

Experiment Replication

Solnick and Hemenway (1996) challenged Waldfogel’s results claiming that his sample was heavily biased.¹ In this case, Wadfogel’s sample consisted of relatively wealthy people, who were young (aged between 18-22) and were majoring in economics. The issue with the younger college population is that they were not yet supporting themselves and thus might have distorted perceptions. Furthermore, as they were already majoring in economics, the sample of students might also already be primed to believe that money would be better than the gift (due to learning about economic theories), and thus report their subjective responses to how much they would pay to match this theory.

Solnick and Hemenway, referencing a wide gift-giving literature, suggested that there are many potential reasons why gift-giving might generate positive value. To list a few: people value something more when someone else purchased it, due to self-management or other problems; the gift could have come from a child or grand-child, which would be perceived as far higher value; the gift could be something the person would not buy, but would gladly receive.

Solnick and Hemenway re-did the exercise as Waldfogel did, but their sample was much broader – besides college students, they also asked random members of the public in Boston and Philadelphia train stations. They found that the mean yield from the gift (i.e. the receiver's value of the gift over the cost) was at 214%. The gift-receivers valued their gifts significantly above the cost. More than half of the respondents’ value their gift above its cost, 21% value it at cost, while 28% valued it below cost. Solnick and Hemenway also found that the more the recipient liked and valued their gift, the more likely they also found it to be more sentimental. 

Additionally, controlling for whether something has sentimental value removed the effect Waldfogel found about the proximity of the relationship. Therefore, it was not important that the gift-giver was a sibling, friend, or significant other, but rather that the gift was sentimental (which of course correlates with whom you are receiving the gift from).

The results from Solnick and Hemenway’s research gave a contrasting result to the Waldfogel estimate. So another pair of economists tackled the question.

A Better Estimate

List and Shogren (1998) continued the research into the topic, as they argued that the Solnick and Hemenway paper had an important flaw – because they only collected survey results and thus, the numbers were hypothetical in nature, the survey responses might not be a good metric of the actual value of the received gifts. This is a type of flaw that is found in all surveys.

List and Shogren tackled this problem by undertaking a multi-step experiment on a sample of undergraduates. They undertook three experiments:

1. They asked the participants to list all their gifts that they received and asked them to value all their gifts.

2. Two days later, they asked the participants to value their gifts under a hypothetical random nth-auction scenario. In this scenario, the participants were asked to put a value to each of their gifts. Next, all the gifts from each participant would be pooled together and ranked from lowest value to highest. The auctioneer would then draw a random number n, and all cheaper gifts than gift n would be bought by the auctioneer at the price of the nth gift. For example, suppose there were 20 gifts, if the number 5 is selected, the first four cheapest gifts would each be bought at the price of the 5th gift. Therefore, this setup gives everyone strong incentives to give a genuine value for their gift, given that you can actually make profit in this setup. However, if you over-value your item, it is unlikely it will be ranked low, and thus the gift has a low chance of being selected to be bought.

3. Two days after the hypothetical auction, they conducted a real nth auction about which the participants were not informed about prior. The authors made it clear to the participants that the gifts will be bought by them, without the ability to decline by either party. At the end of the experiment, List and Shogren ended up buying three items for $2.00 each, which were “hosiery, slippers, and a beer intake facilitator”.

From this set of experiments, what List and Shogren found is that, firstly, when real money is being offered (scenario 3), people value their items differently (scenario 1). They found that in the surveys, the participants significantly undervalued their gifts compared to the auction scenarios. Secondly, they found that the yield of the gifts came out at around 121% to 135%. That is, the participants valued the gifts at more than the cost by 21% to 35%.² 

The results of List and Shogren show that gift giving does generate additional value, which matches the findings of Solnick and Hemenway, albeit a much smaller gain. However, Waldfogel results were also not wrong – due to the nature of surveys, even in the List and Shogren data, when participants were asked to value their items in the first step, they undervalued their gifts at around a yield of 97% to 99%, suggesting that Waldfogel survey results can occur.

Why is this important?

The above three papers have shown how tackling the problem with different methodologies can lead to very different results. Ruffle and Tykocinski (2000) also decided to get it in on the fun of analyzing the valuation of gifts. Ruffle and Tykocinski conducted a non-Christmas based experiment, where they showed an item to participants and told them to assume it was a gift from someone they knew. They were asked about the value of the item to them and the cost they thought was paid for it. Naturally, the values were below the cost as there wasn’t any sentimental nor even much utility value of the gifts to the participants. However, what Ruffle and Tykocinski wanted to elucidate was whether priming played an important role. When asking about the value of the gift, they asked one group of participants what amount of money would make them equally happy to have instead of the gift, while another group was asked what amount of money would make them indifferent. Although these questions are asking the same thing, the words being used have strong emotional differences. Unsurprisingly, the equally happy group valued the items more than the indifferent group, but the magnitude of the difference was startling - the equally happy valued the item at 50% more.

Another experiment Ruffle and Tykocinski conducted was whether the order of questions matter – whether you ask about the value of the gift to you or the cost of the gift first, and if you ask these questions jointly. They did not find any strong results, suggesting in this instance, the question order did not matter. Lastly, as a little test of Solnick and Hemenway's idea that economists might be a biased group for experiments, Ruffle and Tykocinski conducted their experiment on a group of psychology students and economics students. Interestingly, they did not find any differences between their personal values of the gifts economists and psychologists assigned, but they did find that the economists gave lower cost estimates (economists assumed the cost was 10% lower than what psychologists thought). Thus, somewhat in jest, Ruffle and Tykocisnki argue that economists get a greater yield from gifts (value-to-cost is higher), but maybe giving them an expensive gift is not the wisest choice.

Conclusion

Economics can be used to tackle any social question and generate very valuable insights. The value of gift-giving is something that may have seemed not possible to evaluate using a rigorous approach. In this sequence of papers, we learned that this question can be addressed and a result can be established. The result is that there is evidence showing that gift-giving does generate value. Furthermore, sentimental value is separate from relationship proximity (it is strongly correlated), while phrasing of questions can result in different answers.

What started out as a light-hearted research paper by Waldfogel, resulted in a lot of interesting questions on methodologies and approaches to economics questions. It is important to keep in mind that certain approaches can seem appropriate, but under closer inspection, the conclusions are erroneous.

Addendum: While looking for research on Christmas topics, I bumped into multiple articles and analysis focusing on Waldfogel’s research (Wired, New Republic, Marginal Revolution), while Waldfogel himself wrote a book on the topic of value destruction due to gift giving. It is worth noting that these sources do not discuss the follow-up findings at all or only briefly mention that other research exists, although those results showed the opposite conclusion.

1. In statistics, bias refers to the idea that in a random sample, certain members of the intended population (in this case the intended population is everyone in the US) have a lower or higher probability of selection into the sample. As the only people in the sample are undergraduate economics majors at Yale, this group is overrepresented in the sample.

2. Note that the participants could have easily just selected the cost of the item as their value (or slightly above), because they could have just re-purchased the item from the store if the auctioneer would have ended up purchasing it from them.