VoxEU Column Productivity and Innovation Taxation

Around a table: Social facilitators for negotiations

Pre-negotiation interactions, such as shared meals, are viewed as a valuable means to build trust and rapport so as to improve the outcomes of the negotiation. Even tax authorities acknowledge this – business meals tend to be tax-deductible, at least in part. This column puts this folk wisdom to the test using a controlled negotiation simulation experiment with MBA students. It finds that there is no difference in negotiation outcomes whether or not pre-negotiation socialising took place.

Negotiations are central to economic life. Both business and government use them, and they also come up frequently in the personal and social spheres. They are hugely important but also quite stressful. Brooks and Schweitzer (2011) have shown that “compared to negotiators experiencing neutral feelings, negotiators who feel anxious expect lower outcomes, make lower first offers, respond more quickly to offers, exit bargaining situations earlier, and ultimately obtain worse outcomes.” For this reason, it is natural that expert negotiators have looked for tools to diminish the impact of negative emotions and to generally facilitate negotiations. 

One such tool has been pre-negotiation interactions, often in the form of a shared meal. Harvard’s Program on Negotiation,  one of the most prestigious training programmes for this field in the world, is clear on this point: “The reciprocal nature of trust reinforces the value of taking time to get to know the other party and build rapport before you begin to negotiate. Don’t assume that you can form a bond simply by exchanging a few friendly e-mails before meeting in person. Rather, try to forge a personal connection by meeting for an informal lunch or two.” 

Policymakers seem to agree, even such notoriously suspicious outfits as the tax authorities: business meals tend to be tax-deductible, at least in part. The US Internal Revenue Service considers that 50% of such expenses are deductible in general.  The UK HM Revenue and Customs (HMRC) allows the deduction of the part of the expense that is “wholly and exclusively” for the purpose of generating profits.  

Given this folk wisdom, it is rather surprising that there is little scientific evidence to back it up. Some of it is good, if a bit indirect. Cooper et al. (1992), Charness (2000), or Clark et al. (2001) show that ‘cheap talk’ (costless, non-binding pre-play communication) increases efficiency in coordination games. Palfrey and Rosenthal (1991) have done the same for public good games. There are also some results in bargaining games that are closer to our object of interest. Valley et al. (2002) studied a double oral auction with and without communication. They found that communication allowed to achieve higher levels of efficiency than predicted by theory. Forsythe et al. (1991) showed that communication during a bargaining game did not improve the efficiency of negotiated outcomes. 

While this evidence is quite suggestive, the experiments are still stylized. For example, in reality, negotiations often involve many issues over which participants usually have diverse preferences, about which their counterparts have incomplete information. In a recent paper, Brañas et al. (2018), we try to address some of those shortcomings. 

One obstacle to studying the impact of social interaction on negotiations is that these interactions are complicated processes. The business meal preceding a negotiation involves communication and other aspects, such as food and beverage intake. For this reason, we designed an experiment that would allow us to distinguish the effect of different factors on the negotiations. 

We conducted experiments with MBA students at the Burgundy School of Business in Dijon. This is an elite business school in the Bourgogne region of France. The students were recruited for a wine-tasting activity followed (or preceded, in the case of the control) by negotiation. After gathering, they read the experimental instructions and then had 30 minutes to interact (except in the control treatment, which had no interaction). During the interaction phase, there were five treatments besides the control: interaction only; water; wine; food and water; and food and wine. We made an extra effort to make the setting as natural as possible. For instance, we allocated participants and desks to circles of four, facing each other. We also divided the participants into all-male and all-female groups, to avoid complications arising from inter-gender issues in negotiation.

After the interaction phase, participants took part in a negotiation simulation common in negotiation classes. In this simulation, they negotiate over a labour contract with many attributes, each of which carries a different number of points for each possible agreement. Participants are paid as a function of their total points, but they know only their own points. This information asymmetry creates an opportunity for social interaction to increase trust and efficiency. There were two treatments in the negotiation phase: one involved hard negotiation (five issues) and the other, simple negotiation (two issues). 

We run many treatments because our assumption was that some form of pre-play interaction would improve negotiations and we wanted to find out the (possibly synergistic) impact of the different constituents. As it turned out, nothing worked better than moving directly into negotiations. 

The first two figures below show the aggregate points in the negotiation for the hard negotiation treatments, in the all-male (Figure 1) and the all-female groups (Figure 2). Figure 3 shows the distribution of points between employers and workers for different groups in relation to the Pareto frontier. These graphs show clearly that pre-play interactions do not improve aggregate results or the distribution of points between the negotiators.

Figure 1 Hard negotiation: Men

Figure 2 Hard negotiation: Women

Figure 3 Point distribution between employers and workers in relation to Pareto frontier

These results indicate that pre-play social interaction is probably similar to a consumption good. It does not appear to have a value in production, which is what would justify the tax exemption. In the words of HMRC, the part of the expense that is wholly and exclusively for the purpose of generating profits is, on the basis of our experiment, a rather precisely estimated zero. As such, its tax status might need to be revised if future research on this topic confirms our results.


Brañas-Garza, P, A Cabrales, G Mateu, A Sánchez and A Sutan (2018), “Does pre-play social interaction improve negotiation outcomes?”, CEPR Discussion Paper 13417.

Brooks, A W, and M E Schweitzer (2011), “Can Nervous Nelly negotiate? How anxiety causes negotiators to make low first offers, exit early, and earn less profit”, Organizational Behavior and Human Decision Processes 115(1): 43–54.

Charness, G (2000), “Self-serving cheap talk: A test of Aumann’s conjecture”, Games and Economic Behavior 33(2): 177–194.

Charness, G, and M Dufwenberg (2006), “Promises and partnership”, Econometrica 74: 1579–1601.

Clark, K, S Kay and M Sefton (2001), “When are Nash equilibria self-enforcing? An experimental analysis”, International Journal of Game Theory 29(4): 495–515.

Cooper, R W, D V DeJong, R Forsythe and T W Ros (1992), “Communication in coordination games”, The Quarterly Journal of Economics 107(2): 739–771.

Forsythe, R, J Kennan, and B Sopher (1991), “An experimental analysis of strikes in bargaining games with one-sided private information”, The American Economic Review 81: 253–278.

Palfrey, T R, and H Rosenthal (1991), “Testing for effects of cheap talk in a public goods game with private information”, Games and Economic Behavior 3: 183–220.

Valley, K, L Thompson, R Gibbons, and M H Bazerman (2002), “How communication improves efficiency in bargaining games”, Games and Economic Behavior 38(1):127–155.


[1] See

[2] See

[3] See

525 Reads