Dynamic strategic interactions characterise many situations of economic decision making. Modern dynamic macro models typically assume that decision makers are rational and forward looking, incorporating the consequences of future interactions into their current decisions. Likewise, many contexts studied in microeconomics involve both strategic and dynamic components. One prominent example is promotion tournaments between employees who are competing for promotion to higher ranks in organisations. Similar features are also observed in procurement tournaments with multiple rounds and shortlisting, in political campaigns with multiple stages like the US presidential election, as well as in the playoff stage of sports leagues.
While the power and usefulness of tournament models for gaining a better understanding of behaviour in strategic interactions has been obtained with static models, several contributions show that optimal behaviour in early stages of multi-stage elimination tournaments is determined by the value of participation in future stages of the tournament (e.g. Rosen 1986). Intuitively, winning the current match might have some inherent value, but it is also a prerequisite to participate in future matches of the elimination tournament. This implies that forward-looking agents will adjust their behaviour in the present if it becomes more or less valuable to participate in future stages of the tournament.
The only existing study that indirectly accounts for such future effects in strategic interactions is by Brown and Minor (2014) on tennis tournaments. They restrict attention to the outcomes of games and investigate whether the strength of the expected future competitor affects the probability that the stronger player wins in a given game. Outcomes are determined by unobserved decisions of competitors, however, and direct evidence for the prediction that future effects influence decisions of competitors is still lacking.
Testing whether competitors are forward looking or not
In a new paper (Lackner et al. 2015), we try to open the black box of how outcomes are achieved and use data from the playoffs of sports tournaments to investigate whether and when teams are forward looking. In particular, we consider playoff tournaments from the National Basketball Association (NBA). We have access to data for 30 NBA seasons from 1983/84 through 2013/14, covering a total of 2,199 individual playoff games. Each NBA playoff tournament is structured in four stages, and two teams compete in best-of-seven match-ups in each stage, i.e. the winner is determined as the team that wins four games. To investigate whether the institutional settings (best-of-seven, home versus away games) matter for our findings, we collected data for NCAA tournaments for ten seasons from 2003 through 2013 where each stage involves only one game between two teams (i.e. a best-of-one winning rule) and the game takes place on neutral ground.
These data provide an ideal setting for the purpose of our study. On the one hand, these tournaments involve considerable stakes. On the other hand, they provide precise information about all required elements such as heterogeneity between teams, the intensity of play by each team, and the outcomes of each game. In addition, both datasets also include information on performance in the regular season preceding the playoffs that can serve as a measure of strength. Finally, the differences between the NBA and NCAA tournament rules allow us to cross-validate the empirical findings.
Based on the intuition that it is more valuable to participate in the next stage of the tournament if the future opponent team is weak and thus likely to lose the game on the next stage than if the future opponent team is strong and thus likely to win, we investigate whether the intensity of play in the current stage is affected by the ability of their future opponent in next stage of the tournament. To measure whether teams defend more or less aggressively in a particular playoff game, we use the number of personal fouls a team is called for. In order to account for ability and team-specific styles of play, we adjust the number of observed fouls by the number of fouls a team recorded during the entire regular season preceding the playoffs.
In principle, fouls measure how intensely the defender attacks his opponent, how physically close he is in coverage, which may sometimes result in a personal foul. Fouls are not necessarily committed on purpose, as it is presumably more likely that players try to avoid fouls in most instances, but are still more likely to foul the opponent when defending intensely. In that sense, personal fouls are an almost natural outcome of an intense game with close physical contact. The higher the intensity of play, the higher the probability that a foul is inadvertently committed and called, implying that fouls are highly correlated with the effort provided.
Our findings support the view that decision makers are indeed forward looking.
- In particular, our main results show that, everything else equal, the expectation of a weaker future opponent increases the intensity of play in the current interaction.
- In addition, we find that the intensity of play is negatively affected by the ability of the current opponent, consistent with the theoretical prediction and previous evidence based on static interactions.
Additional analyses indicate that teams only accommodate their behaviour to the strength of their future opponents so long as they can ‘afford’ to do so. In particular, we find that the effect for favourites is robust and highly significant across different specifications, while the effect for underdogs is sometimes close to zero. The probability that favourite teams actually reach the next stage is high, implying that forward-looking behaviour is relatively more prevalent for these teams. At the same time, underdog teams have a comparably high probability of losing the series anyway, implying that they cannot afford to slack off anytime, independent of their future opponents. In line with this argument, our results also suggest that even favourite teams focus entirely on their immediate opponent in pivotal games where they can lose the series.
When disaggregating the data by the stages of the tournament, we find that the strategic aspect implied by the expectation of a weaker future opponent becomes more important the shorter the remaining tournament, i.e. the closer the final interaction in the tournament. One candidate explanation for this pattern is that the expectation of a weak opponent in the next round increases the chances of winning the championship in any stage, but more so the fewer the number rounds to go and the less uncertainty involved regarding the ability of future opponents.
Our findings provide evidence from the field for forward-looking behaviour of decision makers in strategic interactions. The heterogeneity of results (e.g. across favourites and underdogs) raises the question about the size, the subjective determinants, and the salience of future interactions that induces forward-looking behaviour. Moreover, our analysis considers a setting where the next stage of the tournament is in the very near future, while the competition for future promotions in organisations might take several years. A systematic analysis of these issues provides a logical next step and a promising avenue for future research.
Brown, J, and D Minor (2014), “Selecting the Best: Effort Spillover and Shadows in Elimination Tournaments”, Management Science, 60, 3087 – 3102.
Lackner, M, R Stracke, U Sunde, and R Winter-Ebmer (2015) “Are Competitors Forward Looking in Strategic Interactions? Evidence from the Field”, CEPR Discussion Paper No. 10977.
Rosen, S (1986). “Prizes and Incentives in Elimination Tournaments”, American Economic Review, 76, 701 – 715.